feat: 新增 API（capabilities / 退群监控 / 自动化 / 群统计）

feat: 持久化聊天记录导出设置
style: 修复图标样式 #46
2026-10-05 04:20:34 +08:00 · 2026-10-04 02:31:34 +07:00 · 2026-10-02 19:37:09 +07:00 · 2026-10-01 17:35:47 +07:00 · 2026-09-29 18:02:09 +07:00 · 2026-09-29 16:24:20 +07:00
223 changed files with 33129 additions and 9252 deletions
@@ -10,7 +10,8 @@ out
 coverage/
 playwright-report/
 test-results/
-resources/connectors/wechat-personal/
+resources/runtime/darwin-arm64
+.native-runtime-source
 .omc
 .codex/
 skills-lock.json
@@ -25,16 +25,14 @@
 </p>

 <p align="center">
-  <img src="./public/日报.png" alt="TraceMemo 主界面" />
+  <img src="./public/日报.png" alt="TraceMemo 日报" />
 </p>

 <p align="center">
-  <img src="./public/问问微信.png" alt="TraceMemo 问问微信" />
+  <img src="./public/自动化.png" alt="TraceMemo 自动化" />
 </p>

-<p align="center">
-  <img src="./public/退群监控.png" alt="TraceMemo 退群监控" />
-</p>
+
 ---

 ## 🎨 社区日报模板
@@ -49,7 +47,7 @@ TraceMemo 日报除了内置版式，也支持从社区模板市场安装更多

 在 TraceMemo 中打开：

-**日报 → 今日日报 → 日报模板 → 模板市场**
+**日报 → 社区模板市场**

 即可查看、预览、安装和切换已发布的社区模板。

@@ -67,18 +65,21 @@ TraceMemo（迹忆）原名 **WechatExplorer** 是一款本地优先的微信数

 ## 核心能力

- 💬 **聊天档案与搜索**：浏览会话，按关键词或身份信息查找。
+- 💬 **聊天档案与搜索**：浏览会话，按关键词、备注、昵称或 wxid 查找消息。
 - 🔍 **AI Search / 问问微信**：用自然语言找回模糊记忆，并查看来源。
- 🧠 **本地知识库**：建立索引，提升跨会话查询稳定性。
- 📊 **群聊日报**：生成今日、昨日或近 7 天的群聊总结。
- 👀 **群成员变化监控**：记录指定群聊的退群动态。
- 🔊 **文字转语音**：生成语音，试听后发送到选定会话。
- 🤖 **Agent Hub**：在微信里调用本机 TraceMemo。
- 🔌 **外部 Agent / Local HTTP API**：让外部 Agent 查询本机微信历史。
+- 🧠 **本地知识库**：在本机建立索引，让跨会话、跨时间的查询更稳定。
+- 🖼️ **图片文字索引**：在本机识别微信图片里的文字（截图、公告、报价图），识别结果可以在搜索和「问问微信」里被检索。识别全程不联网，原始图片不会因为本地识别而上传。
+- 📊 **群聊日报**：生成今日、昨日或近 7 天的群聊总结，可保存为 HTML 与 PNG。
+- 🗣️ **群发言统计**：统计群成员的发言量和沉默成员，看清一个群里谁在说、谁一直没说。
+- 👀 **退群监控**：用成员快照对比记录群成员退出事件，支持多群与事件历史。
+- ⚙️ **自动化**：把上面几步按规则串起来——定时生成并发送日报、成员退群时发送通知；能发到哪里取决于当前的发送能力。
+- 🔊 **文字转语音**：把文字生成语音，试听后发送到当前会话。
+- 🤖 **Agent Hub**：在微信里向本机 TraceMemo 提问。
+- 🔌 **外部 Agent / Local HTTP API**：让 Codex 等外部 Agent 查询本机微信历史。

 ## 💻 平台支持

-TraceMemo 2.4.0 支持：
+TraceMemo 2.5.0 支持：

 - **Windows x64**
 - **macOS Apple Silicon（M 系列 / arm64）**
@@ -86,6 +87,12 @@ TraceMemo 2.4.0 支持：

 Windows 与 macOS 均支持微信本地数据库连接与数据库 Key 获取。

+### 关于“发送能力”
+
+浏览、搜索、日报生成、导出、知识库和图片文字索引都不需要额外的发送组件。只有**把内容真正发回微信**这一步——自动发送日报、退群通知、把语音发到会话——依赖本机发送能力：
+
+发送能力未就绪、未绑定或发送失败时，报告本身仍会正常生成并保存在本机，执行记录会显示为“已生成，但未发送”或“已生成，发送失败”，可以稍后重试。
+
 ## 项目缘起

 <details>
@@ -138,11 +145,14 @@ TraceMemo 最早叫 **WechatExplorer**。
 | 想做什么                           | 使用入口                      |
 | ---------------------------------- | ----------------------------- |
 | 找记得原文或关键词的消息           | 档案搜索                      |
-| 找记得大意、但不知道在哪聊过的内容 | AI Search / 问问微信          |
+| 找记得大意、但不知道在哪聊过的内容 | 问问微信（AI Search）         |
+| 找到截图、公告图里写过的文字       | 问问微信 → 图片文字索引       |
 | 长期跨群查询历史                   | 本地知识库                    |
-| 了解一个群今天或近 7 天聊了什么    | 群聊日报                      |
+| 了解一个群今天或近 7 天聊了什么    | 日报                          |
+| 看群里谁最活跃、谁一直没说话       | 档案 → 群聊 → 群发言统计      |
 | 持续关注群成员退出                 | 退群监控                      |
-| 按计划生成并发送群聊日报           | 定时日报                      |
+| 按计划自动生成并发送群聊日报       | 自动化                        |
+| 成员退群时自动发一条通知           | 自动化 → 退群通知             |
 | 把文字生成微信语音                 | 文字转语音                    |
 | 在微信里向本机 TraceMemo 提问      | Agent Hub                     |
 | 让 Codex 等工具查询微信历史        | Reader Skill / Local HTTP API |
@@ -159,9 +169,9 @@ TraceMemo 最早叫 **WechatExplorer**。

 ## 文档

- [用户指南](./docs/README.md#用户指南)
+- [用户指南](./docs/README.md#档案与搜索)
 - [AI / Knowledge](./docs/README.md#ai-与知识库)
- [Monitor / Automation](./docs/README.md#日报与自动化)
+- [日报与自动化](./docs/README.md#日报与自动化)
 - [Agent / API](./docs/README.md#agent--api)
 - [开发文档](./docs/development/overview.md)
 - [隐私与安全](./docs/user-guide/privacy.md)
@@ -213,7 +223,7 @@ TraceMemo 在早期适配微信 4.x 时，曾参考 **[WeFlow](https://github.co

 这个项目起初只是一个一时兴起的项目，所以它大概也不会有一份特别严肃的产品路线图。

-我可能会按照自己的兴趣继续折腾，也可能突然加入一些奇奇怪怪、但觉得有意思的功能—— 比如让AI给某个好友, 某个群发一个语音条(逗逗群友) 或者定时生成群聊日报并做成微信卡片。
+我可能会按照自己的兴趣继续折腾，也可能突然加入一些奇奇怪怪、但觉得有意思的功能

 也因此，这个项目随时可能继续折腾，也可能因为其他事情暂时搁置。如果你有想要的功能，可以提Issue；如果觉得现有实现不符合你的需求，也欢迎直接 Fork 后自己改。

@@ -6,29 +6,29 @@

 - [第一次使用](./user-guide/getting-started.md)：安装、连接微信并完成第一次搜索。
 - [Intel Mac 获取微信密钥](./user-guide/intel-mac-key.md)：按页面检查结果准备环境并获取密钥。
- [聊天档案与搜索](./user-guide/chat-archive.md)：浏览联系人和群聊，按关键词、备注、昵称、微信号或 wxid 查找消息；也包含档案中的文字转语音入口。
+- [聊天档案与搜索](./user-guide/chat-archive.md)：浏览联系人和群聊，按关键词、备注、昵称、微信号或 wxid 查找消息；也包含档案中的文字转语音入口，以及群聊里的「群发言统计」。

 ## AI 与知识库

 - [AI Search / 问问微信](./user-guide/ai-search.md)：用自然语言找回记得大意、但不知道在哪个会话里的内容，并查看 Evidence、Citation 和 Search Trace。
- [本地知识库](./user-guide/knowledge.md)：主动建立本地索引，提升跨会话、跨时间查询的稳定性。
+- [本地知识库](./user-guide/knowledge.md)：主动建立本地索引，提升跨会话、跨时间查询的稳定性；也包括在本机识别图片文字、让截图和公告图变得可搜索的「图片文字索引」。
 - [如何核对 AI 的回答来源](./concepts/answer-sources.md)：从来源回到原始消息，检查上下文和覆盖范围。
 - [从微信数据到回答、日报和导出](./concepts/how-it-works.md)：了解哪些步骤在本机完成，哪些 AI 功能可能调用 Provider。

 ## 日报与自动化

- [群聊日报](./user-guide/report.md)：手动生成今日、昨日或近 7 天的群聊报告，也可以创建定时日报。
- 定时日报会依次生成报告、保存 Report History，再按当前微信发送能力尝试通知；发送失败时可复用已有 PNG 重试。
+- [群聊日报](./user-guide/report.md)：手动生成今日、昨日或近 7 天的群聊报告，也可以在「自动化」里创建定时日报。
+- 「自动化」按三类规则执行：**@我生成日报**、**定时日报**、**退群通知**。定时日报会依次生成报告、保存 Report History，再按当前微信发送能力尝试通知；发送失败时可复用已有 PNG 重试。
 - 自动发送和监控动作通过统一执行边界，并保留执行记录；简要说明见[产品工作方式](./concepts/how-it-works.md#动作执行与审计)。

-## Monitor
+## 退群监控

-退群监控会比较当前成员与上一份有效快照，记录成员退出事件。它支持多群、Last Good Snapshot 和事件历史；监控关闭期间的变化不会在重新开启后补报。工作方式见[产品工作方式](./concepts/how-it-works.md#退群监控)。
+退群监控会比较当前成员与上一份有效快照，记录成员退出事件。它支持多群、Last Good Snapshot 和事件历史；监控关闭期间的变化不会在重新开启后补报。成员退出同时是「自动化 → 退群通知」的触发条件。工作方式见[产品工作方式](./concepts/how-it-works.md#退群监控)。

 ## 语音能力

 - [语音转文字](./user-guide/voice.md)：在本机转写微信语音，结果可用于搜索、Knowledge 和导出。
- [聊天档案与搜索](./user-guide/chat-archive.md#文字转语音)：把文字生成微信语音，试听后发送到当前联系人或群聊。
+- [聊天档案与搜索](./user-guide/chat-archive.md#文字转语音)：把文字生成微信语音，试听后发送到当前联系人或群聊；实际发送依赖本机发送能力。

 ## Agent / API

@@ -61,7 +61,6 @@ Agent Hub 让微信机器人调用本机 TraceMemo；Reader Skill / Local HTTP A
 - [实验性：自托管微信分享卡片](./deployment/experimental-wechat-share-card.md)
 - [微信分享卡片自动部署 Skill](./skill/setup-wechat-share-card/SKILL.md)
 - [TraceMemo Reader Skill 文件](./skill/tracememo-reader/SKILL.md)
- [第三方组件说明](./third-party/wechat-chatter/NOTICE.md)

 ## 版本说明

@@ -21,14 +21,15 @@ Agent Hub 是 TraceMemo 内置的微信机器人入口，也是应用一级导
 - “帮我看看最近跟某人聊了些什么。”
 - “生成产品交流群今天的群聊总结图片。”

-当前已实现的实时任务包括：
+Hub 把入站文字分成两类处理。

- 查看最近会话（数量限制为 1–20）；
- 查询你和某位联系人的近期聊天；
- 用已配置的 AI 总结你和某位联系人近 7 天的聊天；
- 生成今天、昨天或近 7 天的群聊总结图片；
- 总结指定群成员在群里的近期发言；
- 对不需要读取聊天的普通文字请求返回简短 AI 回复。
+**确定性的快捷动作**（不经过模型，命中就执行）：
+
+- 查看最近会话：数量限制为 1–20；
+- 生成群聊总结图片：今天、昨天或近 7 天（需要同时提到“群”和“图片 / 长图 / 日报 / 报告”）；
+- 分析某个群成员的近期发言：可以指定“今天 / 昨天 / 最近 N 天”。
+
+**其余问题**交给本机的 Query Agent：它和桌面端“问问微信”使用的是同一个实现，可以按需读取联系人、会话和时间范围来回答，必要时调用你在“设置 → AI 模型”里配置的 AI。例如“帮我看看最近跟某人聊了些什么”“上个月讨论过的项目地址在哪里”。

 任务完成后，回复会发送回触发这次请求的微信用户。群聊总结会先发送进度提示，完成后发送图片。

@@ -56,12 +57,12 @@ Agent Hub 页面会记录系统、Agent Hub 和微信连接器日志。日志支

 ## 安全与边界

- Hub 使用本机通信，不把数据库直接暴露到公网；
+- Hub 在主进程内运行，不开放本地监听端口；它不会把数据库暴露到公网；
 - 机器人账号和个人微信账号是不同的登录边界，请确认你连接的是正确账号；
 - 机器人回复会发送给当前发消息的人；开发者 API 另有受保护的测试发送入口，使用前必须确认接收者；
- Hub 生成群聊总结时仍可能调用你配置的 AI Provider；
- 当前实时自然语言入口主要处理文字消息。底层连接器可以接收图片、语音、文件和视频，但 Agent Hub 尚未为这些媒体提供同等的实时意图处理；
- 当前没有实现群发、广播、定时任务或通用自主操作微信；
+- Hub 理解请求或生成总结时，会调用你在“设置 → AI 模型”配置的 Provider；
+- 当前实时入口只处理文字消息。连接器会归一化收到的消息条目，但 Agent Hub 只把文本条目当作意图处理，尚未为图片、语音、文件和视频提供同等能力；
+- 当前没有实现群发、广播、定时任务或通用自主操作微信（定时日报属于「自动化」，不是 Agent Hub）；
 - 本页面的“Agent Hub 状态”可以通过 Local HTTP API 查询，但不要把它误认为外部 Agent 的实时消息订阅接口或 MCP Server。

 ## 无法连接时
@@ -4,9 +4,11 @@

 TraceMemo 的本地 API 默认监听 `127.0.0.1:6131`。它面向同一台电脑上的 API Center、Reader Skill、CLI 和 Agent，不是公网网关，也不是带用户账户和细粒度权限 Scope 的服务。

+它同时包含**写入型**端点：生成报告并渲染 PNG（`/report`）、通过已连接机器人发送微信消息（`/agent/send`）、创建/修改/删除/启停定时日报任务并触发立即执行（`/scheduled-reports*`）。因此这个 Token 相当于本机敏感凭据，而不是一个只读查询键。
+
 ## Bearer Token

-新 Agent 配置使用 `TRACEMEMO_API_TOKEN`。v2.2.0 仍兼容读取历史变量 `WECHATEXPLORER_API_TOKEN`，优先级为新变量高于旧变量。
+新 Agent 配置使用 `TRACEMEMO_API_TOKEN`。历史变量名 `WECHATEXPLORER_API_TOKEN` 仍被兼容读取，优先级为新变量高于旧变量；当前没有设定旧变量名的移除时间，新配置不要再使用它。

 - `/api/v1/health` 是公开健康检查；
 - 其他所有端点都要求 `Authorization: Bearer <TOKEN>`；
@@ -14,7 +16,8 @@ TraceMemo 的本地 API 默认监听 `127.0.0.1:6131`。它面向同一台电脑
 - Token 由 Electron `safeStorage` 加密保存在用户数据目录的 `local-api-token.bin`；
 - 文件权限设置为 `0600`；
 - 在“API Center”中可以显示、复制和重新生成；
- 重新生成后旧 Token 立即失效。
+- 重新生成后旧 Token 立即失效；
+- 服务端只认这个 Token，**不接受用环境变量覆盖**——Agent 一侧的环境变量只是把 Token 交给 Agent 自己的方式，不是鉴权来源。

 应用不会自动把 Token 写入 Codex、Claude Code、OpenClaw 或其他 Agent 配置。请把它放进 Agent 自己的本地 secret/environment，例如：

@@ -8,7 +8,9 @@
 - API 前缀：`/api/v1`
 - 默认只监听 loopback；不要把它当作公网服务。
 - `/api/v1/health` 无需 Token；其他端点需要 `Authorization: Bearer <TOKEN>`。
- 请求体使用 JSON；响应为 JSON。
+- 请求体使用 JSON，单个请求体最大 `1 MiB`；超限返回 `413`。
+- 错误响应包含 `requestId`，响应头包含 `X-Request-Id`。客户端可传入 1-128 位的 `[A-Za-z0-9._:-]` 标识，否则服务端会生成 UUID。
+- 不支持的 HTTP method 返回 `405` 和 `Allow` 响应头。

 ## 最小请求

@@ -24,31 +26,168 @@ curl -H "Authorization: Bearer $TRACEMEMO_API_TOKEN" \

 不要把 Token 放入 URL、Skill 文件、仓库或命令历史可被共享的脚本中。

-新配置必须优先使用 `TRACEMEMO_API_TOKEN`。已安装的旧 Reader Skill 可在 v2.2.0 兼容期内继续读取 `WECHATEXPLORER_API_TOKEN`；如果两个变量都存在，以新变量为准。
+新配置必须优先使用 `TRACEMEMO_API_TOKEN`。应用生成的安装指令仍会提示：尚未升级的旧配置可以继续读取 `WECHATEXPLORER_API_TOKEN`，但新配置必须使用新变量名；如果两个变量都存在，以新变量为准。当前没有设定旧变量名的移除时间。
+
+Token 由应用生成并保存在本机，**不接受用环境变量覆盖**：Agent 侧的环境变量只是把 Token 传给 Agent 自己的方式，不是服务端的鉴权来源。

 ## 端点

-| 方法 | 路径                         | 作用                                   | 参数/请求体                                                     |
-| ---- | ---------------------------- | -------------------------------------- | --------------------------------------------------------------- |
-| GET  | `/api/v1/health`             | 服务与数据库健康状态                   | 无                                                              |
-| GET  | `/api/v1/current_time`       | 本机时间、时区和 Unix 时间戳           | 无                                                              |
-| GET  | `/api/v1/contact`            | 联系人和群聊列表                       | `filter`、`type=user\|group`                                    |
-| GET  | `/api/v1/chatroom`           | 群聊列表                               | `keyword`                                                       |
-| GET  | `/api/v1/recent_chat`        | 最近会话                               | `limit`，默认 50                                                |
-| GET  | `/api/v1/chatlog`            | 指定会话的聊天记录                     | 必填 `talker`；可选 `time` 或 `startTime`/`endTime`             |
-| GET  | `/api/v1/media/{mediaId}`    | 获取图片消息的二进制资源               | 原样使用 `/chatlog` 返回的 `media.url`，不要用消息 `id` 拼接     |
-| GET  | `/api/v1/group_snapshot`     | 群成员快照                             | 必填 `md5`                                                      |
-| GET  | `/api/v1/resolve`            | 将昵称、wxid 或 md5 解析为会话         | 必填 `q`                                                        |
-| POST | `/api/v1/report`             | 将结构化日报渲染为 HTML 与 PNG         | `GroupReportExportRequest` JSON                                 |
-| GET  | `/api/v1/agent/status`       | Agent Hub、连接器和数据库状态          | 无                                                              |
-| POST | `/api/v1/agent/group-report` | 读取群聊并生成总结图片                 | `{ "group": "群名或标识", "range": "today\|yesterday\|7days" }` |
-| POST | `/api/v1/agent/send`         | 通过已连接机器人测试发送文字或本地图片 | `{ "to": "接收者", "text": "...", "media_url": "..." }`         |
+| 方法   | 路径                                                            | 作用                                   | 参数/请求体                                                     |
+| ------ | --------------------------------------------------------------- | -------------------------------------- | --------------------------------------------------------------- |
+| GET    | `/api/v1/health`                                                | 服务与数据库健康状态                   | 无                                                              |
+| GET    | `/api/v1/current_time`                                          | 本机时间、时区和 Unix 时间戳           | 无                                                              |
+| GET    | `/api/v1/contact`                                               | 联系人和群聊列表                       | `filter`、`type=user\|group`                                    |
+| GET    | `/api/v1/chatroom`                                              | 群聊列表                               | `keyword`                                                       |
+| GET    | `/api/v1/recent_chat`                                           | 最近会话                               | `limit`，默认 50                                                |
+| GET    | `/api/v1/chatlog`                                               | 指定会话的聊天记录                     | 必填 `talker`；可选 `time` 或 `startTime`/`endTime`             |
+| GET    | `/api/v1/media/{mediaId}`                                       | 获取图片消息的二进制资源               | 原样使用 `/chatlog` 返回的 `media.url`，不要用消息 `id` 拼接    |
+| GET    | `/api/v1/group_snapshot`                                        | 群成员快照                             | 必填 `md5`                                                      |
+| GET    | `/api/v1/resolve`                                               | 将昵称、wxid 或 md5 解析为会话         | 必填 `q`                                                        |
+| POST   | `/api/v1/report`                                                | 将结构化日报渲染为 HTML 与 PNG         | `GroupReportExportRequest` JSON                                 |
+| GET    | `/api/v1/agent/status`                                          | Agent Hub、连接器和数据库状态          | 无                                                              |
+| POST   | `/api/v1/agent/group-report`                                    | 读取群聊并生成总结图片                 | `{ "group": "群名或标识", "range": "today\|yesterday\|7days" }` |
+| POST   | `/api/v1/agent/send`                                            | 通过已连接机器人测试发送文字或本地图片 | `{ "to": "接收者", "text": "...", "media_url": "..." }`         |
+| GET    | `/api/v1/wechat-personal/send-capability`                       | 个人微信发送能力状态                   | 无                                                              |
+| GET    | `/api/v1/scheduled-reports`                                     | 定时日报任务列表                       | 无                                                              |
+| POST   | `/api/v1/scheduled-reports`                                     | 创建定时日报任务                       | `ScheduledReportApiCreateRequest` JSON                          |
+| GET    | `/api/v1/scheduled-reports/{id}`                                | 查询单个定时日报任务                   | 无                                                              |
+| PATCH  | `/api/v1/scheduled-reports/{id}`                                | 修改定时日报任务                       | `ScheduledReportApiUpdateRequest` JSON                          |
+| DELETE | `/api/v1/scheduled-reports/{id}`                                | 删除定时日报任务                       | 无                                                              |
+| POST   | `/api/v1/scheduled-reports/{id}/enable`                         | 启用定时日报任务                       | 无                                                              |
+| POST   | `/api/v1/scheduled-reports/{id}/disable`                        | 暂停定时日报任务                       | 无                                                              |
+| POST   | `/api/v1/scheduled-reports/{id}/run`                            | 立即执行一次并返回 execution           | 无                                                              |
+| GET    | `/api/v1/scheduled-reports/{id}/executions`                     | 查询某个任务的执行记录                 | 无                                                              |
+| POST   | `/api/v1/scheduled-reports/executions/{executionId}/retry-send` | 兼容占位路由；当前返回 `501 not_supported` | 无                                                              |
+| GET    | `/api/v1/capabilities`                                            | TraceMemo 应用能力和可用状态             | 无                                                              |
+| GET    | `/api/v1/automations`                                             | 自动化规则列表                           | 可选 `type`、`enabled`                                           |
+| POST   | `/api/v1/automations`                                             | 创建默认停用的自动化规则                 | Automation draft JSON                                           |
+| POST   | `/api/v1/automations/validate`                                    | 校验规则，不保存、不执行                 | Automation draft JSON                                           |
+| GET    | `/api/v1/automations/{id}`                                        | 查询单条自动化规则                       | 无                                                              |
+| PATCH  | `/api/v1/automations/{id}`                                        | 更新规则配置                             | 可变配置字段 JSON                                                |
+| DELETE | `/api/v1/automations/{id}`                                        | 删除自定义规则                           | 系统内置规则受保护                                               |
+| POST   | `/api/v1/automations/{id}/enable`                                 | 校验并启用规则                           | 无                                                              |
+| POST   | `/api/v1/automations/{id}/disable`                                | 停用规则                                 | 无                                                              |
+| GET    | `/api/v1/automations/executions`                                  | 查询自动化执行记录                       | `ruleId`、`status`、`since`、`until`、`limit`                   |
+| GET    | `/api/v1/monitors/group-exits`                                     | 查看退群监控状态                         | 无                                                              |
+| PATCH  | `/api/v1/monitors/group-exits`                                     | 配置监控群范围或启停                     | `enabled`、`monitoredConversationIds`                          |
+| GET    | `/api/v1/monitors/group-exits/events`                             | 查询退群事件历史                         | `conversationId`、`since`、`until`、`limit`                    |
+| GET    | `/api/v1/groups/{conversationId}/member-stats`                     | 查询群成员活跃统计                       | 必填 `conversationId`、`start`、`end`                          |
+
+`/api/v1/query/*` 是一组结构化的 Query 端点，见下方[LLM-friendly Query Tool API](#llm-friendly-query-tool-api)。
+
+## Application Capabilities
+
+`GET /api/v1/capabilities` 描述 TraceMemo 应用级能力和当前运行环境；`GET /api/v1/query/capabilities` 只描述结构化 Query primitive，两者不是同一份目录。应用能力使用 `supported` 和 `available` 分开表示“代码支持”与“当前可用”；运行时原因使用稳定的简短 code，不返回 Token、数据库路径、微信密钥或 sender 诊断路径。
+
+响应包含应用版本、数据库 readiness、Query、Automation、退群监控、群统计，以及个人微信/iLink 的能力状态。`groupExitMonitor.operations` 当前声明 `read_state`、`configure_scope`、`enable`、`disable`、`list_events`；`groupStats.operations` 当前声明 `member_stats`。能力声明不会触发监控扫描或群统计查询。
+
+```bash
+: "${TRACEMEMO_API_TOKEN:?Set TRACEMEMO_API_TOKEN from API Center}"
+BASE="http://127.0.0.1:6131/api/v1"
+AUTH="Authorization: Bearer $TRACEMEMO_API_TOKEN"
+curl -H "$AUTH" "$BASE/capabilities"
+```
+
+## Group Exit Monitor API
+
+退群监控只负责“监测哪些群、发现了哪些退群事实”。退群后是否通知、通知到哪里以及通知模板，仍由 `leave_notification` Automation singleton 负责；修改监控范围不会隐式修改该 Automation。
+
+### 查看和配置监控
+
+```bash
+curl -H "$AUTH" "$BASE/monitors/group-exits"
+
+curl -X PATCH -H "$AUTH" -H 'Content-Type: application/json' \
+  "$BASE/monitors/group-exits" \
+  -d '{"enabled":true,"monitoredConversationIds":["123@chatroom"]}'
+```
+
+`monitoredConversationIds` 只接受当前联系人列表中精确存在的群 `roomId`（例如 `xxx@chatroom`），不接受群名、md5、个人联系人、重复或空 ID。请求至少提供 `enabled` 或 `monitoredConversationIds` 其中一个；传空数组表示清空监控范围。服务会先校验全部群，再执行一次原子配置。PATCH 返回最终完整状态。
+
+状态中的 `eventCount` 是持久化退群事件总数，`lastCheckedAt`/`lastReadAt` 为空时返回 `null`。GET 不会调用 `checkNow()`，也不会触发通知发送。
+
+### 查询退群事件
+
+```bash
+curl -G -H "$AUTH" "$BASE/monitors/group-exits/events" \
+  --data-urlencode 'conversationId=123@chatroom' \
+  --data-urlencode 'since=2026-10-01T00:00:00+07:00' \
+  --data-urlencode 'until=2026-10-02T23:59:59+07:00' \
+  --data-urlencode 'limit=50'
+```
+
+时间参数必须是带 offset 的 ISO-8601；默认 `limit=50`，最大 200。事件按 `detectedAt` 升序返回。事件 DTO 使用 `eventId`、稳定的 `conversationId` 和 `memberId`，并把时间输出为 ISO-8601；当前整体已读状态不会伪造成 event-level `read` 字段。当前未开放 clear events、markRead 或 checkNow HTTP 路由。
+
+一个典型 Agent 工作流是：先通过 `/resolve` 或 `/contact` 找到稳定群 ID，再 PATCH monitor scope；如需通知，再单独 PATCH `leave_notification` Automation，调用 `/automations/validate`，最后启用规则。
+
+## Group Member Stats API
+
+```bash
+curl -G -H "$AUTH" "$BASE/groups/123%40chatroom/member-stats" \
+  --data-urlencode 'start=2026-09-01T00:00:00+07:00' \
+  --data-urlencode 'end=2026-10-01T00:00:00+07:00'
+```
+
+`conversationId` 必须是当前联系人列表中精确存在的群 `roomId`；不存在返回 `NOT_FOUND`，个人联系人返回 `NOT_GROUP_CONVERSATION`。`start` 和 `end` 必须同时提供，且使用带 offset 的 ISO-8601，`start` 不能晚于 `end`。HTTP adapter 只负责把稳定群 ID 解析为内部 md5 并调用现有 `GroupStatsService`，不会在 HTTP 层重新统计消息。
+
+响应中的 `activeMembers` 和 `silentMembers` 都只描述当前成员名单；成员使用 `memberId`，活跃成员的 `lastMessageAt` 和 `range` 时间均为 ISO-8601。`freshness`、`complete`、`limitations` 必须原样保留，`unattributedMessages` 与 `excludedSystemMessages` 用于诊断，`firstMessageAt` 没有消息时为 `null`。当前成员统计不等于完整历史成员统计，`limitations` 表达的退群成员或未归档时段不能从文本中推导成额外的 `formerMembers`，也不会伪造 `totalMessageCount`。
+
+## Automation API
+
+`/api/v1/automations*` 是 Automation 的 canonical HTTP API，读写唯一的 `AutomationRuleStore`。它支持当前真实规则类型：`daily_report`、`scheduled_report`、`leave_notification`。本 API 不提供立即执行、重试或清理执行记录。
+
+旧 `/api/v1/scheduled-reports*` 保持兼容，不设移除日期；它是面向旧 DTO 的受限 compatibility API，不是第二份存储，也不能表示所有新的定时日报目标和配置。新的 Agent 集成应使用 `/automations`。
+
+创建和校验规则时 `enabled` 只能缺省或为 `false`。创建成功后必须调用 `/automations/{id}/enable` 才会启用。启用会重新校验当前规则；数据库未就绪、目标无法解析或配置无效时不会启用。`PATCH` 只接受规则配置字段，不可改 `ruleType`、`id`、创建/更新时间或 `enabled`；启停必须使用独立 endpoint。未知字段和未知枚举会被拒绝。
+
+会话范围优先传 `wxid`、`roomId`（如 `xxx@chatroom`）或 canonical conversation ID。唯一匹配的联系人名可被解析为稳定 ID；重名会返回 `ambiguous_contact`，不会猜测。`daily_report.conditions.conversationIds` 在对外 API 中使用稳定 ID，Store 内部仍沿用既有 md5 口径。
+
+校验请求不落盘、不发消息，也不执行规则。`valid: false` 时查看 `issues`；有效时 `normalized` 是经 ID 解析后的草稿，`effects` 描述启用后的动作，定时日报另外返回按本机时区计算的 `nextRunAt`。
+
+```json
+{
+  "name": "产品群每日日报",
+  "ruleType": "scheduled_report",
+  "scheduledReport": {
+    "schedule": { "time": "20:00" },
+    "report": {
+      "sourceConversationId": "wxid_product@chatroom",
+      "range": "today",
+      "messageTypes": ["text", "image"],
+      "templateId": "v1",
+      "memberNameMode": "groupNickname",
+      "timeoutSeconds": 300
+    },
+    "target": { "type": "file_transfer" },
+    "postfixText": ""
+  }
+}
+```
+
+执行历史只读，默认最多返回 50 条，`limit` 范围是 1-200。`since` 和 `until` 接受带时区的 ISO-8601 时间；execution 本身最多留存 200 条。`running` 记录的 `finishedAt` 为 `null`。
+
+新 Agent API 的错误格式：
+
+```json
+{
+  "error": {
+    "code": "VALIDATION_FAILED",
+    "message": "自动化规则校验失败",
+    "details": []
+  },
+  "requestId": "..."
+}
+```
+
+常见错误码包括 `UNAUTHORIZED`、`METHOD_NOT_ALLOWED`、`PAYLOAD_TOO_LARGE`、`INVALID_ARGUMENT`、`NOT_FOUND`、`NOT_GROUP_CONVERSATION`、`DATABASE_NOT_READY`、`VALIDATION_FAILED`、`SINGLETON_RULE`、`PROTECTED_RULE` 和 `PERSISTENCE_FAILED`。`leave_notification` 是固定单例：可读取、修改和启停，但不能创建第二条或删除。内置 `@我生成日报` 同样不能通过 HTTP 删除。

 ### 这些端点与实时机器人有什么关系

 - `/api/v1/agent/status` 只用于查询 Agent Hub、微信连接器和数据库状态；
 - `/api/v1/agent/group-report` 由外部 Agent 或脚本主动请求生成群聊总结图片；
 - `/api/v1/agent/send` 是受 Bearer Token 保护的开发者/测试发送入口，用于通过已经连接的机器人发送文字或本地图片；它不是任意群发能力，也不是实时消息订阅接口；
+- `/api/v1/scheduled-reports*` 会**写入**应用状态：创建、修改、删除、启停定时日报任务，以及立刻执行一次。加上 `/report` 和 `/agent/send`，这个 API 并非只读接口——拿到 Token 就能改配置、生成报告并发送微信消息，请按本机敏感凭据对待；
+- `POST /api/v1/scheduled-reports/{id}/run` 与定时触发共用同一条链路：读取群聊 → 生成报告 → 保存 Report History → 尝试发送；
 - 当前 API 没有对外暴露实时入站 webhook。微信消息由应用内部的 Agent Hub 和微信连接器接收、处理和回复。

 ## 时间查询
@@ -83,10 +222,12 @@ curl -H "$AUTH" "$BASE/chatlog?talker=技术交流群&time=2026-08-07"
 ## 响应与错误

 - `200`：请求成功；
+- `201`：定时日报任务创建成功；
 - `401`：缺少、错误或已失效的 Bearer Token；
 - `400`：参数或 JSON 请求体无效；
 - `422`：媒体标识格式错误，或目标消息不是可读取的图片（`NOT_IMAGE`）；
 - `403`：浏览器 Origin 不在允许的 loopback 列表；
+- `409`：定时日报任务重复（`error === "duplicate"`，响应里会带回已存在的任务），或群聊名称匹配到多个目标（`ambiguous_contact`）；
 - `404`：端点、会话或群聊不存在；媒体标识未登记、已过期、有歧义，或图片文件不存在（`NOT_FOUND`）。媒体请求遇到此状态时，先重新读取 `/chatlog` 并使用新的 `media.url`；若仍失败，再检查本地图片文件是否存在；
 - `503`：数据库或 Agent Hub 尚未就绪；
 - `500`：服务端处理或报告渲染失败。
@@ -154,12 +295,12 @@ curl -X POST -H "$AUTH" -H 'Content-Type: application/json' "$BASE/query/convers

 `query/messages`、`query/search`、`query/message-context` 和 `query/conversation-overview` 都接受一个可选的 `scope`，用来把检索限制在一个确定的语料边界内：

-| scope | 含义 |
-| ----- | ---- |
-| `{"kind":"all"}` | 所有可读会话（默认；省略 `scope` 等价于此） |
-| `{"kind":"groups"}` | 只搜群聊语料，**且包含群成员实际发送的消息**（不是群名称或群元数据） |
-| `{"kind":"contact","conversationId":"…"}` | 只搜该一对一会话 |
-| `{"kind":"current","conversationId":"…"}` | 只搜指定的那个会话（单聊或群聊） |
+| scope                                     | 含义                                                                 |
+| ----------------------------------------- | -------------------------------------------------------------------- |
+| `{"kind":"all"}`                          | 所有可读会话（默认；省略 `scope` 等价于此）                          |
+| `{"kind":"groups"}`                       | 只搜群聊语料，**且包含群成员实际发送的消息**（不是群名称或群元数据） |
+| `{"kind":"contact","conversationId":"…"}` | 只搜该一对一会话                                                     |
+| `{"kind":"current","conversationId":"…"}` | 只搜指定的那个会话（单聊或群聊）                                     |

 `conversationId` 是会话标识，可用 `/api/v1/resolve` 或 `/api/v1/contact` 得到。`scope` 一旦给出就是**权威边界**：`target` 落在范围之外会被拒绝（`status: "invalid_tool_arguments"`、`constraint: "target_outside_scope"`），不会静默扩大范围；范围里包含多个会话时，`query/messages` 与 `query/conversation-overview` 必须显式指定 `target`（`constraint: "target_required_for_scope"`）。

@@ -186,11 +327,11 @@ curl -X POST -H "$AUTH" -H 'Content-Type: application/json' "$BASE/query/convers

 `query/search` 依赖本地索引，而本地索引是异步建立的派生数据，可能落后于聊天数据库。因此它的响应会显式给出覆盖口径：

-| 字段 | 含义 |
-| ---- | ---- |
-| `indexLatestAt` | 索引目前覆盖到的源数据时间（epoch ms），`null` 表示无法判定 |
-| `sourceLatestAt` | 聊天数据库里最新的活跃时间（epoch ms），`null` 表示无法判定 |
-| `coverage.state` | `complete` 只在索引确实覆盖了所请求的时间范围时出现 |
+| 字段                | 含义                                                                |
+| ------------------- | ------------------------------------------------------------------- |
+| `indexLatestAt`     | 索引目前覆盖到的源数据时间（epoch ms），`null` 表示无法判定         |
+| `sourceLatestAt`    | 聊天数据库里最新的活跃时间（epoch ms），`null` 表示无法判定         |
+| `coverage.state`    | `complete` 只在索引确实覆盖了所请求的时间范围时出现                 |
 | `freshness.catchUp` | 本次为追赶索引做了什么：`none` / `reused` / `completed` / `pending` |

 调用方**必须**把 `coverage` 当真：`coverage.state` 不是 `complete` 且 `evidence` 为空时，只能说明"这段范围暂时无法确认"，**不能**下"没有找到"的结论。索引落后时服务端会自动请求一次追赶同步，但不会让请求无限等待；`freshness.catchUp` 为 `pending` 表示追赶仍在后台进行，稍后重试即可拿到更新的覆盖。
@@ -8,7 +8,7 @@ Reader Skill 是一份给 Agent 的操作说明。安装后，Codex、Claude Cod

 Reader Skill 只负责“外部 Agent 主动查询历史微信数据”。它不负责二维码登录、监听微信实时消息、接收机器人消息或管理 Agent Hub。想让机器人收到微信消息后处理并回复，请阅读[Agent Hub](./agent-hub.md)。

-正式 Reader Skill 名称和目录是 `tracememo-reader`，新安装使用 `TRACEMEMO_API_TOKEN`。已安装的旧 `wechatexplorer-reader` 可在 v2.2.0 兼容期内继续使用旧变量。
+正式 Reader Skill 名称和目录是 `tracememo-reader`，新安装使用 `TRACEMEMO_API_TOKEN`。已安装的旧 `wechatexplorer-reader` 仍可继续使用旧变量 `WECHATEXPLORER_API_TOKEN`（当前没有设定移除时间），但新安装请使用新名称与新变量名。

 ## 推荐安装流程

@@ -53,8 +53,13 @@ Reader Skill 可以指导 Agent 使用：
 - 指定会话、日期或时间戳范围的聊天记录；
 - 群成员快照；
 - 结构化日报渲染和按群聊生成总结图片；
+- 定时日报任务的查询、创建、修改、启停、删除、立即执行和执行记录；删除不可逆，Skill 要求先列出唯一任务并取得用户明确确认；
+- 个人微信发送能力状态查询（`/wechat-personal/send-capability`）；
+- `query/*` 一组结构化 Query 端点：`messages`、`search`、`message-context`、`conversation-overview`；
 - Agent Hub 状态检查与已连接机器人发送测试。这里的发送接口是开发者/测试用途，不是实时机器人入口，也不会让 Reader Skill 自动监听微信消息。

+注意这个 API 不只是只读的：`/report`、`/agent/send` 和 `/scheduled-reports*` 会写入状态或真的发出微信消息。
+
 端点、参数、错误码和鉴权细节以[Local HTTP API](./api.md)为准。Skill 文件保持短小，避免在多个文档中复制会变化的完整响应 schema。

 ## 隐私边界
@@ -4,39 +4,54 @@

 ```mermaid
 flowchart LR
-  A[本机微信数据] --> B[读取与解析]
-  B --> C[聊天档案与普通搜索]
-  B --> D[本地知识索引]
-  D --> E[筛选相关消息]
-  E --> F[用户配置的 AI Provider]
-  F --> G[回答与可核对来源]
-  B --> H[聊天导出]
-  B --> I[整理日报输入]
-  I --> F
-  F --> J[本地保存 HTML 与 PNG]
-  B --> K[Local HTTP API]
-  K --> L[外部 Agent]
-  M[微信机器人消息] --> N[Agent Hub]
-  N --> B
-  N --> F
-  B --> O[Monitor / Snapshot]
-  O --> P[Proposed Action]
-  F --> P
-  P --> Q[Policy]
-  Q --> R[Action Gateway]
-  R --> S[Personal WeChat Send Capability]
-  S --> T[Action Audit / Logs]
+  WX["本机微信数据"] --> PARSE["读取与解析"]
+
+  PARSE --> ARCHIVE["聊天档案与普通搜索"]
+  PARSE --> EXPORT["聊天导出"]
+
+  PARSE --> IDX["本机索引"]
+  IDX --> TEXTIDX["聊天记录索引"]
+  IDX --> IMGIDX["图片文字索引（本机识别）"]
+
+  TEXTIDX --> UNDERSTAND["Understand：AI Search / 问问微信"]
+  IMGIDX --> UNDERSTAND
+  UNDERSTAND --> PROVIDER["你配置的 AI Provider"]
+  PROVIDER --> ANSWER["回答与可核对来源"]
+
+  PARSE --> REPORTINPUT["整理日报输入"]
+  REPORTINPUT --> PROVIDER
+  PROVIDER --> REPORTFILE["本机保存 HTML 与 PNG"]
+
+  PARSE --> MONITOR["Monitor：退群监控 / 成员快照"]
+  MONITOR --> RULE["自动化规则"]
+  REPORTFILE --> RULE
+  RULE --> POLICY["Policy"]
+  POLICY --> GATEWAY["Action Gateway"]
+  GATEWAY --> CAP["本机发送能力"]
+  CAP --> AUDIT["执行记录与审计"]
+
+  PARSE --> API["Local HTTP API"]
+  API --> EXTAGENT["外部 Agent / Reader Skill"]
+  BOT["微信机器人消息"] --> HUB["Agent Hub"]
+  HUB --> PARSE
+  HUB --> PROVIDER
 ```

-## Remember → Understand → Monitor → Act
+## Remember → 图片文字 → Understand → Monitor → Act

 TraceMemo 的工作方式可以概括为：

 ```text
-Remember → Understand → Monitor → Act
+Remember → 图片文字 → Understand → Monitor → Act
 ```

-先读取和整理微信信息，再由 AI、Knowledge 或日报帮助理解；Monitor 负责发现成员变化，明确的业务动作再进入执行边界。回答和动作结果都应能回到来源或记录核对。
+- **Remember**：读取并解析本机微信数据，建立聊天档案、普通搜索和导出。
+- **图片文字**：在本机识别图片里的文字，把截图、公告、报价图也变成可检索的内容。这一步不联网。
+- **Understand**：Knowledge、AI Search / 问问微信、群聊日报。需要模型时，只把完成这次任务所需的受控上下文交给 Provider。
+- **Monitor**：用成员快照对比发现群成员变化，产出成员退出事件。
+- **Act**：自动化规则把前面的步骤串起来（定时日报、退群通知）；动作经过统一执行边界，并留下执行记录。
+
+回答和动作结果都应能回到来源或记录核对。

 ## 退群监控

@@ -46,7 +61,9 @@ Remember → Understand → Monitor → Act
 Current Membership → Snapshot Diff → Member Event
 ```

-上一份有效快照（Last Good Snapshot）不会被不完整读取覆盖，因此重启后仍可继续监控通知。
+上一份有效快照（Last Good Snapshot）不会被不完整读取覆盖，因此重启后仍可继续监控通知。监控关闭期间发生的变化，不会在重新开启后补报。
+
+成员退出事件同时是「自动化」里「退群通知」规则的触发条件。

 ## 动作执行与审计

@@ -58,17 +75,20 @@ Feature → Policy → Gateway → Capability → Execution → Audit

 Policy blocked 表示策略不允许，Capability unavailable 表示当前发送能力不可用，Send failed 表示已经尝试但执行失败。Action Audit / Logs 会保留执行结果；定时日报即使发送失败，也会保留已生成的报告记录。

+这些动作统一由「自动化」管理，当前有三类规则：**@我生成日报**、**定时日报**、**退群通知**。发送目标支持当前群聊、文件传输助手、自己、指定好友，不是任意群发。
+
 ## 哪些步骤在本机

 - 微信数据库读取与解析；
 - 聊天档案浏览和普通搜索；
 - Knowledge 索引与增量同步；
+- 图片文字索引：识别图片中的文字完全在本机进行，原始图片不会因为本地识别而上传；
 - 离线语音转写；
 - 聊天导出文件、日报 HTML/PNG 和本地历史记录的保存。

 ## 哪些步骤可能调用外部服务

-当你主动使用 AI Search、群聊日报或图片理解时，应用会把完成任务所需的受控问题和上下文发送给你配置的 Provider。它不会因为打开软件就自动上传完整数据库。
+当你主动使用 AI Search、群聊日报或图片理解时，应用会把完成任务所需的受控问题和上下文发送给你配置的 Provider。它不会因为打开软件就自动上传完整数据库，本机 OCR、离线语音转写和普通搜索也不会触发外发。

 Agent Hub 收到微信机器人的文字后，也可能为了理解请求或生成总结调用已配置的 Provider。Reader Skill 调用的是本机 API；外部 Agent 是否把读取结果继续交给云端模型，取决于外部 Agent 自己的配置。

@@ -76,13 +96,15 @@ Agent Hub 收到微信机器人的文字后，也可能为了理解请求或生

 ## 产品名词和用户任务的对应关系

-| 用户想做什么             | 产品中可能看到的名称         |
-| ------------------------ | ---------------------------- |
-| 让 AI 找相关聊天         | AI Search、Retrieval         |
-| 让答案能回到原消息       | Evidence、Citation           |
-| 查看 AI 查找过程         | Search Trace                 |
-| 让跨会话查找更稳定       | Knowledge、FTS 索引          |
-| 让外部 Agent 读取聊天    | Reader Skill、Local HTTP API |
-| 让微信机器人调用本机能力 | Agent Hub                    |
+| 用户想做什么                   | 产品中可能看到的名称         |
+| ------------------------------ | ---------------------------- |
+| 让 AI 找相关聊天               | AI Search、Retrieval         |
+| 让答案能回到原消息             | Evidence、Citation           |
+| 查看 AI 查找过程               | Search Trace                 |
+| 让跨会话查找更稳定             | Knowledge、FTS 索引          |
+| 搜到截图、公告图里写过的文字   | 图片文字索引、本机 OCR       |
+| 让日报、退群通知按规则自动执行 | 自动化、Policy、执行记录     |
+| 让外部 Agent 读取聊天          | Reader Skill、Local HTTP API |
+| 让微信机器人调用本机能力       | Agent Hub                    |

 先按任务使用，再在需要排查或开发集成时阅读术语。
@@ -0,0 +1,308 @@
+# TraceMemo Local HTTP API / Agent API 能力审计
+
+本报告基于当前代码、文档、IPC、Renderer 调用和相关测试做静态审计，对应当前发布应用版本 `2.5.0`；不连接真实微信数据库，也不修改生产代码、API Center 或测试。
+
+主要代码入口：[http-server.ts](../../src/main/http-server.ts)、[automation-rule-store.ts](../../src/main/services/automation-rule-store.ts)、[group-exit-monitor-service.ts](../../src/main/services/group-exit-monitor-service.ts)、[group-stats-service.ts](../../src/main/services/group-stats-service.ts)、[api.md](../agent/api.md)。
+
+## 1. Executive Summary
+
+当前 HTTP 层声明了 **30 个 Method + Path 模板**：15 个 GET、12 个 POST、1 个 PATCH、1 个 DELETE、1 个 HEAD。共享 `LOCAL_API_ENDPOINTS` 定义 **14 个测试项**，但 Renderer 的 `API_ENDPOINTS` 实际只展示 **12 项**；个人微信能力和 `GET /scheduled-reports` 虽有共享定义，界面没有展示。界面也没有结构化 Query、媒体读取、定时日报写操作、退群监控、群统计、Automation 通用资源或执行日志查询。
+
+当前不缺成熟的 Query primitive：结构化读取消息、Knowledge 搜索、消息前后文、会话概览和图片 OCR 搜索已经有 HTTP 契约。真正的能力缺口集中在 **监控配置、自动化规则通用管理、群员统计、执行历史、全局能力发现、发送状态和报告历史**。
+
+几个影响后续设计的代码事实：
+
+1. 定时日报已迁入 `AutomationRuleStore`。当前 `/scheduled-reports` 是旧 HTTP 契约的兼容投影，不是第二份活动规则存储；但它只表达“生成后发回来源群”，其他当前合法目标不会出现在该兼容 API 列表中。
+2. 退群监控只负责监测范围、快照和事件历史；退群后如何通知由唯一的 `leave_notification` Automation 规则负责。二者应分别建资源。
+3. 群统计 Service 足以提供活跃成员、当前沉默成员、时间范围、新鲜度和限制说明；它没有结构化的 former member 列表，也没有覆盖所有发言者的群总消息数。
+4. 现有微信 Action Gateway 有策略、审计和幂等骨架，但策略目前主要校验收件人，以及 Automation purpose allowlist；手动用户 purpose 默认可放行。它只发个人微信，现有 `/agent/send` 则走 iLink 的 `WechatSendGateway`。两者不能直接视为统一的 Agent 安全边界。
+5. HTTP 静态 GET handler 大多没有 method guard；对这些路径发 POST、PATCH 或 DELETE 仍会执行读处理。Automation store 的规则落盘失败会记录日志，但仍将内存中的规则返回为成功。
+
+建议先做应用级能力发现、退群监控资源、严格校验后的 Automation CRUD/执行查询；发送和“立即执行”放到有 dry-run、明确确认、幂等键和 Action 审计的后续阶段。
+
+## 2. Current API Inventory
+
+以下按 `http-server.ts` 的路由分派和动态 route factory 盘点。静态读路由的 `Method` 是当前文档和产品语义的预期方法；实际接受方法的差异见本节末尾。
+
+| Method | Path | 能力 | Read/Write/Execute | Service | API Center | 文档 | Agent 价值 |
+|---|---|---|---|---|---|---|---|
+| GET | `/api/v1/health` | HTTP 与数据库 ready 状态；唯一免 Token 路径 | Read | `isReady()` | 是 | 是 | 高：连通性 |
+| GET | `/api/v1/current_time` | 本机时间、时区、日期 | Read | JavaScript `Date` | 是 | 是 | 中：相对日期换算 |
+| GET | `/api/v1/contact` | 联系人和群列表；`filter`、`type`；使用异步 hydration | Read | `chat-service.listContactsAsync` | 是 | 是 | 高：标识发现与 resolve 前置 |
+| GET | `/api/v1/chatroom` | 群聊列表；`keyword`；使用异步 hydration | Read | `chat-service.listContactsAsync` | 是 | 是 | 高：群标识发现 |
+| GET | `/api/v1/recent_chat` | 最近会话；`limit` 默认 50 | Read | `chat-service.listRecentChat` | 是 | 是 | 高：导航/摘要 |
+| GET | `/api/v1/chatlog` | 按 talker 和时间读取原始消息；移除 `contentData.aeskey` | Read | `chat-service.listMessages`、`resolveMd5` | 是 | 是 | 高，但旧式、未做结构化分页 |
+| GET | `/api/v1/group_snapshot` | 群成员快照；必填 `md5` | Read | `chat-service.getGroupSnapshot` | 是 | 是 | 高：成员身份解析 |
+| GET | `/api/v1/resolve` | 昵称、wxid、md5 解析为会话 | Read | `chat-service.resolveMd5` | 是 | 是 | 高；新资源宜返回稳定 ID 和歧义候选 |
+| POST | `/api/v1/report` | 接收完整结构化日报并导出 HTML/PNG；当前拒绝外部 `templateRef` | Write：本地文件 | `group-report-service.exportGroupReport` | 是 | 是 | 低/中：低层渲染契约，Agent 须先拼完整结构 |
+| POST | `/api/v1/agent/group-report` | 读群消息、调用 AI 生成群总结、导出 HTML/PNG | Execute：AI/本地文件 | `agent-group-report-service.generateAgentGroupReport` | 是 | 是 | 高但有模型费用/数据出站；结果不等同于报告历史记录 |
+| GET | `/api/v1/agent/status` | Agent Hub、connector、数据 API、数据库状态 | Read | `agentHubService.getStatus` | 是 | 是 | 中：只覆盖 Agent Hub，不是应用总能力 |
+| POST | `/api/v1/agent/send` | 通过 Agent Hub 连接器发送文字或媒体 | Execute：微信发送 | `agentHubService.testSend` → `WechatSendGateway`/iLink | 是 | 是 | 高但 R2；是测试入口，不含统一 Action policy/幂等确认 |
+| GET | `/api/v1/wechat-personal/send-capability` | 个人微信 text/image/voice 能力状态 | Read | `PersonalWechatCapabilityService`，由 `ScheduledReportApiService` 包装 | 否（共享定义有，界面未展示） | 是 | 高但只代表 personal，不代表 iLink |
+| GET | `/api/v1/scheduled-reports` | 列出旧 DTO 可表达的定时日报规则 | Read | `ScheduledReportApiService.list` → 规则投影 | 否（共享定义有，界面未展示） | 是 | 高但不完整：仅 `source_chat` 目标 |
+| POST | `/api/v1/scheduled-reports` | 创建旧型定时日报，来源群即发送目标 | Write：规则配置 | `ScheduledReportApiService.create` → `AutomationRuleStore` | 否 | 是 | 高；功能受旧 DTO 限制 |
+| GET | `/api/v1/scheduled-reports/{id}` | 单条旧型日报规则投影 | Read | `ScheduledReportApiService.get` → `AutomationRuleStore` | 否 | 是 | 中/高：只支持兼容投影规则 |
+| PATCH | `/api/v1/scheduled-reports/{id}` | 修改旧型日报规则 | Write：规则配置 | `ScheduledReportApiService.update` → `AutomationRuleStore` | 否 | 是 | 高但只能改旧字段/目标 |
+| DELETE | `/api/v1/scheduled-reports/{id}` | 删除旧型日报规则 | Write：删除配置 | `ScheduledReportApiService.delete` → `AutomationRuleStore` | 否 | 是 | 中；新 API 应标 R3 并防护系统规则 |
+| POST | `/api/v1/scheduled-reports/{id}/enable` | 启用规则 | Write：配置/未来执行 | `ScheduledReportApiService.setEnabled` → `AutomationRuleStore` | 否 | 是 | 高；启用后未来可能发送微信 |
+| POST | `/api/v1/scheduled-reports/{id}/disable` | 暂停规则 | Write：配置 | `ScheduledReportApiService.setEnabled` → `AutomationRuleStore` | 否 | 是 | 高 |
+| POST | `/api/v1/scheduled-reports/{id}/run` | 手动触发完整日报规则 | Execute：可能调用 AI、保存历史、发微信 | `ScheduledReportService.runScheduledReportNow` → `AutomationService` | 否 | 是 | 高但 R2；当前无 HTTP 幂等键/确认 |
+| GET | `/api/v1/scheduled-reports/{id}/executions` | 旧型 execution 历史和新 Automation execution 投影 | Read | `ScheduledReportService.listExecutions` | 否 | 是 | 高但旧响应模型/有限留存 |
+| POST | `/api/v1/scheduled-reports/executions/{executionId}/retry-send` | 旧文档称复用 PNG 重发 | Execute 路由存在但当前固定 `501 not_supported` | `ScheduledReportApiService.retrySend` | 否 | **路径有，语义已失效** | 无：不能重发 |
+| GET | `/api/v1/media/{mediaId}` | 读取消息关联的图片二进制 | Read | `http-media-service.readImageMedia` | 否 | 是 | 高：图像证据 |
+| HEAD | `/api/v1/media/{mediaId}` | 图片资源存在性/响应头 | Read | `http-media-service.readImageMedia` | 否 | 否 | 低/中 |
+| GET | `/api/v1/query/capabilities` | Query Tool 支持的结构化操作、范围和上限 | Read | `LocalQueryApiService.capabilities` | 否 | 是（独立章节） | 高，但不是 TraceMemo 应用能力清单 |
+| POST | `/api/v1/query/messages` | 单会话、范围、时间、方向、类型等确定性消息读取 | Read | `LocalQueryApiService.messages` | 否 | 是 | 高：推荐 Query primitive |
+| POST | `/api/v1/query/search` | Knowledge 关键词检索，返回覆盖、新鲜度和 OCR 命中 | Read | `LocalQueryApiService.search` + `KnowledgeSearchService` | 否 | 是 | 高：必须读取 coverage/freshness |
+| POST | `/api/v1/query/message-context` | 通过 opaque `messageRef` 读取前后文 | Read | `LocalQueryApiService.context` | 否 | 是 | 高：稳定消息引用 |
+| POST | `/api/v1/query/conversation-overview` | 单会话范围的概览证据和 source coverage | Read | `LocalQueryApiService.overview` | 否 | 是 | 高：broad summary |
+
+路由来源：[http-server.ts](../../src/main/http-server.ts#L212)、scheduled/query/media route factory（同文件 L404-L670）。总数是代码中声明的业务方法模板，不代表静态 handler 都正确拒绝其他动词：`/health`、`/current_time`、`/contact`、`/chatroom`、`/recent_chat`、`/chatlog`、`/group_snapshot`、`/resolve`、`/agent/status` 没有检查 `req.method`。这些路径携带错误动词仍会走同一 handler；特别是 `POST /health` 也绕过 Token 检查，因为鉴权例外按 pathname 判断。新/旧路由都应 fail closed 并对不支持的方法返回 405。
+
+全局 `OPTIONS` 在路由和鉴权前处理；媒体额外支持 HEAD。非 health 路径要求 `Authorization: Bearer …`，但 `readBody` 没有大小上限。路由外层目前没有统一请求 schema、统一错误 envelope 或 request id。
+
+## 3. Internal Capability Inventory
+
+这里按产品能力追 Service → IPC/Renderer → HTTP → 外部 Agent，而不是按当前 API 名字扩展。
+
+| 业务域 | 已有能力及实现 | IPC / Renderer | HTTP 现状 | Agent 结论 |
+|---|---|---|---|---|
+| Chat / Contact | 联系人、群聊、最近会话、解析、历史消息、群快照、消息周边上下文、媒体定位；`contact-resolution-service` 可精确匹配别名并返回歧义候选 | `db:getContacts`、`db:getGroupSnapshot`、消息查询/around 等，Chat/Contact/档案 UI | 旧 Reader routes + `/query/*` + 图片 `/media/{id}` | 已有 Read API 基础完整。新配置应使用 `m_nsUsrName` 对应的 wxid/roomId 等稳定 ID，不应把昵称当长期键 |
+| Query / Knowledge | `messages`、`search`、`message-context`、`conversation-overview`；Knowledge 索引覆盖/新鲜度；Image OCR 可进入 Knowledge 搜索和证据 | `knowledge:search/getStatus/startIndex/cancelIndex`、AI Search UI、`LocalQueryToolExecutor` | 四个 Query 操作均已 HTTP 化；`query/capabilities` 只描述 Query Tool | 没有需要重做的基础 Query primitive。缺的是统一应用 capability/status、更多外层筛选和 API Center 展示 |
+| Group Analytics | 活跃/沉默当前成员、每人 messageCount/lastMessageTime、窗口时间、memberCount、unattributed/system 消息、firstMessageTime、freshness/complete/limitations；索引未新鲜时最多等待 2 秒并如实降级 | `group-stats:getMemberStats`；聊天页群统计 UI | 无 | P1 候选。当前 query 要求群 `userMd5` + epoch 毫秒。former sender 只以限制文案给出数量，没有 formerMembers 数组/结构化计数；没有覆盖 former sender 的群总消息数字段，需扩 DTO 后再承诺 |
+| Group Exit Monitor | enabled/running/nativeMonitor 状态、监控 roomId 集合、lastChecked、unread、事件历史；每次回传最多 500 条，但完整事件历史 append-only 长期保存；可立即 check、改范围、启停、筛事件、清历史、mark read | `group-exit-monitor:*`；`GroupExitMonitorWorkspace` | 无 | P0。拆成 monitor state/config、events、check。`checkNow` 可能发现事件并触发自动通知，不是纯读操作；重启监控会重建快照基线，暂停期成员变化不会补报 |
+| Automation | 实际规则类型：`daily_report`、`scheduled_report`、`leave_notification`。`daily_report` 是现有消息触发条件/动作链，不是任意流程引擎；退群通知为固定 ID singleton。规则 CRUD、enable、执行记录读写均已有 Service/Store | `automation:*`；AutomationWorkspace 有规则、日志、定时执行、启停、删除和状态 UI | 没有通用 Automation API；仅 scheduled-report 兼容映射 | P0。使用 typed rule union；不要把内部任意 draft 原样开放。规则创建当前缺省 enabled=true，未知值会被归一化成默认；需 HTTP 严格校验，先 disabled + validate，再显式 enable |
+| WeChat Send / Action | `WechatSendGateway` 有 personal/iLink transport resolution、text/image/voice/file 统一类型及 Send Log；`WechatActionGateway` 做 capability preflight、Automation purpose allowlist、Action audit、幂等和 Automation 3 秒间隔 | `wechat-personal:send`、`sendGeneratedTtsVoice`、`wechat-action-log:list`、Agent Hub connector 相关 IPC/UI | `/agent/send` 只走 Agent Hub/iLink 测试发送；个人 capability 有独立 GET；两类日志没有 HTTP | 分 transport 公布 capability；R2 send 通过经审计的业务门面，不直接暴露底层 gateway。现有 Action Gateway 还不是普适安全策略：`triggerType=user` 不按 purpose 限制；普通 `/agent/send` 没传 idempotency key，也没有 Action audit |
+| Agent Hub | status、connector login/reconnect/disconnect、notification recipient/send、logs、conversation list/detail/clear、入站 inbox retry | `agent-hub:*`；Agent Hub UI | 仅 `/agent/status` 和 `/agent/send`；无 conversation/log HTTP | status 有只读价值。对话记录包含完整收发正文；inbox 包含 context token/raw items。Connector 生命周期、登录 QR/验证码、收件箱和通知 recipient 应保持 internal |
+| Reports / Templates | 手动 report render/export；AI group report；本地 Report History list/save/update template/delete；内置/已安装模板和市场 catalog/install/uninstall | `report:*`、`report-template:*`、`report-template-market:*`；Reports 与 Template Market UI | `/report` 低层 export，`/agent/group-report` AI 生成；没有 history/template API | P1：只读 Report History 元数据/资产可分离设计。现有 `listGeneratedReports` 会读取每张 PNG 为 base64，并返回结构快照和本机绝对路径，不可原样直出。模板目录可读列入 P2；安装/卸载涉及网络与本地包写入，不宜第一批开放 |
+| Recall Archive | 后台监听撤回变化，最多按会话存归档消息/撤回记录；chat-service 将 archive merge 到历史读结果 | 没有独立 CRUD IPC；设置开关和消息渲染 | 没有独立 archive API；旧 `/chatlog` 可能随底层消息返回 `recalled` 标记；Query DTO 未声明 recalled 字段 | 不开放原始 archive 管理。后续 Query 应明确返回 `recalled`/来源，避免把已撤回归档当普通消息证据 |
+| OCR / Image Insight | System OCR 本地识别；image-text-index status/count/start/pause/resume/cancel/clear/repair；Image Insight 读/解密图片并可调用 AI Provider | `system-ocr:*`、`image-text-index:*`、`image:*` IPC；Search/Report UI | OCR 派生文本可通过 `/query/search` 得到；索引管理、单图 AI 分析无 HTTP | 已有搜索能力可用。状态可纳入 capability/status；索引删除、key/decoder 配置、任意图像 AI 分析涉及成本、私密图片和索引破坏，不列第一批 |
+| 系统 / 数据 / 其他 | account discovery、DB key 管理、数据库 connect/root 重开、设置写入、cache summary/clear、app update、voice/TTS、export/import、Reader Skill 本地安装信息 | 多组 IPC；Settings、Cache、Export、Update、Voice UI | 无相应 HTTP API | 只读脱敏运行状态可按需求列 P2；DB key/root、通用 settings patch、cache 清理、任意文件路径、更新安装、TTS synthesis 等保持 internal |
+
+### A. Chat / Contact 与 ID 语义
+
+- `/contact`、`/chatroom` 改用 `listContactsAsync`，因为 macOS Session 可能只有原始 wxid/chatroom id，需要 hydrate 显示名；`ScheduledReportApiService` 却使用同步 `listContacts()` 解析群名。稳定 `talker` 可直接解析，但名称输入在需要 hydration 的运行时可能失败/退化。这是可复用 adapter 应统一异步解析的理由。
+- ID 现在不是一个口径：旧 Query 的 `scope.conversationId`/`target` 解析为 `Contact.md5`；`group-stats` 传 `userMd5`；监控用 `roomId`（`xxx@chatroom`）；新的 scheduled automation 用 `sourceConversationId`（wxid/roomId）；老 HTTP 路由混用昵称、wxid、md5。保持已有 Reader 契约不动，新 API facade 应统一对外 canonical `conversationId`（底层当前联系人的稳定 username/wxid 或 roomId），并在 main adapter 转为服务所需 md5。名称只做 resolve，不持久化到规则。
+- `chatlog` 时间边界是 Unix 秒，Query `absolute` 内部也是秒，而 group stats IPC 是 epoch 毫秒；新 Agent 资源建议用带时区 ISO-8601 输入/输出，并在 facade 单点转换。
+- `/query/messages` 有 200 上限，`messageRef` 是 opaque 稳定引用；图片 OCR 文字和 Coverage 分开呈现。`/chatlog` 则支持旧 talker/time 风格但读取结果没有同等结构边界；作为兼容 Reader 保留，不作为新 Agent 配置/分析的默认接口。
+
+### B. Group Exit Monitor 与 Leave Notification
+
+`GroupExitMonitorService` 的真实 IPC 有 `getState`、`setEnabled`、`setGroups`、`checkNow`、`listEvents`、`clearEvents`、`markRead`。事件是群成员差异事实，包含 roomId、member wxid/name、previous/current count、detectedAt；monitor state 中 `events` 只是最近 500 条快照，`totalEventCount` 对应完整内存历史。
+
+`AutomationService.handleGroupExit` 只处理 `BUILTIN_LEAVE_NOTIFICATION_RULE_ID` 对应的 singleton 规则。规则另有 `notifyScope` / `notifyRoomIds` 二次范围、target、template。Agent 配“监控 A/B/C”需改 monitor 范围；配置通知目标/通知哪些被监控群则另改这条 leave notification automation。两者不能合并为 `/monitors/{id}/notify`。
+
+### C. Automation 与 Scheduled Report
+
+`AutomationRuleStore` 的真实方法有 `listRules/getRule/createRule/updateRule/saveLeaveNotificationRule/deleteRule/setRuleEnabled`；`AutomationExecutionLogService` 提供 `list/record/clear/countSince`。执行日志最多留存 200 条，clear 属于破坏性操作。规则在 `{userData}/automation/rules.json` 中 JSON 持久化。
+
+`scheduled-report-service.ts` 的调度来源是 `automationRuleStore.listRules()`，执行交给 `AutomationService.executeScheduledRule()`，execution 从 Automation Log 投影。迁移后的旧 `tasks.json`/`executions.json` 是只读历史存档。旧 HTTP API 通过 `ScheduledReportApiService` 转换旧 DTO；创建、修改、删除、启停最终也是读写 `AutomationRuleStore`。所以正确方案是保留兼容 facade，并建立 Automation canonical API，不要继续增加第二个 scheduled-report store。
+
+旧 facade 的限制：只列/操作可投影为 `target.type === 'wechat_group'`、且目标等于来源群的规则。如今 scheduled automation 支持 source_chat/self/file_transfer/contact，故通过新 UI 创建为文件传输助手或联系人目标的规则，会从旧 `/scheduled-reports` 列表隐藏。旧 API 输入 schema 也不能表示完整 scheduled config（成员名、消息类型、模板、timeout、postfix 等）。
+
+写 API 前还需处理 `AutomationRuleStore` 的归一化和持久化契约：未知 ruleType 会降成 `daily_report`，大部分错误枚举会静默落安全默认；缺省 enabled 是 true；`persist()` catch 写盘错误后只记 warning，Store 仍返回创建/更新后的对象。HTTP adapter 必须先 strict validate，且 Store 需要可观察的持久化结果，不能把内存态冒充成功。
+
+### D. Group Analytics 确认项
+
+`GroupStatsService.getMemberStats` 已有可直接复用的核心计算；接口具体有：
+
+- 当前群成员：`memberCount`、`activeMembers`、`silentMembers`、各活跃成员 `messageCount`/`lastMessageTime`；
+- 查询窗口：`startTime`、`endTime`（epoch ms）、`firstMessageTime`；
+- 数据完整性：`freshness = fresh|stale|unknown`、`complete`、`limitations`；
+- 诊断：`unattributedMessages`、`excludedSystemMessages`。
+
+成员名单是当前成员集合；知识库统计的 sender 不在当前集合时被排除，只在 `limitations` 中增加“另有 N 位窗口内发言者已不在当前群成员名单”。Service 不返回其身份/每人消息数，也没有 `totalMessageCount`。若 Agent 需要“前成员榜”或全群消息数，需要先扩展 Service/shared type；不能由 API adapter 从 limitation 文案反解析。
+
+## 4. API / Docs / API Center Drift
+
+| 项目 | 代码事实 | 漂移/影响 |
+|---|---|---|
+| HTTP、共享定义与界面列表 | HTTP 有 30 个 method/path 模板；`LOCAL_API_ENDPOINTS` 定义 14 项，Renderer `API_ENDPOINTS` 实际展示 12 项 | 16 个 HTTP 操作模板没有共享定义；另有 2 个已定义项（个人微信能力、定时日报列表）没有展示。界面仅呈现 12/30 项，不能作为完整 API catalog |
+| Scheduled Report 展示 | 共享定义只有 `GET /scheduled-reports`；该项本身也未进入 Renderer 列表 | POST 和 task action 不显示，GET 列表也不显示；Agent 在 API Center 里无法发现这组 API |
+| WeChat Capability 展示 | 共享定义有 `GET /wechat-personal/send-capability`；Renderer 列表未包含它 | API Center 看不到个人微信发送能力状态，用户可能误把 Agent Hub 状态当成完整发送能力 |
+| Query 展示 | Query 文档在 `api.md` 的独立 LLM-friendly 章节，Service/HTTP 实现完整 | API Center 看不到；用户可能误认为 Reader API 仍只有旧 chatlog |
+| Media 方法 | `/media/{mediaId}` 支持 GET、HEAD | 文档仅列 GET；API Center 都未列 |
+| Retry Send | 文档表称 retry-send“复用已有 PNG 重试发送” | `ScheduledReportApiService.retrySend()` 当前无条件抛 `501 not_supported`；integration/unit tests 也未覆盖 retry 路由的这项现状 |
+| Scheduled Report 完整性 | 旧 facade 只 project `source_chat` | UI 可保存的其他 scheduled target 会从旧 API list/get 隐藏；不是两份存储，但旧 API 不是 Automation API 的完整别名 |
+| Health 版本 | `/health` 固定返回 `version: "1.0.0"` | 与当前 package version `2.5.0` 不同，Agent 无法据此判断应用版本 |
+| HTTP 动词 | 九个静态 GET 语义路由无 method guard | POST/PATCH/DELETE 等也可能调用读取逻辑；`/health` 任意 method 均免 Token。测试目前未锁定统一 405 契约 |
+| 请求/错误 schema | JSON parsing 和错误形状分散：通用 `sendError`、ScheduledReport 专用 error、Query status body、业务自身 result | Agent 要写多套解析逻辑；共享 API schema 和统一错误 code 不存在 |
+| 命名 | `/contact`、`/chatroom`、`/recent_chat`、`/group_snapshot` 与 `/scheduled-reports`、`/query/*`、`/agent/*`、`/wechat-personal/*` 并存 | snake_case 旧路径、资源路径和“Agent 为业务 owner”的命名混杂；新接口不能继续沿用此漂移 |
+
+文档 [api.md](../agent/api.md#L35) 基本列出当前 HTTP 路径，Query 在后续单独说明；除 HEAD 外没有发现漏写的当前业务路径，但 retry-send 的成功语义过期。API Center 的来源是单独的 [local-api-test.ts](../../src/shared/local-api-test.ts) 和 [apiEndpoints.ts](../../src/renderer/src/features/api-center/model/apiEndpoints.ts)，没有从 HTTP route/schema 派生。测试现有 `local-api-auth` 覆盖鉴权、媒体、部分 Query 和 Agent send；`scheduled-report-api` 覆盖旧生命周期；`local-api-contact-search` 覆盖 hydrate。它们没有自动比对 HTTP route、文档、Catalog 三者，也没有覆盖全部 method guard 和 retry-send。
+
+## 5. Candidate API Matrix
+
+风险按本任务口径：R0 只读；R1 本地配置/应用状态修改；R2 微信发送、AI/provider 调用等外部副作用；R3 删除或清空不可轻易恢复的数据。R1 不代表没有后续行为：enable 一条定时规则会武装未来的 R2 执行。
+
+| Capability | 当前实现 | 当前 API | 建议 | Agent 用例 | Risk | Priority |
+|---|---|---|---|---|---|---|
+| 联系人/群/会话 resolve | Chat Service + Contact Resolution | 有旧 routes；Query 内 resolve | 保留旧路由；新 resource 返回稳定 ID、歧义候选 | 查找群并取得 roomId | R0 | P0（复用） |
+| 结构化消息/搜索/上下文/概览 | LocalQueryApiService + Knowledge | `/query/*` | 保持契约；加 route schema/catalog，后续可升级稳定 ID | 查聊天、关键词/OCR、补上下文 | R0 | P0（复用） |
+| 应用 capability discovery | 各 Service 能回答局部状态 | 无；`query/capabilities` 仅 Query Tools | 新 `GET /capabilities`，区分 supported/available/reason/operations | 发现自动化、监控、统计、发送 transport | R0 | P0 |
+| Group Exit Monitor 状态/范围 | GroupExitMonitorService | 仅 IPC | GET state + PATCH enabled/roomIds | 查看监控、监控/停止一个群 | R0/R1 | P0 |
+| Group Exit events | Monitor JSONL + listEvents | 仅 IPC | GET 带 stable roomId/time/cursor/limit | 最近 7 天谁退群 | R0 | P0 |
+| 手动检查退群 | checkNow 会扫描并触发事件 handler | 仅 IPC | 有外部通知时按 R2 操作开放，先 validate effects + confirm | 立即检查一次 | R2 | P1 |
+| Automation 规则 CRUD | AutomationRuleStore | 通用 IPC；HTTP 仅旧 scheduled facade | typed union CRUD；create disabled；validate 再 enable；保护 singleton/system rules | 创建、列出、修改、暂停自动化 | R1/R3(delete) | P0 |
+| Automation validation/dry-run | 现有编辑器 preview 分散；无通用 validator API | 无 | `POST /automations/validate`；不落盘、不发送 | 确认群、目标、模板、下次运行和能力 | R0 | P0 |
+| Automation execution history | AutomationExecutionLogService，最多 200 条 | schedule 专属旧投影 | 规范化 Automation execution 读接口；清日志不开放第一批 | 查看失败、按 rule 过滤 | R0/R3(clear) | P0 |
+| Group member stats | GroupStatsService | 仅 IPC | 按稳定 group ID + ISO window 读统计；先补 former/total 语义 | 近 30 天活跃榜 | R0 | P1 |
+| WeChat capability | personal capability service；Agent Hub status | personal GET + Agent status | 新全局 capability 含 personal/iLink 和内容能力；旧路由保留 | 检查发送当前是否可用 | R0 | P0 |
+| 手动微信发送 | WechatSendGateway + Action Gateway | `/agent/send` iLink test send | 新 send command 经受限 Action facade，强制 stable recipient、confirm、idempotency | 文件助手测试消息 | R2 | P1 |
+| Send Log / Action audit | Send Log 500 条；Action audit 500 条；IPC action-log | 无 HTTP | 分层只读分页，preview 脱敏；按 executionId/requestId 关联 | 查最近发送失败、审计规则动作 | R0 | P1 |
+| Agent Hub status | AgentHubService.getStatus | `/agent/status` + IPC | 保留 alias，新资源名归 `/agent-hub/status`，与 app capabilities 分开 | 查 Hub/connector online | R0 | P1（复用） |
+| Agent Hub 对话内容 | Conversation Store，最多 50 会话×500 条 | 仅 IPC | 默认为 Internal；若产品确认需要，另做显式 opt-in、分页/时间过滤 | 查看机器人与某人的对话 | R0（高隐私） | 不建议第一批 |
+| AI 群日报 | AgentGroupReportService + export | `/agent/group-report` | 保留兼容；未来先 validate model/range/group/data egress，再异步 job | 生成临时总结图片 | R2（provider/本地文件） | P1 |
+| 日报历史 | ReportHistory Service，IPC CRUD | 无 | 分页 metadata DTO；图片 asset 单独下载；不返回 base64/路径/完整 snapshot | 昨天生成过哪些日报 | R0 | P1 |
+| 模板列表 | Template Service + market catalog | 仅 IPC | 仅已安装模板只读列表列 P2 | 有哪些日报模板 | R0 | P2 |
+| 模板安装/删除/历史改版 | Template Service/Market + Report History | 仅 IPC | 不开放通用 HTML/路径写入；将来单独授权且保留校验 | 安装或修改模板 | R1/R3 | Maybe/P2 |
+| Recall archive 查询 | 内部 archive merge 到历史消息 | 无独立 API | 不单独开放磁盘 Archive；给 Query 增 `recalled` 来源标记 | 找被撤回消息 | R0（敏感/语义风险） | P2 |
+| Image OCR index 操作 | image-text-index service | IPC（含 clear/repair） | coverage 状态可汇入 capabilities/status；不让 Agent 操作 clear/reset | 查 OCR 覆盖 | R0/R1/R3 | P2 |
+| AI image insight | ImageInsightService 读/解密图片并调 vision provider | IPC | 不暴露任意 hash/message AI 分析，除非有成本/隐私授权 | 理解群图片 | R2 | 不建议第一批 |
+| DB key、根目录、settings、cache | 多个设置/DB/cache Service | IPC/UI | 禁止通用 settings patch / 文件路径 / DB key API；只加白名单状态字段 | 修改本机数据库、安全设置 | R1/R3 | 不建议开放 |
+| Connector 生命周期/inbox | AgentHubService + WechatInboundInbox | 仅 IPC/内部 | connector 登录、验证码、QR、inbox、context token 不对 Agent 暴露 | 重连或直接拿入站 token | R1/R2 | 不建议开放 |
+
+## 6. P0 Recommendation
+
+第一批目标是“让 Agent 能配置和核验 TraceMemo，但不意外发消息”。建议只包括：
+
+1. `GET /api/v1/capabilities`：应用级 capability，不与现有 `/query/capabilities` 合并。返回版本、DB/readiness、supported vs available、不可用原因和依赖；发送分 personal/iLink 与 text/image/voice 能力。
+2. Group Exit Monitor：读取状态、显式配置 monitored roomIds、读取历史事件。`PATCH` 只接 canonical 群 ID，拒绝不存在/非群 ID；修改范围响应明确显示后台 baseline/check 状态。`check` 先列 P1，因为它可能启动退群通知发送。
+3. Automation typed CRUD：列/读规则、创建 disabled 规则、更新、显式启停、执行历史读取。Leave Notification 仍使用固定 singleton id，不允许创建重复规则；拒绝未知字段/未知 enum，而不是靠 `normalizeRuleDraft` 静默修正。
+4. `POST /automations/validate`：验证目标群、稳定通知 recipient、模板变量、report/template 配置、send capability 和 nextRunAt；只返回 plan，不落盘、不发送。
+5. 运行状态与错误契约：一致的 405、最大 body、请求 ID、错误 envelope；这是任何新 Agent 写接口前的 foundation，不是大规模权限系统。
+
+Agent 实现示例（概念流程）：
+
+- “监控 A/B/C 退群”：resolve 三个群为 roomId → validate scope → PATCH monitor group IDs。
+- “A 群有人退出就通知文件助手”：读取 monitor scope 和 singleton leave rule → validate 类型/notify scope/target → 更新规则但保持 disabled → 用户/Agent 明确 enable。监控与 leave-notification 是两份正交配置。
+- “每天 20:00 生成产品群日报”：resolve sourceConversationId → validate scheduled rule（包含 target/transport/模板/时区/next run）→ 创建 disabled → 显式 enable。旧 `/scheduled-reports` 无法表达所有当前 config，不承担新 Agent CRUD。
+- “昨天哪些自动化失败”：读 automation executions，按本机 timezone/UTC offset 和 status 查询，不清理日志。
+
+## 7. Proposed Resource Model
+
+采用业务 capability 资源，HTTP server 只负责 transport、auth、body、统一错误；每个 domain route 调用独立的 main-process API facade/Service adapter。Facade 复用当前 Service/Store，不把 UI IPC 当 HTTP RPC 转发层。
+
+| Resource | 职责 | 现有路径的处理 |
+|---|---|---|
+| `system` | health、版本、应用级 capabilities、运行状态 | `/health` 保留；新增 `/capabilities`；`query/capabilities` 不改语义 |
+| `contacts` / `groups` | 稳定 ID 列表、resolve、群成员快照/统计 | `/contact`、`/chatroom`、`/resolve`、`/group_snapshot` 保留兼容 |
+| `query` | 消息/搜索/上下文/概览证据 | 现有 `/query/*` 保持；query ID 口径升级需兼容 reader skill |
+| `monitors` | 退群监控范围、启停、事件、显式 check | 新 `/monitors/group-exits`，与通知规则分离 |
+| `automations` | 规则 typed CRUD、validate、启停和运行 | 新 `/automations` 是 canonical HTTP resource；底层仍由 `AutomationRuleStore` 存储 |
+| `executions` | 跨 Automation/Action/Send 的只读运行视图 | 新 `/executions` read model，不合并各自写存储 |
+| `wechat` | transport capabilities、受控 send command、Send Log/Action audit | `/agent/send` 与 `/wechat-personal/send-capability` 保留兼容 |
+| `reports` | report history 元数据/asset；installed templates read-only | `/report` 与 `/agent/group-report` 保留为不同兼容操作 |
+| `agent-hub` | Hub/connector 状态；conversation API 默认 internal | `/agent/status` 保留 alias，不复用 app capability |
+| `developer` | 高级 raw request tester 与诊断 | API Center 的 tester；不作为 Agent 业务 API |
+
+共享资源原则：新 API 输入以 stable id 为主、名字只用于 resolve；所有写请求 strict validate；时间对新资源用 offset ISO-8601；分页使用 `limit` + `cursor`；成功/失败使用一个 typed envelope；不直接返回 app userData path、token、context token、AES key 或原始 transport payload。
+
+## 8. Safety Model
+
+### 当前边界
+
+- 默认监听 `127.0.0.1:6131`，但 host/port 由设置和 `api:start` 调用传入，API Center 会警告非 loopback。CORS 只允许 loopback Origin，但不带 Origin 的 curl/Agent 请求仍可用 Token；CORS 不是本地进程授权边界。
+- `/health` 公开，其余 endpoint 共用单一 Bearer Token。Token 为 32 random bytes、safeStorage 加密存储并设 `0600`，没有 read/config/send scope。拿到 token 即可读取聊天，也能建/删/启停规则、立即发送。
+- HTTP `/agent/send` 的消息进入 `WechatSendGateway`，因此有低层 Send Log；但没有 `WechatActionGateway` 的业务 Action audit/策略决策，也没有调用方 Idempotency-Key。Personal capability 路由只报告个人微信状态。
+- `WechatActionGateway` 的请求包含 `purpose`、`triggerType`、recipient、content、`idempotencyKey`、`executionId`；Automation trigger 有 purpose allowlist，sender capability 会先检查，审计记录会保存 content preview/hash。当前 `evaluateWechatActionPolicy` 对 `triggerType=user` 不做 purpose allowlist，`shouldUseAiPolicy` 固定 false；幂等只对显式 key 或历史特定 scheduled request 生效。它目前只发个人微信，不能直接替换 iLink send。
+- Action Audit 和 Send Log 各自上限 500；Automation Execution Log 上限 200。三类日志粒度不同，不是重复的同一事实。
+
+### 分级建议
+
+| 风险 | 操作 | 建议控制 |
+|---|---|---|
+| R0 | 查询聊天/成员/事件/状态/统计/日志/报告元数据 | 保留本地 Token；响应明确 scope、coverage、freshness、隐私字段 |
+| R1 | 修改监控群、Automation 配置、启停、模板设置 | typed validation、dry-run plan、显示持久化成功；enable 需确认它武装未来发送 |
+| R2 | 微信发送、立即执行日报/退群通知 check、调用 AI Provider 生成报告 | 明确 recipient 和内容/规则；`Idempotency-Key` 必填；统一 Action policy + Action audit + Send Log；重复请求回放原结果；返回发送状态 |
+| R3 | 清理退群事件、删除 report、清空执行/发送/Action 日志、清 cache/index、删规则 | 初始不开放；如以后开放，独立权限、预览计数、可恢复备份/本地确认，不接受批量 wildcard |
+
+现有 Bearer Token 不足以支撑“查询、配置、发送、清理”都对一个不受信 Agent 开放。近期不必造完整账户系统，但应先修路由 method/body/validation，提供默认只读或 disabled 配置工作流；后续可加入多个 named token + scope（`read`, `configure`, `send`, `destructive`），R2 请求确认和 durable idempotency。风险分类也需承认本地 artifact write（如 `/report` 导出）不是配置本身，可先按 R1 local-write 处理。
+
+## 9. Compatibility Plan
+
+1. 不 rename/remove 任何现有 route。Reader Skill 依赖旧 contacts/chatlog/media 路径；新 structured query 已经是更合适的 Agent query，但两者并行。
+2. 新 `/automations` 读写同一 `AutomationRuleStore`。旧 `/scheduled-reports*` 改为明确标注 Deprecated 的 compatibility adapter；维持现有 request/response shape 和 source_chat 子集，不维护第二个任务存储。响应可加 `Deprecation` header/文档说明，未定 sunset 前不返回 breaking error。
+3. `/scheduled-reports` 的投影不能假装覆盖所有 scheduled automation。旧 list/get 只反映 source_chat；新 clients 必须迁到 `/automations?type=scheduled_report`。旧 `retry-send` 保留返回 501，文档明确废弃；不能伪造已发送成功。
+4. `/agent/send` 继续表示现有 iLink/Agent Hub 测试发送。新 `/wechat/send` 必须先明确 transport/recipient schema，再通过能统一 personal+iLink 的受控 Action facade；如果不能保留旧发送语义，就将旧路径作为 adapter 而不是简单 alias。
+5. `/wechat-personal/send-capability` 保持 personal-only 兼容 view；新全局 capability 返回 transport map。应用能力 `/capabilities` 和 Query Tool 的 `/query/capabilities` 各自有清晰不同的契约。
+6. 同一 shared contract/catalog 应供 HTTP 验证、API Center、文档和 route contract tests 使用；把实际路由、文档和 API Center 三者 drift 变成测试失败，而不是发布后人工发现。
+
+## 10. Proposed Phase Plan
+
+### Phase A — HTTP Contract / Facade
+
+给现有 routes 加明确 method guard、body size limit、严格 shared request schema、统一错误 envelope/request id；补 `AutomationRuleStore` 写盘成功/失败结果；建立稳定 conversation ID adapter 和 route contract tests。保留 raw Node HTTP，不需要为第一阶段换 web framework。
+
+### Phase B — Read + Validation P0
+
+增加应用 `/capabilities`、monitor state/events、automation rules/executions 读取、`/automations/validate`、group stats adapter。monitor `check` 因可能触发 notification 暂留 P1 或先加 side-effect confirmation。把新资源路径、schema、风险元数据接进 EndpointCatalog/生成文档。
+
+### Phase C — Automation Configuration
+
+开放 disabled create、PATCH、启停和 DELETE 防护；退群通知专用 singleton upsert 接入同一规则资源；scheduled report 使用真实完整 config；旧 scheduled API 只做 compatibility projection。先做 dry-run 再允许 enable。
+
+### Phase D — Side Effects / Execution Read Model
+
+扩展一层统一 `Action` facade 支持 iLink + personal transports，并有 per-purpose policy、recipient allowlist/validation、确认语义、强幂等、Action audit 到 Send Log correlation。再开放 send/check/run；按需增加 `/executions` read projection，不合并底层日志存储。
+
+### Phase E — API Center
+
+Overview、Query、Monitors、Automations、WeChat Actions、Reports、Developer 分区；按业务任务做 schema-aware 表单/效果预览，Raw Request Tester 留在 Developer。展示 Token scope/host 范围/transport capability，而不只是固定 URL 测试器。
+
+## 11. Concrete Endpoint Proposal
+
+下表是下一阶段建议契约，不表示当前已实现。新写 API 应统一错误：`{"error":{"code":"...","message":"...","details":{...}},"requestId":"..."}`。时间使用带 offset 的 ISO-8601；接口只接受 stable IDs。
+
+| Method | Path | Request → Response | Risk | Underlying Service |
+|---|---|---|---|---|
+| GET | `/api/v1/capabilities` | 无 → app version/readiness + `query`,`groups.memberStats`,`groupExitMonitor`,`automations`、每种 `wechat.transport/content` 的 `supported/available/reason` | R0 | 新薄 facade 汇总 LocalQuery、GroupStats、Monitor、AutomationStore、personal capability、AgentHub status |
+| GET | `/api/v1/monitors/group-exits` | 无 → `{enabled,running,monitoredConversationIds,lastCheckedAt,eventCount}` | R0 | `GroupExitMonitorService.getState` |
+| PATCH | `/api/v1/monitors/group-exits` | `{enabled?,monitoredConversationIds?}` → 保存后的 state；监控 ID 必须 resolve 到现有群 | R1 | `setEnabled` / `setMonitoredRoomIds` |
+| GET | `/api/v1/monitors/group-exits/events?conversationId=&since=&until=&limit=&cursor=` | ISO 时间和稳定群 ID → `{events,nextCursor}`，事件显式 `eventId`、group/member、counts、detectedAt | R0 | `GroupExitMonitorService.listEvents`；为无 cursor 的现有 list 加稳定分页 adapter |
+| POST | `/api/v1/monitors/group-exits/check` | `{confirmSideEffects:true}` + `Idempotency-Key` → checkedAt、新事件数、notification execution refs | R2 | `checkNow`；因检查可发现事件并调用 leave-notification Automation，不应标成纯读 |
+| GET | `/api/v1/automations?type=&enabled=` | 无 → typed rules page；包括 singleton leave rule | R0 | `AutomationRuleStore.listRules` |
+| POST | `/api/v1/automations` | typed `AutomationRuleDraft`，create 默认 `enabled:false` → `{rule}` | R1 | `AutomationRuleStore.createRule`（需先强化 strict validation/persist result） |
+| GET | `/api/v1/automations/{ruleId}` | 无 → `{rule}` | R0 | `AutomationRuleStore.getRule` |
+| PATCH | `/api/v1/automations/{ruleId}` | typed partial config → `{rule}`；`ruleType` 不可变 | R1 | `AutomationRuleStore.updateRule` |
+| DELETE | `/api/v1/automations/{ruleId}` | 无 → `{deletedId}`；默认拒绝 builtin/system singleton 删除 | R3 | `AutomationRuleStore.deleteRule` + protected-id policy |
+| POST | `/api/v1/automations/validate` | typed draft → `{valid,normalizedDraft,issues,effects,resolvedTargets,nextRunAt,capabilities,validationId}`；不保存、不发送 | R0 | 新 validator adapter：contact resolve、template validator、capability services、schedule pure functions |
+| POST | `/api/v1/automations/{ruleId}/enable` | `{validationId,confirmFutureEffects:true}` → `{rule}`；validation hash 必须匹配当前规则 | R1（武装未来 R2） | `AutomationRuleStore.setRuleEnabled` + validator |
+| POST | `/api/v1/automations/{ruleId}/disable` | 无 → `{rule}` | R1 | `AutomationRuleStore.setRuleEnabled` |
+| POST | `/api/v1/automations/{ruleId}/run` | `{confirmSideEffects:true}` + `Idempotency-Key` → `{execution}` | R2 | `AutomationService.executeScheduledRule`；只对具有 `run` 语义的 rule type 开放 |
+| GET | `/api/v1/automations/executions?ruleId=&status=&trigger=&since=&until=&limit=&cursor=` | 无 → `{executions,nextCursor}` | R0 | `AutomationExecutionLogService.list`；需加 timestamp/filter/cursor adapter |
+| GET | `/api/v1/groups/{conversationId}/member-stats?start=&end=` | ISO 时间窗口 → active/silent/counts/freshness/complete/limitations；未来要 former members 则先扩 shared result | R0 | `GroupStatsService.getMemberStats`；adapter 把 canonical roomId 转 md5 |
+| GET | `/api/v1/wechat/capabilities` | 无 → personal/iLink 分 transport 状态与内容能力 | R0 | `PersonalWechatCapabilityService` + `WechatSendGateway.hasIlinkSender`/Agent Hub connector health |
+| POST | `/api/v1/wechat/send` | `{recipient:{type,id},content:{type:"text",text},confirm:true}` + 必填 `Idempotency-Key` → `{actionId,status,transport,sendLogRef}` | R2 | 扩展后的 `WechatActionGateway`/统一 Action facade；不得裸调 `WechatSendGateway.send` |
+| GET | `/api/v1/wechat/send-logs?status=&since=&limit=&cursor=` | 无 → 只读、脱敏 Send Log page | R0 | `WechatSendGateway.listSendLog` / `WechatSendLogService.list` |
+| GET | `/api/v1/wechat/action-logs?executionId=&status=&since=&limit=&cursor=` | 无 → 业务 Action 审计 page；preview 默认截短/可省略 | R0 | `WechatActionLogService.list` / `WechatActionGateway.listAuditRecords` |
+| GET | `/api/v1/executions?source=&status=&since=&until=&limit=&cursor=` | 无 → 统一只读 projection，每项保留 source/executionId/trigger/action/status/error/times/correlation IDs | R0 | 组合 `AutomationExecutionLogService`、Action audit、Send Log；不合并其存储 |
+| GET | `/api/v1/reports/history?groupId=&since=&until=&limit=&cursor=` | 无 → 仅元数据分页，不带 `generatedImage`、本地绝对路径或完整 report snapshot | R0 | `listGeneratedReports` 经 summary adapter，最好先让 Service 支持 metadata-only |
+| GET | `/api/v1/reports/history/{reportId}/image` | 无 → PNG binary | R0 | Report History 的 file resolver；仅按内部 report id 解析，不接收任意路径 |
+| GET | `/api/v1/report-templates` | 无 → 已安装模板的 stable id/name/version/available 列表 | R0 | `reportTemplateService.list` |
+| GET | `/api/v1/agent-hub/status` | 无 → hub/connector/dataApi/dbReady 的脱敏状态 | R0 | `AgentHubService.getStatus`；`/api/v1/agent/status` 保留 alias |
+
+`POST /wechat/send` 建议先只开放 text + 明确目标；image path/url、voice 和 file 类型各自扩大本机文件/外传风险，需独立 validation/permission，不从现有任意 `msg` 输入自动继承。
+
+## 12. Do Not Expose
+
+- 任意数据库 key、图片 AES key、微信登录凭据、iLink `context_token`/bot token、inbound inbox raw items；这些是密钥或未处理消息，不是业务 API。
+- 通用 `settings:set`、任意 `dbRoot`、数据库 disconnect/reopen、cache 全清、Knowledge/OCR 索引 clear/reset；配置或 destructive blast radius 远高于 Agent automation 管理需求。
+- Agent Hub QR/login/verify-code/reconnect/disconnect/connector lifecycle。它会影响进程状态/账户登录，也会暴露登录材料。
+- `WechatInboundInbox` pending/clear/complete/recordFailure 等队列控制。它是 at-least-once 消息交付内部机制，外部 ack 会破坏不丢消息语义。
+- Agent Hub 完整对话正文默认不暴露。若明确产品需要，独立设计用户授权、最小时间窗口、分页、清理和脱敏；现有 Conversation Store 为本机完整收发留档。
+- 任意本地 file path、path traversal 类导出/报告资产操作；不要把 IPC 的 file chooser、reveal、delete path 形状直接变 HTTP 参数。
+- 单图 AI insight/任意 Vision analyze、批量 TTS synthesis、AI Provider 配置/测试和 app update download/install；它们具有费用、敏感图片/文本上传或应用安装影响。
+- 清除退群、Automation、发送/Action、报告历史或批量发送接口作为 P0。清理类不是“管理配置”的必要前提，应使用 R3 独立权限及本地可恢复流程。
+- 通用“任意 Automation DAG/任意脚本/任意 purpose”的创建。当前实现只有明确的三类规则和固定动作，不是 workflow engine；扩展功能应有显式 ruleType/schema，而不是暴露内存对象。
+
+## 13. Open Questions
+
+以下属于产品边界选择，无法只从代码决定：
+
+1. 新 Agent API 是否允许同时操作个人微信和 Agent Hub/iLink，还是第一版限定一个 transport？现有 capability 与 send endpoint 分属两套连接状态。
+2. Agent 是否默认只拿只读 scope；配置和 R2 send 是否要求独立 token/用户确认？当前只有单一 Bearer Token。
+3. 多账号是否属于本阶段？当前 Query/GroupExit 绑定当前活动数据库账号，Agent Hub 有自己的 connector accountId；没有统一 account-scoped API model。
+4. Agent Hub 完整对话是否要作为 API capability？它含完整消息正文，和从微信数据库按 query 搜历史是不同隐私边界。
+
+真实微信 runtime 可用性、平台 hydration 和 transport 能力需在后续实现集成测试中验证；静态代码审计无法替代真实数据库/微信连接的运行时验证。
@@ -33,17 +33,24 @@ pnpm test:e2e:build

 ## 代码变更对应文档

-| 代码区域                                                  | 需要同步检查的文档                                      |
-| --------------------------------------------------------- | ------------------------------------------------------- |
-| `src/shared/ai-search.ts`、AI Search pipeline             | `user-guide/ai-search.md`、`concepts/answer-sources.md` |
-| `src/shared/knowledge.ts`、`src/main/knowledge/`          | `user-guide/knowledge.md`、`concepts/how-it-works.md`   |
-| `src/shared/voice-recognition.ts`                         | `user-guide/voice.md`                                   |
-| `src/shared/group-report.ts`、报告 UI                     | `user-guide/report.md`、API/Agent 文档                  |
-| `src/shared/export.ts`、导出服务/UI                       | `user-guide/export.md`                                  |
-| `src/main/services/recall-archive-service.ts`             | `user-guide/privacy.md`                                 |
-| `src/shared/local-api-test.ts`、`src/main/http-server.ts` | `agent/api.md`、`api-security.md`、打包 Skill           |
-| Agent Hub service/UI                                      | `agent/agent-hub.md`、`user-guide/privacy.md`           |
-| 设置导航、连接页面                                        | `user-guide/getting-started.md`、`docs/README.md`       |
+| 代码区域                                                                 | 需要同步检查的文档                                                             |
+| ------------------------------------------------------------------------ | ------------------------------------------------------------------------------ |
+| `src/shared/ai-search.ts`、AI Search pipeline                            | `user-guide/ai-search.md`、`concepts/answer-sources.md`                        |
+| `src/shared/knowledge.ts`、`src/main/knowledge/`                         | `user-guide/knowledge.md`、`concepts/how-it-works.md`                          |
+| `src/shared/voice-recognition.ts`                                        | `user-guide/voice.md`                                                          |
+| `src/shared/group-report.ts`、报告 UI                                    | `user-guide/report.md`、API/Agent 文档                                         |
+| `src/shared/export.ts`、导出服务/UI                                      | `user-guide/export.md`                                                         |
+| `src/main/services/system-ocr-service.ts`、`image-text-index-service.ts` | `user-guide/knowledge.md`、`concepts/how-it-works.md`、`user-guide/privacy.md` |
+| `src/main/services/image-insight-service.ts`、AI Provider                | `user-guide/report.md`、`user-guide/privacy.md`                                |
+| `src/shared/automation.ts`、自动化服务与执行网关                         | `user-guide/report.md`、`concepts/how-it-works.md`、`docs/README.md`           |
+| `src/shared/local-api-test.ts`、`src/main/http-server.ts`                | `agent/api.md`、`api-security.md`、打包 Skill                                  |
+| Agent Hub service/UI                                                     | `agent/agent-hub.md`、`user-guide/privacy.md`                                  |
+| 设置导航、连接页面                                                       | `user-guide/getting-started.md`、`docs/README.md`                              |
+
+两条容易被写错的边界：
+
+- **防撤回已下线**（`src/main/services/recall-archive-service.ts` 保留但不再启动）：设置入口隐藏，`recallProtectionEnabled` 在所有读写路径上被强制收敛为 `false`。不要把它写回用户指南。
+- **图片文字索引（本机 OCR）与图片理解（需要 Provider）是两条不同的路径**：前者写入本地索引、能被搜索，且不联网；后者只在日报和设置里的模型检测中使用。改其中一条时不要把另一条的隐私口径带过去。

 ## 文档检查

@@ -51,7 +58,10 @@ pnpm test:e2e:build

 ```bash
 git diff --check
-rg -n "v2\.1\.7|TraceMemo|迹忆|mcpServers|无鉴权" README.md docs --glob '*.md' --glob '!development/overview.md'
+# 过时版本号、旧品牌名、旧结构叙述、MCP 误解
+rg -n "v2\.1\.7|2\.4\.0|v2\.2\.0 兼容期|无鉴权|mcpServers" README.md docs --glob '*.md' --glob '!development/overview.md'
+# 不存在的产品结构（定时日报已并入自动化）
+rg -n "日报 → 定时日报|Monitor / Automation" README.md docs --glob '*.md'
 ```

-历史迁移说明可以出现旧版本号；正式使用指南不要把过时版本写成当前版本。负向澄清“6131 不是 MCP Server”可以保留，以防用户照抄错误配置。
+历史迁移说明可以出现旧版本号；正式使用指南不要把过时版本写成当前版本。负向澄清“6131 不是 MCP Server”可以保留，以防用户照抄错误配置。发版前额外确认 `README.md` 里的版本号与 `package.json` 的 `version` 一致。
@@ -1,6 +1,6 @@
 ---
 name: tracememo-reader
-description: 通过 TraceMemo 本地 HTTP API 按需读取用户有权访问的微信聊天数据和图片媒体。当用户要求查看微信消息、查找联系人或群聊、总结聊天、查看或理解图片、生成群聊总结时使用。此 Skill 由本机 TraceMemo 提供数据，不是 MCP Server。
+description: 通过 TraceMemo 本地 HTTP API 按需读取用户有权访问的微信聊天数据和图片媒体，并管理定时日报任务。当用户要求查看微信消息、查找联系人或群聊、总结聊天、查看或理解图片、生成群聊总结、查询或修改定时日报时使用。此 Skill 由本机 TraceMemo 提供数据，不是 MCP Server。
 ---

 # TraceMemo Reader
@@ -27,31 +27,39 @@ description: 通过 TraceMemo 本地 HTTP API 按需读取用户有权访问的

 ## 端点速查

-| 方法   | 路径                                | 用途                                              |
-| ------ | ----------------------------------- | ------------------------------------------------- |
-| GET    | `/health`                           | 健康和数据库状态                                  |
-| GET    | `/current_time`                     | 本机时间与时区                                    |
-| GET    | `/contact`                          | 联系人/群聊列表；可传 `filter`、`type`            |
-| GET    | `/chatroom`                         | 群聊列表；可传 `keyword`                          |
-| GET    | `/recent_chat`                      | 最近会话；可传 `limit`                            |
-| GET    | `/chatlog`                          | 会话消息；必填 `talker`，可传 `time` 或时间戳范围 |
-| GET    | `/media/{mediaId}`                  | 按消息返回的 `media.url` 获取图片二进制资源       |
-| GET    | `/group_snapshot`                   | 群成员快照；必填 `md5`                            |
-| GET    | `/resolve`                          | 昵称、wxid、md5 解析；必填 `q`                    |
-| GET    | `/wechat-personal/send-capability`  | 个人微信图片发送能力状态                          |
-| GET    | `/scheduled-reports`                | 查询全部定时日报任务                              |
-| GET    | `/scheduled-reports/:id`            | 查询单个定时日报任务                              |
-| POST   | `/scheduled-reports`                | 创建定时日报任务                                  |
-| PATCH  | `/scheduled-reports/:id`            | 修改定时日报任务                                  |
-| DELETE | `/scheduled-reports/:id`            | 删除定时日报任务（执行前必须获得用户确认）        |
-| POST   | `/scheduled-reports/:id/enable`     | 启用定时日报任务                                  |
-| POST   | `/scheduled-reports/:id/disable`    | 暂停定时日报任务                                  |
-| POST   | `/scheduled-reports/:id/run`        | 立即执行一次并返回 execution                      |
-| GET    | `/scheduled-reports/:id/executions` | 查询执行记录                                      |
-| POST   | `/report`                           | 将已有日报结构渲染为 HTML/PNG                     |
-| GET    | `/agent/status`                     | Agent Hub、连接器和数据库状态                     |
-| POST   | `/agent/group-report`               | 按群和 `today`/`yesterday`/`7days` 生成总结图片   |
-| POST   | `/agent/send`                       | 已连接机器人发送测试                              |
+| 方法   | 路径                                                    | 用途                                              |
+| ------ | ------------------------------------------------------- | ------------------------------------------------- |
+| GET    | `/health`                                               | 健康和数据库状态                                  |
+| GET    | `/current_time`                                         | 本机时间与时区                                    |
+| GET    | `/contact`                                              | 联系人/群聊列表；可传 `filter`、`type`            |
+| GET    | `/chatroom`                                             | 群聊列表；可传 `keyword`                          |
+| GET    | `/recent_chat`                                          | 最近会话；可传 `limit`                            |
+| GET    | `/chatlog`                                              | 会话消息；必填 `talker`，可传 `time` 或时间戳范围 |
+| GET    | `/media/{mediaId}`                                      | 按消息返回的 `media.url` 获取图片二进制资源       |
+| GET    | `/group_snapshot`                                       | 群成员快照；必填 `md5`                            |
+| GET    | `/resolve`                                              | 昵称、wxid、md5 解析；必填 `q`                    |
+| POST   | `/query/messages`                                       | 按目标与时间范围取消息（结构化，不调用 AI）       |
+| POST   | `/query/search`                                         | 受限语义关键词检索（依赖本地索引，见 freshness）  |
+| POST   | `/query/message-context`                                | 用 `messageRef` 取某条消息的前后文                |
+| POST   | `/query/conversation-overview`                          | 按会话与时间范围提取可总结的证据                  |
+| GET    | `/query/capabilities`                                   | Query 端点能力目录                                |
+| GET    | `/wechat-personal/send-capability`                      | 个人微信发送能力状态（文字 / 图片 / 语音）        |
+| GET    | `/scheduled-reports`                                    | 查询全部定时日报任务                              |
+| GET    | `/scheduled-reports/:id`                                | 查询单个定时日报任务                              |
+| POST   | `/scheduled-reports`                                    | 创建定时日报任务                                  |
+| PATCH  | `/scheduled-reports/:id`                                | 修改定时日报任务                                  |
+| DELETE | `/scheduled-reports/:id`                                | 删除定时日报任务（执行前必须获得用户确认）        |
+| POST   | `/scheduled-reports/:id/enable`                         | 启用定时日报任务                                  |
+| POST   | `/scheduled-reports/:id/disable`                        | 暂停定时日报任务                                  |
+| POST   | `/scheduled-reports/:id/run`                            | 立即执行一次并返回 execution                      |
+| GET    | `/scheduled-reports/:id/executions`                     | 查询执行记录                                      |
+| POST   | `/scheduled-reports/executions/:executionId/retry-send` | 复用已有 PNG 重试发送                             |
+| POST   | `/report`                                               | 将已有日报结构渲染为 HTML/PNG                     |
+| GET    | `/agent/status`                                         | Agent Hub、连接器和数据库状态                     |
+| POST   | `/agent/group-report`                                   | 按群和 `today`/`yesterday`/`7days` 生成总结图片   |
+| POST   | `/agent/send`                                           | 已连接机器人发送测试（文字或本地图片）            |
+
+这个 API **不只是只读的**：`/report` 会渲染并写文件，`/agent/send` 会真的发出微信消息，`/scheduled-reports*` 会创建、修改、删除或立刻执行定时任务。这些调用都要先确认用户意图；`DELETE` 与 `/agent/send` 尤其需要用户明确确认。

 ## 定时日报管理

@@ -78,7 +86,7 @@ description: 通过 TraceMemo 本地 HTTP API 按需读取用户有权访问的
 }
 ```

-如果 API 返回 `409` 且 `error === "duplicate"`，告诉用户相同任务已经存在，不要再次创建。能力状态为 `unsupported`、`unconfigured`、`needs_binding`、`needs_verification` 或 `error` 时，直接说明需要先在 TraceMemo 设置中完成个人微信绑定和消息能力检测。
+如果 API 返回 `409` 且 `error === "duplicate"`，告诉用户相同任务已经存在，不要再次创建。能力状态不是 `ready` 时（`unsupported`、`unconfigured`、`needs_binding`、`initializing` 或 `error`），直接说明需要先在 TraceMemo 的“设置 → 发送能力”里完成个人微信绑定和能力检测，不要继续创建任务。

 ### 查看、修改和执行

@@ -102,6 +110,20 @@ description: 通过 TraceMemo 本地 HTTP API 按需读取用户有权访问的
 - 根据多条消息整理出的总结；
 - 没有来源支持的推断。

+## 结构化查询（query/\*）
+
+需要按目标 + 时间范围稳定取数时，优先使用 `query/*`，而不是自己拼 `chatlog`：
+
+- `query/messages`：按 `target`、`timeRange`、`direction`、`messageTypes` 取消息；
+- `query/search`：受限语义关键词检索，依赖本地索引；
+- `query/message-context`：用返回的 `messageRef` 取前后文；
+- `query/conversation-overview`：按会话与时间范围提取可总结的证据。
+
+两个要点：
+
+- `messageRef` 是服务端生成的不透明引用，**不要**自行构造 wxid、md5 或数据库路径；
+- `query/search` 依赖异步建立的本地索引。`coverage.state` 不是 `complete` 且 `evidence` 为空时，只能说“这段范围暂时无法确认”，**不能**下“没有找到”的结论。
+
 ## 媒体消息

 当 `/chatlog` 返回图片消息时：
@@ -1,674 +0,0 @@
-                    GNU GENERAL PUBLIC LICENSE
-                       Version 3, 29 June 2007
-
- Copyright (C) 2026 yincongcyincong
- Everyone is permitted to copy and distribute verbatim copies
- of this license document, but changing it is not allowed.
-
-                            Preamble
-
-  The GNU General Public License is a free, copyleft license for
-software and other kinds of works.
-
-  The licenses for most software and other practical works are designed
-to take away your freedom to share and change the works.  By contrast,
-the GNU General Public License is intended to guarantee your freedom to
-share and change all versions of a program--to make sure it remains free
-software for all its users.  We, the Free Software Foundation, use the
-GNU General Public License for most of our software; it applies also to
-any other work released this way by its authors.  You can apply it to
-your programs, too.
-
-  When we speak of free software, we are referring to freedom, not
-price.  Our General Public Licenses are designed to make sure that you
-have the freedom to distribute copies of free software (and charge for
-them if you wish), that you receive source code or can get it if you
-want it, that you can change the software or use pieces of it in new
-free programs, and that you know you can do these things.
-
-  To protect your rights, we need to prevent others from denying you
-these rights or asking you to surrender the rights.  Therefore, you have
-certain responsibilities if you distribute copies of the software, or if
-you modify it: responsibilities to respect the freedom of others.
-
-  For example, if you distribute copies of such a program, whether
-gratis or for a fee, you must pass on to the recipients the same
-freedoms that you received.  You must make sure that they, too, receive
-or can get the source code.  And you must show them these terms so they
-know their rights.
-
-  Developers that use the GNU GPL protect your rights with two steps:
-(1) assert copyright on the software, and (2) offer you this License
-giving you legal permission to copy, distribute and/or modify it.
-
-  For the developers' and authors' protection, the GPL clearly explains
-that there is no warranty for this free software.  For both users' and
-authors' sake, the GPL requires that modified versions be marked as
-changed, so that their problems will not be attributed erroneously to
-authors of previous versions.
-
-  Some devices are designed to deny users access to install or run
-modified versions of the software inside them, although the manufacturer
-can do so.  This is fundamentally incompatible with the aim of
-protecting users' freedom to change the software.  The systematic
-pattern of such abuse occurs in the area of products for individuals to
-use, which is precisely where it is most unacceptable.  Therefore, we
-have designed this version of the GPL to prohibit the practice for those
-products.  If such problems arise substantially in other domains, we
-stand ready to extend this provision to those domains in future versions
-of the GPL, as needed to protect the freedom of users.
-
-  Finally, every program is threatened constantly by software patents.
-States should not allow patents to restrict development and use of
-software on general-purpose computers, but in those that do, we wish to
-avoid the special danger that patents applied to a free program could
-make it effectively proprietary.  To prevent this, the GPL assures that
-patents cannot be used to render the program non-free.
-
-  The precise terms and conditions for copying, distribution and
-modification follow.
-
-                       TERMS AND CONDITIONS
-
-  0. Definitions.
-
-  "This License" refers to version 3 of the GNU General Public License.
-
-  "Copyright" also means copyright-like laws that apply to other kinds of
-works, such as semiconductor masks.
-
-  "The Program" refers to any copyrightable work licensed under this
-License.  Each licensee is addressed as "you".  "Licensees" and
-"recipients" may be individuals or organizations.
-
-  To "modify" a work means to copy from or adapt all or part of the work
-in a fashion requiring copyright permission, other than the making of an
-exact copy.  The resulting work is called a "modified version" of the
-earlier work or a work "based on" the earlier work.
-
-  A "covered work" means either the unmodified Program or a work based
-on the Program.
-
-  To "propagate" a work means to do anything with it that, without
-permission, would make you directly or secondarily liable for
-infringement under applicable copyright law, except executing it on a
-computer or modifying a private copy.  Propagation includes copying,
-distribution (with or without modification), making available to the
-public, and in some countries other activities as well.
-
-  To "convey" a work means any kind of propagation that enables other
-parties to make or receive copies.  Mere interaction with a user through
-a computer network, with no transfer of a copy, is not conveying.
-
-  An interactive user interface displays "Appropriate Legal Notices"
-to the extent that it includes a convenient and prominently visible
-feature that (1) displays an appropriate copyright notice, and (2)
-tells the user that there is no warranty for the work (except to the
-extent that warranties are provided), that licensees may convey the
-work under this License, and how to view a copy of this License.  If
-the interface presents a list of user commands or options, such as a
-menu, a prominent item in the list meets this criterion.
-
-  1. Source Code.
-
-  The "source code" for a work means the preferred form of the work
-for making modifications to it.  "Object code" means any non-source
-form of a work.
-
-  A "Standard Interface" means an interface that either is an official
-standard defined by a recognized standards body, or, in the case of
-interfaces specified for a particular programming language, one that
-is widely used among developers working in that language.
-
-  The "System Libraries" of an executable work include anything, other
-than the work as a whole, that (a) is included in the normal form of
-packaging a Major Component, but which is not part of that Major
-Component, and (b) serves only to enable use of the work with that
-Major Component, or to implement a Standard Interface for which an
-implementation is available to the public in source code form.  A
-"Major Component", in this context, means a major essential component
-(kernel, window system, and so on) of the specific operating system
-(if any) on which the executable work runs, or a compiler used to
-produce the work, or an object code interpreter used to run it.
-
-  The "Corresponding Source" for a work in object code form means all
-the source code needed to generate, install, and (for an executable
-work) run the object code and to modify the work, including scripts to
-control those activities.  However, it does not include the work's
-System Libraries, or general-purpose tools or generally available free
-programs which are used unmodified in performing those activities but
-which are not part of the work.  For example, Corresponding Source
-includes interface definition files associated with source files for
-the work, and the source code for shared libraries and dynamically
-linked subprograms that the work is specifically designed to require,
-such as by intimate data communication or control flow between those
-subprograms and other parts of the work.
-
-  The Corresponding Source need not include anything that users
-can regenerate automatically from other parts of the Corresponding
-Source.
-
-  The Corresponding Source for a work in source code form is that
-same work.
-
-  2. Basic Permissions.
-
-  All rights granted under this License are granted for the term of
-copyright on the Program, and are irrevocable provided the stated
-conditions are met.  This License explicitly affirms your unlimited
-permission to run the unmodified Program.  The output from running a
-covered work is covered by this License only if the output, given its
-content, constitutes a covered work.  This License acknowledges your
-rights of fair use or other equivalent, as provided by copyright law.
-
-  You may make, run and propagate covered works that you do not
-convey, without conditions so long as your license otherwise remains
-in force.  You may convey covered works to others for the sole purpose
-of having them make modifications exclusively for you, or provide you
-with facilities for running those works, provided that you comply with
-the terms of this License in conveying all material for which you do
-not control copyright.  Those thus making or running the covered works
-for you must do so exclusively on your behalf, under your direction
-and control, on terms that prohibit them from making any copies of
-your copyrighted material outside their relationship with you.
-
-  Conveying under any other circumstances is permitted solely under
-the conditions stated below.  Sublicensing is not allowed; section 10
-makes it unnecessary.
-
-  3. Protecting Users' Legal Rights From Anti-Circumvention Law.
-
-  No covered work shall be deemed part of an effective technological
-measure under any applicable law fulfilling obligations under article
-11 of the WIPO copyright treaty adopted on 20 December 1996, or
-similar laws prohibiting or restricting circumvention of such
-measures.
-
-  When you convey a covered work, you waive any legal power to forbid
-circumvention of technological measures to the extent such circumvention
-is effected by exercising rights under this License with respect to
-the covered work, and you disclaim any intention to limit operation or
-modification of the work as a means of enforcing, against the work's
-users, your or third parties' legal rights to forbid circumvention of
-technological measures.
-
-  4. Conveying Verbatim Copies.
-
-  You may convey verbatim copies of the Program's source code as you
-receive it, in any medium, provided that you conspicuously and
-appropriately publish on each copy an appropriate copyright notice;
-keep intact all notices stating that this License and any
-non-permissive terms added in accord with section 7 apply to the code;
-keep intact all notices of the absence of any warranty; and give all
-recipients a copy of this License along with the Program.
-
-  You may charge any price or no price for each copy that you convey,
-and you may offer support or warranty protection for a fee.
-
-  5. Conveying Modified Source Versions.
-
-  You may convey a work based on the Program, or the modifications to
-produce it from the Program, in the form of source code under the
-terms of section 4, provided that you also meet all of these conditions:
-
-    a) The work must carry prominent notices stating that you modified
-    it, and giving a relevant date.
-
-    b) The work must carry prominent notices stating that it is
-    released under this License and any conditions added under section
-    7.  This requirement modifies the requirement in section 4 to
-    "keep intact all notices".
-
-    c) You must license the entire work, as a whole, under this
-    License to anyone who comes into possession of a copy.  This
-    License will therefore apply, along with any applicable section 7
-    additional terms, to the whole of the work, and all its parts,
-    regardless of how they are packaged.  This License gives no
-    permission to license the work in any other way, but it does not
-    invalidate such permission if you have separately received it.
-
-    d) If the work has interactive user interfaces, each must display
-    Appropriate Legal Notices; however, if the Program has interactive
-    interfaces that do not display Appropriate Legal Notices, your
-    work need not make them do so.
-
-  A compilation of a covered work with other separate and independent
-works, which are not by their nature extensions of the covered work,
-and which are not combined with it such as to form a larger program,
-in or on a volume of a storage or distribution medium, is called an
-"aggregate" if the compilation and its resulting copyright are not
-used to limit the access or legal rights of the compilation's users
-beyond what the individual works permit.  Inclusion of a covered work
-in an aggregate does not cause this License to apply to the other
-parts of the aggregate.
-
-  6. Conveying Non-Source Forms.
-
-  You may convey a covered work in object code form under the terms
-of sections 4 and 5, provided that you also convey the
-machine-readable Corresponding Source under the terms of this License,
-in one of these ways:
-
-    a) Convey the object code in, or embodied in, a physical product
-    (including a physical distribution medium), accompanied by the
-    Corresponding Source fixed on a durable physical medium
-    customarily used for software interchange.
-
-    b) Convey the object code in, or embodied in, a physical product
-    (including a physical distribution medium), accompanied by a
-    written offer, valid for at least three years and valid for as
-    long as you offer spare parts or customer support for that product
-    model, to give anyone who possesses the object code either (1) a
-    copy of the Corresponding Source for all the software in the
-    product that is covered by this License, on a durable physical
-    medium customarily used for software interchange, for a price no
-    more than your reasonable cost of physically performing this
-    conveying of source, or (2) access to copy the
-    Corresponding Source from a network server at no charge.
-
-    c) Convey individual copies of the object code with a copy of the
-    written offer to provide the Corresponding Source.  This
-    alternative is allowed only occasionally and noncommercially, and
-    only if you received the object code with such an offer, in accord
-    with subsection 6b.
-
-    d) Convey the object code by offering access from a designated
-    place (gratis or for a charge), and offer equivalent access to the
-    Corresponding Source in the same way through the same place at no
-    further charge.  You need not require recipients to copy the
-    Corresponding Source along with the object code.  If the place to
-    copy the object code is a network server, the Corresponding Source
-    may be on a different server (operated by you or a third party)
-    that supports equivalent copying facilities, provided you maintain
-    clear directions next to the object code saying where to find the
-    Corresponding Source.  Regardless of what server hosts the
-    Corresponding Source, you remain obligated to ensure that it is
-    available for as long as needed to satisfy these requirements.
-
-    e) Convey the object code using peer-to-peer transmission, provided
-    you inform other peers where the object code and Corresponding
-    Source of the work are being offered to the general public at no
-    charge under subsection 6d.
-
-  A separable portion of the object code, whose source code is excluded
-from the Corresponding Source as a System Library, need not be
-included in conveying the object code work.
-
-  A "User Product" is either (1) a "consumer product", which means any
-tangible personal property which is normally used for personal, family,
-or household purposes, or (2) anything designed or sold for incorporation
-into a dwelling.  In determining whether a product is a consumer product,
-doubtful cases shall be resolved in favor of coverage.  For a particular
-product received by a particular user, "normally used" refers to a
-typical or common use of that class of product, regardless of the status
-of the particular user or of the way in which the particular user
-actually uses, or expects or is expected to use, the product.  A product
-is a consumer product regardless of whether the product has substantial
-commercial, industrial or non-consumer uses, unless such uses represent
-the only significant mode of use of the product.
-
-  "Installation Information" for a User Product means any methods,
-procedures, authorization keys, or other information required to install
-and execute modified versions of a covered work in that User Product from
-a modified version of its Corresponding Source.  The information must
-suffice to ensure that the continued functioning of the modified object
-code is in no case prevented or interfered with solely because
-modification has been made.
-
-  If you convey an object code work under this section in, or with, or
-specifically for use in, a User Product, and the conveying occurs as
-part of a transaction in which the right of possession and use of the
-User Product is transferred to the recipient in perpetuity or for a
-fixed term (regardless of how the transaction is characterized), the
-Corresponding Source conveyed under this section must be accompanied
-by the Installation Information.  But this requirement does not apply
-if neither you nor any third party retains the ability to install
-modified object code on the User Product (for example, the work has
-been installed in ROM).
-
-  The requirement to provide Installation Information does not include a
-requirement to continue to provide support service, warranty, or updates
-for a work that has been modified or installed by the recipient, or for
-the User Product in which it has been modified or installed.  Access to a
-network may be denied when the modification itself materially and
-adversely affects the operation of the network or violates the rules and
-protocols for communication across the network.
-
-  Corresponding Source conveyed, and Installation Information provided,
-in accord with this section must be in a format that is publicly
-documented (and with an implementation available to the public in
-source code form), and must require no special password or key for
-unpacking, reading or copying.
-
-  7. Additional Terms.
-
-  "Additional permissions" are terms that supplement the terms of this
-License by making exceptions from one or more of its conditions.
-Additional permissions that are applicable to the entire Program shall
-be treated as though they were included in this License, to the extent
-that they are valid under applicable law.  If additional permissions
-apply only to part of the Program, that part may be used separately
-under those permissions, but the entire Program remains governed by
-this License without regard to the additional permissions.
-
-  When you convey a copy of a covered work, you may at your option
-remove any additional permissions from that copy, or from any part of
-it.  (Additional permissions may be written to require their own
-removal in certain cases when you modify the work.)  You may place
-additional permissions on material, added by you to a covered work,
-for which you have or can give appropriate copyright permission.
-
-  Notwithstanding any other provision of this License, for material you
-add to a covered work, you may (if authorized by the copyright holders of
-that material) supplement the terms of this License with terms:
-
-    a) Disclaiming warranty or limiting liability differently from the
-    terms of sections 15 and 16 of this License; or
-
-    b) Requiring preservation of specified reasonable legal notices or
-    author attributions in that material or in the Appropriate Legal
-    Notices displayed by works containing it; or
-
-    c) Prohibiting misrepresentation of the origin of that material, or
-    requiring that modified versions of such material be marked in
-    reasonable ways as different from the original version; or
-
-    d) Limiting the use for publicity purposes of names of licensors or
-    authors of the material; or
-
-    e) Declining to grant rights under trademark law for use of some
-    trade names, trademarks, or service marks; or
-
-    f) Requiring indemnification of licensors and authors of that
-    material by anyone who conveys the material (or modified versions of
-    it) with contractual assumptions of liability to the recipient, for
-    any liability that these contractual assumptions directly impose on
-    those licensors and authors.
-
-  All other non-permissive additional terms are considered "further
-restrictions" within the meaning of section 10.  If the Program as you
-received it, or any part of it, contains a notice stating that it is
-governed by this License along with a term that is a further
-restriction, you may remove that term.  If a license document contains
-a further restriction but permits relicensing or conveying under this
-License, you may add to a covered work material governed by the terms
-of that license document, provided that the further restriction does
-not survive such relicensing or conveying.
-
-  If you add terms to a covered work in accord with this section, you
-must place, in the relevant source files, a statement of the
-additional terms that apply to those files, or a notice indicating
-where to find the applicable terms.
-
-  Additional terms, permissive or non-permissive, may be stated in the
-form of a separately written license, or stated as exceptions;
-the above requirements apply either way.
-
-  8. Termination.
-
-  You may not propagate or modify a covered work except as expressly
-provided under this License.  Any attempt otherwise to propagate or
-modify it is void, and will automatically terminate your rights under
-this License (including any patent licenses granted under the third
-paragraph of section 11).
-
-  However, if you cease all violation of this License, then your
-license from a particular copyright holder is reinstated (a)
-provisionally, unless and until the copyright holder explicitly and
-finally terminates your license, and (b) permanently, if the copyright
-holder fails to notify you of the violation by some reasonable means
-prior to 60 days after the cessation.
-
-  Moreover, your license from a particular copyright holder is
-reinstated permanently if the copyright holder notifies you of the
-violation by some reasonable means, this is the first time you have
-received notice of violation of this License (for any work) from that
-copyright holder, and you cure the violation prior to 30 days after
-your receipt of the notice.
-
-  Termination of your rights under this section does not terminate the
-licenses of parties who have received copies or rights from you under
-this License.  If your rights have been terminated and not permanently
-reinstated, you do not qualify to receive new licenses for the same
-material under section 10.
-
-  9. Acceptance Not Required for Having Copies.
-
-  You are not required to accept this License in order to receive or
-run a copy of the Program.  Ancillary propagation of a covered work
-occurring solely as a consequence of using peer-to-peer transmission
-to receive a copy likewise does not require acceptance.  However,
-nothing other than this License grants you permission to propagate or
-modify any covered work.  These actions infringe copyright if you do
-not accept this License.  Therefore, by modifying or propagating a
-covered work, you indicate your acceptance of this License to do so.
-
-  10. Automatic Licensing of Downstream Recipients.
-
-  Each time you convey a covered work, the recipient automatically
-receives a license from the original licensors, to run, modify and
-propagate that work, subject to this License.  You are not responsible
-for enforcing compliance by third parties with this License.
-
-  An "entity transaction" is a transaction transferring control of an
-organization, or substantially all assets of one, or subdividing an
-organization, or merging organizations.  If propagation of a covered
-work results from an entity transaction, each party to that
-transaction who receives a copy of the work also receives whatever
-licenses to the work the party's predecessor in interest had or could
-give under the previous paragraph, plus a right to possession of the
-Corresponding Source of the work from the predecessor in interest, if
-the predecessor has it or can get it with reasonable efforts.
-
-  You may not impose any further restrictions on the exercise of the
-rights granted or affirmed under this License.  For example, you may
-not impose a license fee, royalty, or other charge for exercise of
-rights granted under this License, and you may not initiate litigation
-(including a cross-claim or counterclaim in a lawsuit) alleging that
-any patent claim is infringed by making, using, selling, offering for
-sale, or importing the Program or any portion of it.
-
-  11. Patents.
-
-  A "contributor" is a copyright holder who authorizes use under this
-License of the Program or a work on which the Program is based.  The
-work thus licensed is called the contributor's "contributor version".
-
-  A contributor's "essential patent claims" are all patent claims
-owned or controlled by the contributor, whether already acquired or
-hereafter acquired, that would be infringed by some manner, permitted
-by this License, of making, using, or selling its contributor version,
-but do not include claims that would be infringed only as a
-consequence of further modification of the contributor version.  For
-purposes of this definition, "control" includes the right to grant
-patent sublicenses in a manner consistent with the requirements of
-this License.
-
-  Each contributor grants you a non-exclusive, worldwide, royalty-free
-patent license under the contributor's essential patent claims, to
-make, use, sell, offer for sale, import and otherwise run, modify and
-propagate the contents of its contributor version.
-
-  In the following three paragraphs, a "patent license" is any express
-agreement or commitment, however denominated, not to enforce a patent
-(such as an express permission to practice a patent or covenant not to
-sue for patent infringement).  To "grant" such a patent license to a
-party means to make such an agreement or commitment not to enforce a
-patent against the party.
-
-  If you convey a covered work, knowingly relying on a patent license,
-and the Corresponding Source of the work is not available for anyone
-to copy, free of charge and under the terms of this License, through a
-publicly available network server or other readily accessible means,
-then you must either (1) cause the Corresponding Source to be so
-available, or (2) arrange to deprive yourself of the benefit of the
-patent license for this particular work, or (3) arrange, in a manner
-consistent with the requirements of this License, to extend the patent
-license to downstream recipients.  "Knowingly relying" means you have
-actual knowledge that, but for the patent license, your conveying the
-covered work in a country, or your recipient's use of the covered work
-in a country, would infringe one or more identifiable patents in that
-country that you have reason to believe are valid.
-
-  If, pursuant to or in connection with a single transaction or
-arrangement, you convey, or propagate by procuring conveyance of, a
-covered work, and grant a patent license to some of the parties
-receiving the covered work authorizing them to use, propagate, modify
-or convey a specific copy of the covered work, then the patent license
-you grant is automatically extended to all recipients of the covered
-work and works based on it.
-
-  A patent license is "discriminatory" if it does not include within
-the scope of its coverage, prohibits the exercise of, or is
-conditioned on the non-exercise of one or more of the rights that are
-specifically granted under this License.  You may not convey a covered
-work if you are a party to an arrangement with a third party that is
-in the business of distributing software, under which you make payment
-to the third party based on the extent of your activity of conveying
-the work, and under which the third party grants, to any of the
-parties who would receive the covered work from you, a discriminatory
-patent license (a) in connection with copies of the covered work
-conveyed by you (or copies made from those copies), or (b) primarily
-for and in connection with specific products or compilations that
-contain the covered work, unless you entered into that arrangement,
-or that patent license was granted, prior to 28 March 2007.
-
-  Nothing in this License shall be construed as excluding or limiting
-any implied license or other defenses to infringement that may
-otherwise be available to you under applicable patent law.
-
-  12. No Surrender of Others' Freedom.
-
-  If conditions are imposed on you (whether by court order, agreement or
-otherwise) that contradict the conditions of this License, they do not
-excuse you from the conditions of this License.  If you cannot convey a
-covered work so as to satisfy simultaneously your obligations under this
-License and any other pertinent obligations, then as a consequence you may
-not convey it at all.  For example, if you agree to terms that obligate you
-to collect a royalty for further conveying from those to whom you convey
-the Program, the only way you could satisfy both those terms and this
-License would be to refrain entirely from conveying the Program.
-
-  13. Use with the GNU Affero General Public License.
-
-  Notwithstanding any other provision of this License, you have
-permission to link or combine any covered work with a work licensed
-under version 3 of the GNU Affero General Public License into a single
-combined work, and to convey the resulting work.  The terms of this
-License will continue to apply to the part which is the covered work,
-but the special requirements of the GNU Affero General Public License,
-section 13, concerning interaction through a network will apply to the
-combination as such.
-
-  14. Revised Versions of this License.
-
-  The Free Software Foundation may publish revised and/or new versions of
-the GNU General Public License from time to time.  Such new versions will
-be similar in spirit to the present version, but may differ in detail to
-address new problems or concerns.
-
-  Each version is given a distinguishing version number.  If the
-Program specifies that a certain numbered version of the GNU General
-Public License "or any later version" applies to it, you have the
-option of following the terms and conditions either of that numbered
-version or of any later version published by the Free Software
-Foundation.  If the Program does not specify a version number of the
-GNU General Public License, you may choose any version ever published
-by the Free Software Foundation.
-
-  If the Program specifies that a proxy can decide which future
-versions of the GNU General Public License can be used, that proxy's
-public statement of acceptance of a version permanently authorizes you
-to choose that version for the Program.
-
-  Later license versions may give you additional or different
-permissions.  However, no additional obligations are imposed on any
-author or copyright holder as a result of your choosing to follow a
-later version.
-
-  15. Disclaimer of Warranty.
-
-  THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
-APPLICABLE LAW.  EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
-HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
-OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
-THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
-PURPOSE.  THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
-IS WITH YOU.  SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
-ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
-
-  16. Limitation of Liability.
-
-  IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
-WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
-THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
-GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
-USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
-DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
-PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
-EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
-SUCH DAMAGES.
-
-  17. Interpretation of Sections 15 and 16.
-
-  If the disclaimer of warranty and limitation of liability provided
-above cannot be given local legal effect according to their terms,
-reviewing courts shall apply local law that most closely approximates
-an absolute waiver of all civil liability in connection with the
-Program, unless a warranty or assumption of liability accompanies a
-copy of the Program in return for a fee.
-
-                     END OF TERMS AND CONDITIONS
-
-            How to Apply These Terms to Your New Programs
-
-  If you develop a new program, and you want it to be of the greatest
-possible use to the public, the best way to achieve this is to make it
-free software which everyone can redistribute and change under these terms.
-
-  To do so, attach the following notices to the program.  It is safest
-to attach them to the start of each source file to most effectively
-state the exclusion of warranty; and each file should have at least
-the "copyright" line and a pointer to where the full notice is found.
-
-    <one line to give the program's name and a brief idea of what it does.>
-    Copyright (C) <year>  <name of author>
-
-    This program is free software: you can redistribute it and/or modify
-    it under the terms of the GNU General Public License as published by
-    the Free Software Foundation, either version 3 of the License, or
-    (at your option) any later version.
-
-    This program is distributed in the hope that it will be useful,
-    but WITHOUT ANY WARRANTY; without even the implied warranty of
-    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-    GNU General Public License for more details.
-
-    You should have received a copy of the GNU General Public License
-    along with this program.  If not, see <https://www.gnu.org/licenses/>.
-
-Also add information on how to contact you by electronic and paper mail.
-
-  If the program does terminal interaction, make it output a short
-notice like this when it starts in an interactive mode:
-
-    <program>  Copyright (C) <year>  <name of author>
-    This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
-    This is free software, and you are welcome to redistribute it
-    under certain conditions; type `show c' for details.
-
-The hypothetical commands `show w' and `show c' should show the appropriate
-parts of the General Public License.  Of course, your program's commands
-might be different; for a GUI interface, you would use an "about box".
-
-  You should also get your employer (if you work as a programmer) or school,
-if any, to sign a "copyright disclaimer" for the program, if necessary.
-For more information on this, and how to apply and follow the GNU GPL, see
-<https://www.gnu.org/licenses/>.
-
-  The GNU General Public License does not permit incorporating your program
-into proprietary programs.  If your program is a subroutine library, you
-may consider it more useful to permit linking proprietary applications with
-the library.  If this is what you want to do, use the GNU Lesser General
-Public License instead of this License.  But first, please read
-<https://www.gnu.org/licenses/why-not-lgpl.html>.
@@ -1,28 +0,0 @@
-# wechat_chatter / OneBot 第三方组件说明
-
-TraceMemo 的 macOS 个人微信发送功能会按需使用以下第三方组件：
-
- 项目：`yincongcyincong/wechat_chatter`
- 上游仓库：https://github.com/yincongcyincong/wechat_chatter
- 当前运行时版本：`v0.0.18`
- 运行时文件：`onebot_mac_arm64.tar.gz`
- 许可证：GNU General Public License version 3（GPL-3.0）
- 上游版权：Copyright (C) 2026 yincongcyincong
- TraceMemo 修改日期：2026-08-17
-
-## 集成方式
-
-OneBot 运行时不会随 TraceMemo 安装包一起分发。用户启用该实验性功能时，TraceMemo 会从上述上游项目的 GitHub Release 按需下载运行时，并将其安装到应用的用户数据目录。
-
-运行时作为独立进程启动，TraceMemo 通过本机 HTTP 接口与其通信。
-
-## 本地修改
-
-为适配连续发送、图片上传 Hook 状态检测以及微信核心模块基址定位，TraceMemo 会在用户设备上对上游 `onebot/script.js` 应用兼容性补丁。补丁逻辑位于：
-
- `scripts/prepare-wechat-chatter-runtime.cjs`
- `src/main/services/personal-wechat-runtime-manager.ts`
-
-补丁中源自或修改自上游 `script.js` 的部分，以及补丁应用后产生的修改版 `script.js`，继续按照 GPL-3.0 提供。本说明只针对该第三方组件及相关修改，不用于声明 TraceMemo 仓库其他部分的许可证。
-
-GPL-3.0 的完整文本见本目录下的 `LICENSE`。
@@ -33,7 +33,7 @@ Apple Silicon 和 Intel 均已适配微信 macOS `4.1.13` 系列。首次获取
 - 上表中的版本是当前 TraceMemo 已适配或推荐使用的版本，并不代表只有这些版本可以运行。
 - TraceMemo 必须取得当前微信账号对应的数据库密钥，才能读取聊天记录。
 - 你需要有权访问要读取的微信账号和聊天数据。
- 如果要使用 AI 问答、群聊日报或图片理解，还需要在应用中配置一个 AI 服务。
+- 如果要使用 AI 问答、群聊日报或图片理解，还需要在应用中配置一个 AI 服务。「图片文字索引」不在此列——它在本机识别图片里的文字，不需要配置 AI。

 当前代码按微信 4.x 数据结构处理。不同微信客户端版本、系统权限和数据迁移状态可能影响自动连接；遇到问题时请查看[常见问题与排查](./troubleshooting.md)。

@@ -134,6 +134,7 @@ Apple Silicon 和 Intel 均已适配微信 macOS `4.1.13` 系列。首次获取
 - [生成群聊日报或总结](./report.md)
 - [转写微信语音](./voice.md)
 - [导出聊天档案](./export.md)
+- [让日报、退群通知按规则自动运行](../README.md#日报与自动化)
 - [在微信里向 TraceMemo 提问](../agent/agent-hub.md)
 - [让外部 Agent 查询微信历史](../agent/overview.md)

@@ -159,14 +160,17 @@ Agent Hub 是普通用户可以直接使用的入口，不需要安装 Reader Sk

 ## 8. 需要配置 AI 吗？

-不一定。浏览聊天、普通关键词搜索、建立本地知识库和导出不要求在线 AI 服务。
+不一定。浏览聊天、普通关键词搜索、建立本地知识库、图片文字识别、离线语音转写和导出都不要求在线 AI 服务。

 使用“问问微信”、群聊日报或图片理解时，需要在“设置 → AI 模型”中添加并测试 AI 服务。你主动开始并确认远程 AI 功能后，完成任务所需的内容才可能发送给该服务；计费、留存和地区规则由对应服务商决定。

+这两种情况容易混淆：**本机识别图片里的文字**（图片文字索引）不联网、不需要 AI 服务；**让模型看图并回答**（图片理解）才需要配置 AI 服务。
+
 ## 9. 数据和隐私的最低须知

- 微信数据库、聊天解析和本地索引默认留在本机。
+- 微信数据库、聊天解析、本地索引和图片文字识别默认留在本机。
 - 离线语音转写使用本地模型；它与在线 AI 请求是两条不同的数据路径。
+- 图片文字索引只在本机识别，原始图片不会因为本地识别而上传；识别出的文字会进入本地索引，供搜索和“问问微信”使用。
 - 你主动开始并确认 AI 问答或日报后，完成任务所需的受控上下文才可能发送给你选择的 AI 服务；打开应用不会自动上传全部聊天。
 - 应用内 Local HTTP API 默认只监听 `127.0.0.1:6131`，受保护接口需要 Token。

@@ -24,13 +24,13 @@ Knowledge 不会在第一次连接后自动悄悄建立。进入“问问微信

 ### 状态怎么读

-| 状态 | 含义 |
-| ---- | ---- |
-| 可用 · 已追至最新 | 索引已覆盖到聊天记录的最新位置，可以直接用 |
-| 可用 · 正在追新 | 索引可用，正在后台补充最近新增的消息 |
-| 可用 · 正在补齐历史 | 索引可用，正在后台补齐较早的历史内容 |
-| 可用 · 同步已取消 | 索引仍然可用；上一轮同步被取消，已建立的部分保留 |
-| 可用 · 更新失败 | 索引仍然可用；上一轮同步出错，可以稍后重试 |
+| 状态                | 含义                                             |
+| ------------------- | ------------------------------------------------ |
+| 可用 · 已追至最新   | 索引已覆盖到聊天记录的最新位置，可以直接用       |
+| 可用 · 正在追新     | 索引可用，正在后台补充最近新增的消息             |
+| 可用 · 正在补齐历史 | 索引可用，正在后台补齐较早的历史内容             |
+| 可用 · 同步已取消   | 索引仍然可用；上一轮同步被取消，已建立的部分保留 |
+| 可用 · 更新失败     | 索引仍然可用；上一轮同步出错，可以稍后重试       |

 只有确实追平、且没有待补齐内容时才会出现“已追至最新”。索引不可查询时不会显示“可用”。

@@ -38,6 +38,21 @@ Knowledge 不会在第一次连接后自动悄悄建立。进入“问问微信

 同步过程中可以点击 **取消同步**（点击后显示“正在取消…”）。取消只结束当前这一轮，不会删除已经建立的索引，也不会回滚已完成的部分；下次同步会从上次停下的位置继续，不需要从头重扫。中断过的索引仍然可以正常搜索。

+## 图片文字索引（另一份索引）
+
+本地索引其实有两份，彼此独立：
+
+- **聊天记录索引**（也就是上面说的 Knowledge）：索引文字消息，用于跨会话、跨时间查找；
+- **图片文字索引**：在本机识别微信图片里的文字（截图、公告、报价图等），把识别结果也变成可搜索的文字。
+
+“独立”的意思是：聊天记录索引建好了，并不代表图片里的文字就搜得到。建立图片文字索引后，可以在“问问微信”里直接搜截图或公告图里写过的词。
+
+图片文字索引只在本机识别，原始图片不会因为本地识别而上传。它**不等于“图片理解”**：识别文字不联网、不需要 AI 服务；而让模型看图并回答属于图片理解，需要配置 AI 服务，走的是另一条路径。
+
+识别失败的图片可以单独重试，也有“只重建搜索索引、不重新识别图片”的修复入口——修索引不需要重跑几万张图。
+
+两个索引都可以在“设置 → 本地索引”里集中查看状态、建立、同步和清理。
+
 ## 账号隔离

 每个微信账号使用独立的本地索引。切换账号时，应用不会把一个账号的索引混入另一个账号的搜索结果。
@@ -58,4 +73,3 @@ Knowledge 不会在第一次连接后自动悄悄建立。进入“问问微信
 ## 产品术语（可选）

 源码和日志中可能出现 SQLite、FTS、Chunk、索引等词。它们描述的是本地存储和检索实现，不是你开始使用 TraceMemo 的前置知识。
-
@@ -9,6 +9,7 @@ TraceMemo 的核心路径是本地优先，但“本地优先”不等于所有
 - 读取和解析微信数据库；
 - 聊天档案浏览和普通关键词搜索；
 - 本地 Knowledge 索引及其账号隔离；
+- 图片文字索引：识别图片中的文字在本机完成，原始图片不会因为本地识别而上传；
 - 离线语音转写；
 - 导出文件生成和本地日报历史。

@@ -16,7 +17,7 @@ TraceMemo 的核心路径是本地优先，但“本地优先”不等于所有

 ## 什么时候会请求外部服务

-当你主动使用 AI Search、群聊日报或图片理解，并配置了远程 Provider 时，完成任务所需的内容可能发送给该 Provider。当前设置页给出的边界是：
+当你主动使用 AI Search、群聊日报或图片理解，并配置了远程 Provider 时，完成任务所需的内容可能发送给该 Provider。图片文字索引、离线语音转写、档案浏览和普通搜索不会触发这一步。当前设置页给出的边界是：

 - 当前用户问题；
 - 受控检索所需的有限上下文；
@@ -28,7 +29,14 @@ Ollama 等本机 Provider 可以把模型请求留在本机，但本机服务的

 ## 语音和媒体

-离线语音转写在本机进行。图片理解属于 AI 功能：只有你主动启用并使用相关报告/分析路径时，图片才可能按该 Provider 的请求规则被处理。无法读取的媒体不会被自动“猜出来”。
+离线语音转写在本机进行。
+
+图片有两条完全不同的路径，不要混为一谈：
+
+- **图片文字索引**：在本机识别图片里的文字，产出的是本地索引数据；原始图片不会因为这一步被上传，也不需要配置 AI 服务。
+- **图片理解**：属于 AI 功能。只有你主动启用并使用相关报告/分析路径时，图片才可能按该 Provider 的请求规则被处理。
+
+无法读取的媒体不会被自动“猜出来”。

 ## Local HTTP API

@@ -33,9 +33,11 @@

 生成成功后会保存本地 HTML 与 PNG，并出现在日报历史中。你可以复制图片、打开文件位置或重新生成。删除历史日报只删除本地生成的报告文件，不会影响微信聊天数据库。

-## 定时日报
+## 定时日报（在自动化里）

-在“日报 → 定时日报”中可以创建每天运行的任务。选择群聊、执行时间、日报范围、消息类型和模板后，TraceMemo 会按计划执行：
+定时日报现在是「自动化」里的一种规则，不再单独占一个页面：打开一级导航的「自动化」，新建或编辑一条「定时日报」规则，选择群聊、执行时间、日报范围、消息类型和发送目标。日报页顶部的指引条也会直接跳到自动化。
+
+TraceMemo 会按计划执行：

 ```text
 定时触发 → 读取群聊 → 生成报告 → 保存 Report History → 尝试发送
@@ -45,6 +47,8 @@

 执行记录支持查看已生成的日报。对“等待发送”或“发送失败”的记录，可以直接重试发送，重试会复用已经生成的 PNG，不会重新调用 AI 生成整份报告；完整执行状态和发送边界见[如何把聊天变成可用的信息](../concepts/how-it-works.md#动作执行与审计)。

+发送目标当前支持**当前群聊、文件传输助手、自己、指定好友**——还不是任意群发。
+
 ## 让报告更可靠

 - 先选正确的群和时间范围；
@@ -0,0 +1,16 @@
+extends: ./electron-builder.yml
+extraResources:
+  - from: build/app-update.yml
+    to: app-update.yml
+  - from: resources
+    to: resources
+    filter:
+      - '**/*'
+  - from: docs/skill/tracememo-reader
+    to: skill/tracememo-reader
+    filter:
+      - '**/*'
+mac:
+  target:
+    - dmg
+    - zip
@@ -34,7 +34,7 @@ extraResources:
    to: resources
    filter:
      - '**/*'
-      - '!connectors/wechat-personal/**'
+      - '!runtime/darwin-arm64/**'
  - from: docs/skill/tracememo-reader
    to: skill/tracememo-reader
    filter:
@@ -1,6 +1,6 @@
 {
  "name": "tracememo",
-  "version": "2.4.0",
+  "version": "2.5.0",
  "packageManager": "pnpm@7.33.7",
  "description": "TraceMemo（迹忆）是一款本地优先、可追溯的 AI 微信知识与分析工作台。 原名 WechatExplorer，支持聊天记录搜索、知识库、微信群聊总结和 Agent 助手。",
  "keywords": [
@@ -39,7 +39,7 @@
    "prepare:ffmpeg:mac:arm64": "node scripts/prepare-electron-runtime.cjs --platform darwin --arch arm64",
    "prepare:ffmpeg:mac:x64": "node scripts/prepare-electron-runtime.cjs --platform darwin --arch x64",
    "prepare:win-runtime": "node scripts/prepare-win-runtime.cjs && npm run prepare:ffmpeg:win",
-    "prepare:wechat-personal": "node scripts/prepare-wechat-chatter-runtime.cjs",
+    "prepare:wechat-native": "node scripts/prepare-wechat-native-runtime.cjs",
    "start": "electron-vite preview",
    "predev": "node scripts/ensure-electron-binary.cjs",
    "dev": "node scripts/ensure-env.cjs && electron-vite dev",
@@ -61,6 +61,7 @@
    "build:unpack": "npm run build && electron-builder --config electron-builder.yml --dir",
    "build:win": "npm run typecheck && npm run prepare:win-runtime && electron-vite build && electron-builder --config electron-builder.win.yml --win --x64",
    "build:mac:arm64": "npm run typecheck && npm run prepare:ffmpeg:mac:arm64 && electron-vite build && electron-builder --config electron-builder.yml --mac --arm64",
+    "build:mac:arm64:send-runtime": "npm run typecheck && npm run prepare:ffmpeg:mac:arm64 && electron-vite build && cross-env TM_SEND_RUNTIME_BUILD=1 electron-builder --config electron-builder.send-runtime.yml --mac --arm64",
    "build:mac:x64": "npm run typecheck && npm run prepare:ffmpeg:mac:x64 && electron-vite build && electron-builder --config electron-builder.yml --mac --x64",
    "release": "npm run release:mac && npm run release:win",
    "release:mac": "npm run typecheck && electron-vite build && npm run release:mac:arm64 && npm run release:mac:x64",
@@ -74,8 +75,10 @@
  "dependencies": {
    "@electron-toolkit/preload": "^3.0.2",
    "@electron-toolkit/utils": "^4.0.0",
+    "@koromix/koffi-darwin-x64": "3.1.0",
    "@koromix/koffi-win32-x64": "3.1.0",
    "@napi-rs/system-ocr": "1.2.0",
+    "@napi-rs/system-ocr-darwin-x64": "1.2.0",
    "@napi-rs/system-ocr-win32-x64-msvc": "1.2.0",
    "@radix-ui/react-alert-dialog": "^1.1.23",
    "@radix-ui/react-checkbox": "^1.3.11",
@@ -107,7 +110,9 @@
    "parse5": "^8.0.0",
    "pinyin-pro": "^3.26.0",
    "qrcode": "^1.5.4",
+    "sherpa-onnx-darwin-x64": "1.13.3",
    "sherpa-onnx-node": "1.13.3",
+    "sherpa-onnx-win-x64": "1.13.4",
    "silk-wasm": "^3.7.1",
    "tailwind-merge": "^3.6.0",
    "unzipper": "^0.12.0",
@@ -169,4 +174,4 @@
      "ffmpeg-static"
    ]
  }
-}
+}
@@ -11,8 +11,10 @@ specifiers:
  '@electron-toolkit/preload': ^3.0.2
  '@electron-toolkit/tsconfig': ^2.0.0
  '@electron-toolkit/utils': ^4.0.0
+  '@koromix/koffi-darwin-x64': 3.1.0
  '@koromix/koffi-win32-x64': 3.1.0
  '@napi-rs/system-ocr': 1.2.0
+  '@napi-rs/system-ocr-darwin-x64': 1.2.0
  '@napi-rs/system-ocr-win32-x64-msvc': 1.2.0
  '@playwright/test': ^1.62.1
  '@radix-ui/react-alert-dialog': ^1.1.23
@@ -72,7 +74,9 @@ specifiers:
  react: ^19.2.1
  react-dom: ^19.2.1
  sass: ^1.102.0
+  sherpa-onnx-darwin-x64: 1.13.3
  sherpa-onnx-node: 1.13.3
+  sherpa-onnx-win-x64: 1.13.4
  silk-wasm: ^3.7.1
  tailwind-merge: ^3.6.0
  tailwindcss: 3.4.17
@@ -87,8 +91,10 @@ specifiers:
 dependencies:
  '@electron-toolkit/preload': 3.0.2_electron@43.1.0
  '@electron-toolkit/utils': 4.0.0_electron@43.1.0
+  '@koromix/koffi-darwin-x64': 3.1.0
  '@koromix/koffi-win32-x64': 3.1.0
  '@napi-rs/system-ocr': 1.2.0
+  '@napi-rs/system-ocr-darwin-x64': 1.2.0
  '@napi-rs/system-ocr-win32-x64-msvc': 1.2.0
  '@radix-ui/react-alert-dialog': 1.1.23_eijghdl4n2x4hz6j4cg7ctgbuu
  '@radix-ui/react-checkbox': 1.3.11_eijghdl4n2x4hz6j4cg7ctgbuu
@@ -120,7 +126,9 @@ dependencies:
  parse5: 8.0.1
  pinyin-pro: 3.29.3
  qrcode: 1.5.4
+  sherpa-onnx-darwin-x64: 1.13.3
  sherpa-onnx-node: 1.13.3
+  sherpa-onnx-win-x64: 1.13.4
  silk-wasm: 3.7.1
  tailwind-merge: 3.6.0
  unzipper: 0.12.5
@@ -1585,7 +1593,6 @@ packages:
    cpu: [x64]
    os: [darwin]
    dev: false
-    optional: true

  /@koromix/koffi-freebsd-arm64/3.1.0:
    resolution: {integrity: sha512-vazoPYIhOAlXZksVIqDRMIID4VeUZKx8F3dR90hOobT2ATyOkqNS5dv5UCV7Q7DSq22lQTrdbvENBAhROzCp0w==}
@@ -1689,6 +1696,36 @@ packages:
      - supports-color
    dev: true

+  /@napi-rs/system-ocr-darwin-arm64/1.2.0:
+    resolution: {integrity: sha512-cK8dcDBEl3P4A04xmFJSHEJQxfDytaAIFyDCLqavTp92FVU5plESttWzZsqtTkS81/kzKiBfHyPQffSIndfWbQ==}
+    engines: {node: '>= 10'}
+    cpu: [arm64]
+    os: [darwin]
+    dev: false
+    optional: true
+
+  /@napi-rs/system-ocr-darwin-x64/1.2.0:
+    resolution: {integrity: sha512-u3TBvBGrhmT5Os6AfaxbUEg6VHe8lvrFJNPgThJgshJHyRXUx/wCfTyOroJ22KdVCP5AE4GpwS5tFHMb6p6iaQ==}
+    engines: {node: '>= 10'}
+    cpu: [x64]
+    os: [darwin]
+    dev: false
+
+  /@napi-rs/system-ocr-win32-arm64-msvc/1.2.0:
+    resolution: {integrity: sha512-7ej8uMvmXomw3NXo5gZ5p2Nl6UKsHI+VRU3ELv0mhcxR0sJ6wFifYTu5bJrM1TGcz1/RsaX+TjWMmsDq8vriKQ==}
+    engines: {node: '>= 10'}
+    cpu: [arm64]
+    os: [win32]
+    dev: false
+    optional: true
+
+  /@napi-rs/system-ocr-win32-x64-msvc/1.2.0:
+    resolution: {integrity: sha512-oOoCj3FPWDVctTxx98vMBiMI6m51U+w7SMmMefvmtpcpLelzZ/zYTqdwtWZFAjShaHO+RdaKkcpeVcQuBQiVbA==}
+    engines: {node: '>= 10'}
+    cpu: [x64]
+    os: [win32]
+    dev: false
+
  /@napi-rs/system-ocr/1.2.0:
    resolution: {integrity: sha512-r0f2xNH6U+sth44qF+lUP+2WuHSGUBAry5KSCNuaLDGRbgslFqeROr/qJJ/fb6AjBp3Ov+CJP5MdrOWoaoM3cw==}
    engines: {node: '>= 10'}
@@ -1699,34 +1736,6 @@ packages:
      '@napi-rs/system-ocr-win32-x64-msvc': 1.2.0
    dev: false

-  /@napi-rs/system-ocr-darwin-arm64/1.2.0:
-    resolution: {integrity: sha512-cK8dcDBEl3P4A04xmFJSHEJQxfDytaAIFyDCLqavTp92FVU5plESttWzZsqtTkS81/kzKiBfHyPQffSIndfWbQ==}
-    cpu: [arm64]
-    os: [darwin]
-    engines: {node: '>= 10'}
-    dev: false
-
-  /@napi-rs/system-ocr-darwin-x64/1.2.0:
-    resolution: {integrity: sha512-u3TBvBGrhmT5Os6AfaxbUEg6VHe8lvrFJNPgThJgshJHyRXUx/wCfTyOroJ22KdVCP5AE4GpwS5tFHMb6p6iaQ==}
-    cpu: [x64]
-    os: [darwin]
-    engines: {node: '>= 10'}
-    dev: false
-
-  /@napi-rs/system-ocr-win32-arm64-msvc/1.2.0:
-    resolution: {integrity: sha512-7ej8uMvmXomw3NXo5gZ5p2Nl6UKsHI+VRU3ELv0mhcxR0sJ6wFifYTu5bJrM1TGcz1/RsaX+TjWMmsDq8vriKQ==}
-    cpu: [arm64]
-    os: [win32]
-    engines: {node: '>= 10'}
-    dev: false
-
-  /@napi-rs/system-ocr-win32-x64-msvc/1.2.0:
-    resolution: {integrity: sha512-oOoCj3FPWDVctTxx98vMBiMI6m51U+w7SMmMefvmtpcpLelzZ/zYTqdwtWZFAjShaHO+RdaKkcpeVcQuBQiVbA==}
-    cpu: [x64]
-    os: [win32]
-    engines: {node: '>= 10'}
-    dev: false
-
  /@nodelib/fs.scandir/2.1.5:
    resolution: {integrity: sha512-vq24Bq3ym5HEQm2NKCr3yXDwjc7vTsEThRDnkp2DK9p1uqLR+DHurm/NOTo0KG7HYHU7eppKZj3MyqYuMBf62g==}
    engines: {node: '>= 8'}
@@ -7589,7 +7598,6 @@ packages:
    cpu: [x64]
    os: [darwin]
    dev: false
-    optional: true

  /sherpa-onnx-linux-arm64/1.13.4:
    resolution: {integrity: sha512-RMjMRqT82BgTXypNNGmLe6ZFYhc3WEvnAGl3DdkK7qB/kuXwkL3iHhV31wAecbnWPsnEpUoD+8cFovWSBzsCuw==}
@@ -7628,7 +7636,6 @@ packages:
    cpu: [x64]
    os: [win32]
    dev: false
-    optional: true

  /side-channel-list/1.0.0:
    resolution: {integrity: sha512-FCLHtRD/gnpCiCHEiJLOwdmFP+wzCmDEkc9y7NsYxeF4u7Btsn1ZuwgwJGxImImHicJArLP4R0yX4c2KCrMrTA==}
@@ -0,0 +1,13 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="18" height="18" viewBox="0 0 18 18">
+  <rect
+    x="1.5"
+    y="1.5"
+    width="15"
+    height="15"
+    rx="3.5"
+    fill="none"
+    stroke="#000"
+    stroke-width="1.75"
+  />
+  <circle cx="9" cy="9" r="2.3" fill="#000" />
+</svg>
@@ -125,13 +125,62 @@ function validateSystemOcrRuntime(runtimeResources, platform, arch) {
  }
 }

+/**
+ * koffi 运行期按 `${process.platform}-${process.arch}` 拼出原生包目录名
+ * （node_modules/koffi/src/koffi/index.cjs:153/175），找不到就直接抛
+ * "Cannot find the native Koffi module; did you bundle it correctly?"。
+ * pnpm 7 不支持 supportedArchitectures，会静默跳过外平台可选依赖，所以每个目标平台的
+ * koffi 原生包都必须在 package.json 里显式声明；这里再兜一层，缺了就让构建失败，
+ * 而不是发出一个装得上、却打不开 WCDB 的包。
+ */
+function koffiNativeTarget(platform, arch) {
+  if (platform === 'win32') {
+    return arch === 'x64'
+      ? { label: 'Windows', segments: ['@koromix', 'koffi-win32-x64', 'win32_x64', 'koffi.node'] }
+      : null
+  }
+  if (platform === 'darwin' && (arch === 'x64' || arch === 'arm64')) {
+    return {
+      label: 'macOS',
+      segments: ['@koromix', `koffi-darwin-${arch}`, `darwin_${arch}`, 'koffi.node']
+    }
+  }
+  return null
+}
+
+function validateKoffiRuntime(runtimeResources, platform, arch) {
+  const target = koffiNativeTarget(platform, arch)
+  if (!target) return
+  const nativePath = path.join(
+    runtimeResources,
+    'app.asar.unpacked',
+    'node_modules',
+    ...target.segments
+  )
+  if (!existsSync(nativePath)) {
+    throw new Error(`Missing ${target.label} Koffi native module: ${nativePath}`)
+  }
+}
+
 function normalizeBuilderArch(arch) {
  if (typeof arch === 'string') return arch
  return { 0: 'ia32', 1: 'x64', 2: 'armv7l', 3: 'arm64', 4: 'universal' }[arch] || String(arch)
 }

 function runCodesign(args) {
-  execFileSync('/usr/bin/codesign', args, { stdio: 'ignore' })
+  try {
+    execFileSync('/usr/bin/codesign', args, {
+      encoding: 'utf8',
+      stdio: ['ignore', 'pipe', 'pipe']
+    })
+  } catch (error) {
+    const stderr =
+      error && typeof error === 'object' && 'stderr' in error ? String(error.stderr) : ''
+    if (stderr.trim() && error instanceof Error) {
+      error.message += `\n${stderr.trim()}`
+    }
+    throw error
+  }
 }

 function isMacosCodeValid(targetPath, run = runCodesign) {
@@ -174,8 +223,89 @@ function signMacosHelpers(runtimeResources, run = runCodesign) {
  return helperPaths
 }

+/**
+ * codesign 只把这些位置当作「嵌套代码」并要求它们先各自签好，才肯签外层 app。
+ * 只遍历这一组根目录，而不是整个 bundle：Contents/Resources 下的
+ * app.asar.unpacked 里成千上万个原生文件不属于嵌套代码，逐个签既慢又无意义。
+ */
+const MACOS_CODE_LOCATIONS = [
+  'Frameworks',
+  'MacOS',
+  'PlugIns',
+  'XPCServices',
+  'Helpers',
+  'Library/LoginItems'
+]
+
+function collectNestedMacosCode(dir, depth, targets) {
+  let entries
+  try {
+    entries = readdirSync(dir, { withFileTypes: true })
+  } catch {
+    return
+  }
+  for (const entry of entries) {
+    const entryPath = path.join(dir, entry.name)
+    // framework 里的 Mantle -> Versions/Current/Mantle 这类符号链接指向真实文件，
+    // 真实文件会在更深的层级被走到；这里跳过以免重复签名。
+    if (entry.isSymbolicLink()) continue
+    if (entry.isDirectory()) {
+      if (/\.(app|framework|xpc)$/.test(entry.name)) {
+        targets.push({ path: entryPath, depth, bundle: true })
+      }
+      collectNestedMacosCode(entryPath, depth + 1, targets)
+      continue
+    }
+    if (!entry.isFile()) continue
+    if (readBinaryArchitectures(entryPath).length === 0) continue
+    targets.push({ path: entryPath, depth, bundle: false })
+  }
+}
+
+/**
+ * 返回嵌套代码的签名顺序：深度大的先签（framework 内部的 dylib、无扩展名的
+ * crashpad handler 先于 framework 本身，helper 的可执行文件先于 helper app），
+ * 同深度时文件先于 bundle。
+ */
+function findNestedMacosCodePaths(appBundlePath) {
+  const targets = []
+  for (const location of MACOS_CODE_LOCATIONS) {
+    const root = path.join(appBundlePath, 'Contents', ...location.split('/'))
+    if (existsSync(root)) collectNestedMacosCode(root, 1, targets)
+  }
+  return targets
+    .map((target, index) => ({ ...target, index }))
+    .sort((a, b) => {
+      if (a.depth !== b.depth) return b.depth - a.depth
+      if (a.bundle !== b.bundle) return a.bundle ? 1 : -1
+      return a.index - b.index
+    })
+    .map((target) => target.path)
+}
+
+/**
+ * Electron 43.1.0 的 darwin-x64 官方 zip（sha256 与上游 SHASUMS256.txt 一致）
+ * 里所有嵌套 Mach-O 都是未签名状态，darwin-arm64 那份则是 linker-signed。
+ * codesign 签外层 bundle 时要求子组件已签，否则直接报
+ * "code object is not signed at all" + "In subcomponent: ..."，
+ * 所以 x64 出包时只签外层必然失败，必须先由内向外补签一遍。
+ *
+ * 这里不采用 `--deep`（Apple 已标记 deprecated）：它会把外层的签名选项套用到
+ * 所有子组件上，将来接上 Developer ID + entitlements 时会把 app 的 entitlements
+ * 一并套到 helper 上，属于已知的坑。
+ */
 function signMacosAppBundle(appBundlePath, run = runCodesign) {
  if (isMacosCodeValid(appBundlePath, run)) return appBundlePath
+  for (const nestedPath of findNestedMacosCodePaths(appBundlePath)) {
+    try {
+      run(['--force', '--sign', '-', nestedPath])
+    } catch (error) {
+      throw new Error(
+        'macOS nested code signing failed: ' + path.relative(appBundlePath, nestedPath),
+        { cause: error }
+      )
+    }
+  }
  run(['--force', '--sign', '-', appBundlePath])
  try {
    run(['--verify', '--strict', appBundlePath])
@@ -325,9 +455,70 @@ function pruneForeignArchConnectors(runtimeResources, platform, arch) {
  return removed
 }

+/**
+ * 微信发送运行时打包边界
+ */
+const SEND_RUNTIME_RELATIVE = ['resources', 'runtime', 'darwin-arm64']
+const SEND_RUNTIME_ENTRY = 'tm-wechat-host'
+
+function sendRuntimeLocations(runtimeResources) {
+  return [
+    path.join(runtimeResources, ...SEND_RUNTIME_RELATIVE),
+    path.join(runtimeResources, 'app.asar.unpacked', ...SEND_RUNTIME_RELATIVE)
+  ]
+}
+
+function findSendRuntime(runtimeResources) {
+  return (
+    sendRuntimeLocations(runtimeResources).find((directory) =>
+      existsSync(path.join(directory, SEND_RUNTIME_ENTRY))
+    ) || null
+  )
+}
+
+function isSendRuntimeBuild() {
+  return process.env.TM_SEND_RUNTIME_BUILD === '1'
+}
+
+function enforceSendRuntimeBoundary(
+  runtimeResources,
+  platform,
+  bundlesSendRuntime = isSendRuntimeBuild()
+) {
+  if (platform !== 'darwin') return null
+  const found = findSendRuntime(runtimeResources)
+  if (bundlesSendRuntime) {
+    if (!found) {
+      throw new Error(
+        'This macOS build requires the WeChat send runtime but resources/runtime/darwin-arm64 is missing. ' +
+          'Run `pnpm prepare:wechat-native` first, or point TM_NATIVE_RUNTIME_DIR at the artifact.'
+      )
+    }
+    return found
+  }
+  if (found) {
+    throw new Error(
+      'macOS bundle must not include the WeChat send runtime: ' +
+        found +
+        '. Build with `pnpm build:mac:arm64:send-runtime` (TM_SEND_RUNTIME_BUILD=1) when it is required, ' +
+        'or fix the resources filter in electron-builder.yml.'
+    )
+  }
+  return null
+}
+
 exports.default = async function afterPack(context) {
  const runtimeResources = getRuntimeResources(context)
  const arch = normalizeBuilderArch(context.arch)
+  // 边界先判，越早失败越好。
+  const sendRuntime = enforceSendRuntimeBoundary(runtimeResources, context.electronPlatformName)
+  if (context.electronPlatformName === 'darwin') {
+    console.log(
+      sendRuntime
+        ? `[afterPack] send runtime bundled at ${sendRuntime}`
+        : '[afterPack] send runtime excluded'
+    )
+  }
  validateAsarRuntimeDependencies(runtimeResources)
  validateReaderSkillRuntime(runtimeResources)
  validateSilkWasmRuntime(runtimeResources)
@@ -340,6 +531,7 @@ exports.default = async function afterPack(context) {
  )
  validateSherpaRuntime(runtimeResources, context.electronPlatformName, arch)
  validateSystemOcrRuntime(runtimeResources, context.electronPlatformName, arch)
+  validateKoffiRuntime(runtimeResources, context.electronPlatformName, arch)
  pruneIntelMacKeyTool(runtimeResources, context.electronPlatformName, arch)
  pruneForeignArchConnectors(runtimeResources, context.electronPlatformName, arch)
  pruneForeignArchNativeRuntimes(runtimeResources, context.electronPlatformName, arch)
@@ -352,23 +544,6 @@ exports.default = async function afterPack(context) {
    const productName = context.packager.appInfo.productFilename
    signMacosAppBundle(path.join(context.appOutDir, productName + '.app'))
  }
-
-  if (context.electronPlatformName === 'win32') {
-    const koffiNative = path.join(
-      context.appOutDir,
-      'resources',
-      'app.asar.unpacked',
-      'node_modules',
-      '@koromix',
-      'koffi-win32-x64',
-      'win32_x64',
-      'koffi.node'
-    )
-    if (!existsSync(koffiNative)) {
-      throw new Error(`Missing Windows Koffi native module: ${koffiNative}`)
-    }
-    return
-  }
 }

 exports.getRuntimeResources = getRuntimeResources
@@ -378,6 +553,7 @@ exports.validateFfmpegRuntime = validateFfmpegRuntime
 exports.validateSilkWasmRuntime = validateSilkWasmRuntime
 exports.validateSherpaRuntime = validateSherpaRuntime
 exports.validateSystemOcrRuntime = validateSystemOcrRuntime
+exports.validateKoffiRuntime = validateKoffiRuntime
 exports.pruneIntelMacKeyTool = pruneIntelMacKeyTool
 exports.pruneForeignArchConnectors = pruneForeignArchConnectors
 exports.pruneForeignArchNativeRuntimes = pruneForeignArchNativeRuntimes
@@ -386,3 +562,7 @@ exports.findMacosHelperPaths = findMacosHelperPaths
 exports.isMacosCodeValid = isMacosCodeValid
 exports.signMacosHelpers = signMacosHelpers
 exports.signMacosAppBundle = signMacosAppBundle
+exports.findNestedMacosCodePaths = findNestedMacosCodePaths
+exports.sendRuntimeLocations = sendRuntimeLocations
+exports.findSendRuntime = findSendRuntime
+exports.enforceSendRuntimeBoundary = enforceSendRuntimeBoundary
@@ -1,436 +0,0 @@
-/*
- * wechat_chatter runtime integration
- *
- * Upstream: https://github.com/yincongcyincong/wechat_chatter
- * Runtime version: v0.0.18
- * Upstream license: GNU General Public License version 3 (GPL-3.0)
- *
- * This file applies local compatibility patches to the upstream
- * onebot/script.js. See docs/third-party/wechat-chatter/NOTICE.md.
- */
-
-/* eslint-disable @typescript-eslint/explicit-function-return-type, @typescript-eslint/no-require-imports */
-const { execFileSync } = require('node:child_process')
-const fs = require('node:fs')
-const os = require('node:os')
-const path = require('node:path')
-
-const release = 'v0.0.18'
-const asset = 'onebot_mac_arm64.tar.gz'
-const url = `https://github.com/yincongcyincong/wechat_chatter/releases/download/${release}/${asset}`
-const projectRoot = path.resolve(__dirname, '..')
-const outputDir = path.join(
-  projectRoot,
-  'resources',
-  'connectors',
-  'wechat-personal',
-  'darwin-arm64'
-)
-const archive = path.join(os.tmpdir(), `wechat-chatter-${release}-${asset}`)
-const appleSilicon =
-  process.platform === 'darwin' &&
-  execFileSync('/usr/sbin/sysctl', ['-n', 'hw.optional.arm64'], { encoding: 'utf8' }).trim() === '1'
-
-if (!appleSilicon) {
-  throw new Error('个人微信发送运行时当前仅支持 macOS arm64')
-}
-
-fs.mkdirSync(outputDir, { recursive: true })
-let archiveReady = false
-if (fs.existsSync(archive)) {
-  try {
-    execFileSync('/usr/bin/tar', ['-tzf', archive], { stdio: 'ignore' })
-    archiveReady = true
-    console.log(`[wechat-personal] 复用已下载归档：${archive}`)
-  } catch {
-    // The archive is partial or invalid; curl will resume it below.
-  }
-}
-if (!archiveReady) {
-  console.log(`[wechat-personal] 下载 ${release}，支持断点续传：${archive}`)
-  execFileSync(
-    '/usr/bin/curl',
-    ['--http1.1', '-L', '--fail', '--retry', '3', '--continue-at', '-', '--output', archive, url],
-    { stdio: 'inherit' }
-  )
-}
-
-const extractionDir = fs.mkdtempSync(path.join(os.tmpdir(), 'wechat-chatter-extract-'))
-console.log(`[wechat-personal] 解压并原子安装到 ${outputDir}`)
-try {
-  execFileSync('/usr/bin/tar', ['-xzf', archive, '-C', extractionDir], { stdio: 'inherit' })
-  fs.mkdirSync(path.join(outputDir, 'onebot'), { recursive: true })
-  fs.mkdirSync(path.join(outputDir, 'wechat_version'), { recursive: true })
-  fs.copyFileSync(
-    path.join(extractionDir, 'onebot', 'script.js'),
-    path.join(outputDir, 'onebot', 'script.js')
-  )
-  fs.cpSync(path.join(extractionDir, 'wechat_version'), path.join(outputDir, 'wechat_version'), {
-    recursive: true,
-    force: true
-  })
-  const stagedExecutable = path.join(outputDir, 'onebot', `.onebot-${process.pid}.tmp`)
-  fs.copyFileSync(path.join(extractionDir, 'onebot', 'onebot'), stagedExecutable)
-  fs.chmodSync(stagedExecutable, 0o755)
-  fs.renameSync(stagedExecutable, path.join(outputDir, 'onebot', 'onebot'))
-} finally {
-  fs.rmSync(extractionDir, { recursive: true, force: true })
-}
-
-const executable = path.join(outputDir, 'onebot', 'onebot')
-const script = path.join(outputDir, 'onebot', 'script.js')
-const config = path.join(outputDir, 'wechat_version', '4_1_11_53_mac.json')
-for (const required of [executable, script, config]) {
-  if (!fs.existsSync(required)) throw new Error(`运行时文件缺失：${required}`)
-}
-
-function patchPerSendPayload(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (!source.includes('var activeTriggerX1Payload = ptr(0);')) return
-
-  const activeSend = `    const payloadData = hexToByteArray(payloadHex);
-    activeTriggerX1Payload = Memory.alloc(payloadData.length);
-    activeTriggerX1Payload.writeByteArray(payloadData);
-    activeTriggerX1Payload.add(0x18).writePointer(info.cgiAddr);
-    activeTriggerX1Payload.add(0xb8).writePointer(activeTriggerX1Payload.add(0xc0));
-    activeTriggerX1Payload.add(0x190).writePointer(activeTriggerX1Payload.add(0x198));`
-  const upstreamSend = `    const payloadData = hexToByteArray(payloadHex);
-    triggerX1Payload.writeByteArray(payloadData);
-    triggerX1Payload.add(0x18).writePointer(info.cgiAddr);
-    triggerX1Payload.add(0xb8).writePointer(triggerX1Payload.add(0xc0));
-    triggerX1Payload.add(0x190).writePointer(triggerX1Payload.add(0x198));`
-  if (!source.includes(activeSend)) throw new Error('无法定位 wechat_chatter 连续发送补丁位置')
-  source = source
-    .replace(
-      'var triggerX1Payload;\nvar activeTriggerX1Payload = ptr(0);\nvar triggerX0;',
-      'var triggerX1Payload;\nvar triggerX0;'
-    )
-    .replace(activeSend, upstreamSend)
-    .replace(
-      '        MMStartTask(triggerX0, activeTriggerX1Payload);',
-      '        MMStartTask(triggerX0, triggerX1Payload);'
-    )
-    .replace(
-      '        activeTriggerX1Payload = ptr(0);\n        console.error("[!] Error trigger " + msgType + " MMStartTask: " + e);',
-      '        console.error("[!] Error trigger " + msgType + " MMStartTask: " + e);'
-    )
-    .replace(
-      '\t\t\t\tpendingSendMsgType = "";\n\t\t\t\tactiveTriggerX1Payload = ptr(0);\n\t\t\t\treturn',
-      '\t\t\t\tpendingSendMsgType = "";\n\t\t\t\treturn'
-    )
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已恢复原生发送 payload 布局')
-}
-
-function patchSendContextCapture(scriptPath, strict = true) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (source.includes('function isLikelySendContext(')) return
-  const original = `function AttachSendFunc() {
-    Interceptor.attach(sendFuncAddr.add(0x10), {
-        onEnter: function (args) {
-
-            if (triggerX1Payload) {
-                return
-            }
-
-            triggerX0 = this.context.x0;
-            triggerX1Payload = this.context.x1;
-            console.log(\`[+] 捕获到 StartTask 调用，X0：\${triggerX0}, Payload: \${triggerX1Payload}\`);
-        }
-    })
-}`
-  const patched = `function isLikelySendContext(candidateX0, candidateX1) {
-    try {
-        if (!isReadablePointer(candidateX0) || !isReadablePointer(candidateX1)) return false;
-        var manager = readPointerIfReadable(candidateX0.add(0x18));
-        var cgi = readUtf8StringIfReadable(readPointerIfReadable(candidateX1.add(0x18)));
-        console.log("[debug] StartTask candidate x0=" + candidateX0 + " x1=" + candidateX1 + " x0+0x18=" + manager + " cgi=" + cgi);
-        return !manager.equals(ptr(0));
-    } catch (e) {
-        console.error("[debug] StartTask candidate inspect failed: " + e);
-        return false;
-    }
-}
-
-function AttachSendFunc() {
-    Interceptor.attach(sendFuncAddr.add(0x10), {
-        onEnter: function (args) {
-            if (triggerX1Payload) return;
-            var candidateX0 = this.context.x0;
-            var candidateX1 = this.context.x1;
-            if (!isLikelySendContext(candidateX0, candidateX1)) return;
-            triggerX0 = candidateX0;
-            triggerX1Payload = candidateX1;
-            console.log(\`[+] 捕获到有效 StartTask 上下文，X0：\${triggerX0}, Payload: \${triggerX1Payload}\`);
-        }
-    })
-}`
-  if (!source.includes(original)) {
-    if (strict) throw new Error('无法定位 StartTask 上下文 Hook')
-    return
-  }
-  source = source.replace(original, patched)
-  fs.writeFileSync(scriptPath, source)
-}
-
-function patchVoiceAudioBuffer(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (source.includes('voiceAudioDataAddr = Memory.alloc(audioLen + 1);')) return
-
-  const staticAllocation = 'voiceAudioDataAddr = Memory.alloc(5 * 1024 * 1024); // 预分配5MB'
-  if (!source.includes(staticAllocation)) {
-    throw new Error('无法定位 wechat_chatter 语音缓冲区')
-  }
-  source = source.replace(
-    staticAllocation,
-    'voiceAudioDataAddr = Memory.alloc(1); // 上传前按语音长度重新分配'
-  )
-  const audioLengthMarker = '    const audioLen = audioBytes.length;\n'
-  if (!source.includes(audioLengthMarker)) {
-    throw new Error('无法定位 wechat_chatter 语音上传逻辑')
-  }
-  source = source.replace(
-    audioLengthMarker,
-    `${audioLengthMarker}    voiceAudioDataAddr = Memory.alloc(audioLen + 1);\n`
-  )
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已应用按语音长度分配上传缓冲区补丁')
-}
-
-function patchImageHookReadiness(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (
-    source.includes('捕获到图片上传上下文，uploadGlobalX0') &&
-    source.includes('图片上传 Hook Setup Complete')
-  )
-    return
-  const original = `\t\t\tuploadGlobalX0 = this.context.x0;`
-  const patched = `\t\t\tconst capturedUploadX0 = this.context.x0;
-\t\t\tif (uploadGlobalX0.equals(ptr(0)) && !capturedUploadX0.equals(ptr(0))) {
-\t\t\t\tconsole.log("[+] 捕获到图片上传上下文，uploadGlobalX0：" + capturedUploadX0);
-\t\t\t}
-\t\t\tuploadGlobalX0 = capturedUploadX0;`
-  if (!source.includes(original)) throw new Error('无法定位 wechat_chatter 图片 Hook 状态补丁位置')
-  source = source.replace(original, patched)
-  source = source.replace(
-    '    })\n}\n\n\n\nfunction patchCdnOnComplete()',
-    '    })\n    console.log("[+] 图片上传 Hook Setup Complete.");\n}\n\n\n\nfunction patchCdnOnComplete()'
-  )
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已应用图片 Hook 状态补丁')
-}
-
-// Backport of wechat_chatter PR #36 by @Leslielu:
-// https://github.com/yincongcyincong/wechat_chatter/pull/36
-// TraceMemo adds the verified macOS WeChat 4.1.11.53 addresses.
-// macOS WeChat 4.1.11.53, located and verified by TraceMemo
-function patchCdnColdStart(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (source.includes('function resolveCdnManager()')) return
-
-  const initAddresses = `    uploadImageAddr = baseAddr.add({{.uploadImageAddr}});
-    cndOnCompleteAddr = baseAddr.add({{.cndOnCompleteAddr}});`
-  const patchedInitAddresses = `    uploadImageAddr = baseAddr.add({{.uploadImageAddr}});
-    cndOnCompleteAddr = baseAddr.add({{.cndOnCompleteAddr}});
-    // 冷启动 CdnManager 解析（旧版本缺少可选键时保持 hook 捕获行为）
-    {{if .cdnGetServiceAddr}}cdnGetServiceAddr = baseAddr.add({{.cdnGetServiceAddr}});{{end}}
-    {{if .cdnManagerGetterAddr}}cdnManagerGetterAddr = baseAddr.add({{.cdnManagerGetterAddr}});{{end}}`
-  if (!source.includes(initAddresses)) throw new Error('下载的微信版本配置与当前应用不兼容')
-  source = source.replace(initAddresses, patchedInitAddresses)
-
-  const downloadChunkEnd = `}
-
-function fillUploadX1AndStart`
-  const resolver = `}
-
-// 上传和下载共用同一个 mars::cdn::CdnManager。冷启动时通过服务定位器
-// 取得 [ctx + 0x40]，避免必须先手动发送图片才能让 Hook 捕获上下文。
-function resolveCdnManager() {
-    if (cdnGetServiceAddr.equals(ptr(0)) || cdnManagerGetterAddr.equals(ptr(0))) {
-        return ptr(0);
-    }
-    try {
-        // libc++ SSO 短字符串：数据在 +0，长度写在 +0x17。
-        var strDefault = Memory.alloc(24);
-        strDefault.writeUtf8String("default");
-        strDefault.add(0x17).writeU8(7);
-
-        var getService = new NativeFunction(cdnGetServiceAddr, 'pointer', ['pointer']);
-        var svc = getService(strDefault);
-        if (!isReadablePointer(svc)) {
-            console.error("[!] GetService(\\"default\\") 返回不可读: " + svc);
-            return ptr(0);
-        }
-        var getCtx = new NativeFunction(cdnManagerGetterAddr, 'pointer', ['pointer']);
-        var ctx = getCtx(svc);
-        if (!isReadablePointer(ctx)) {
-            console.error("[!] CdnManager getter 返回不可读: " + ctx);
-            return ptr(0);
-        }
-        var mgr = readPointerIfReadable(ctx.add(0x40));
-        if (!isReadablePointer(mgr)) {
-            console.error("[!] ctx+0x40 管理器指针不可读: ctx=" + ctx);
-            return ptr(0);
-        }
-        return mgr;
-    } catch (e) {
-        console.error("[!] resolveCdnManager 异常: " + e);
-        return ptr(0);
-    }
-}
-
-function ensureCdnManagerX0() {
-    if (uploadGlobalX0.equals(ptr(0)) && downloadGlobalX0 && !downloadGlobalX0.equals(ptr(0))) {
-        uploadGlobalX0 = downloadGlobalX0;
-        console.log("[+] downloadGlobalX0 回填 uploadGlobalX0: " + uploadGlobalX0);
-    }
-    if ((!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) && !uploadGlobalX0.equals(ptr(0))) {
-        downloadGlobalX0 = uploadGlobalX0;
-        console.log("[+] uploadGlobalX0 回填 downloadGlobalX0: " + downloadGlobalX0);
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {
-        var mgr = resolveCdnManager();
-        if (!mgr.equals(ptr(0))) {
-            uploadGlobalX0 = mgr;
-            if (!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) {
-                downloadGlobalX0 = mgr;
-            }
-            console.log("[+] 冷启动服务定位器解析 CdnManager: " + mgr);
-        }
-    }
-    return !uploadGlobalX0.equals(ptr(0));
-}
-
-function fillUploadX1AndStart`
-  if (!source.includes(downloadChunkEnd)) throw new Error('无法定位 wechat_chatter 媒体上传逻辑')
-  source = source.replace(downloadChunkEnd, resolver)
-
-  const declarations = 'var uploadImageAddr;\n'
-  const patchedDeclarations =
-    'var uploadImageAddr;\nvar cdnGetServiceAddr = ptr(0);\nvar cdnManagerGetterAddr = ptr(0);\n'
-  if (!source.includes(declarations)) throw new Error('无法定位 wechat_chatter 媒体地址声明')
-  source = source.replace(declarations, patchedDeclarations)
-
-  const uploadGuard = `function fillUploadX1AndStart(idAddr, pathAddr, x1Buffer, receiver, md5, filePath, payloadHex) {
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  const patchedUploadGuard = `function fillUploadX1AndStart(idAddr, pathAddr, x1Buffer, receiver, md5, filePath, payloadHex) {
-    if (uploadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  if (!source.includes(uploadGuard)) throw new Error('无法定位 wechat_chatter 媒体上传入口')
-  source = source.replace(uploadGuard, patchedUploadGuard)
-
-  const voiceGuard = `function triggerUploadVoice(receiver, voicePath, payloadHex, audioDataHex, durationMs) {
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  const patchedVoiceGuard = `function triggerUploadVoice(receiver, voicePath, payloadHex, audioDataHex, durationMs) {
-    if (uploadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  if (!source.includes(voiceGuard)) throw new Error('无法定位 wechat_chatter 语音上传入口')
-  source = source.replace(voiceGuard, patchedVoiceGuard)
-
-  const uploadHook = `\t\t\tuploadGlobalX0 = capturedUploadX0;`
-  const patchedUploadHook = `\t\t\tuploadGlobalX0 = capturedUploadX0;
-            if ((!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) && !capturedUploadX0.equals(ptr(0))) {
-                downloadGlobalX0 = capturedUploadX0;
-                console.log("[+] 上传hook回填 downloadGlobalX0: " + downloadGlobalX0);
-            }`
-  if (!source.includes(uploadHook)) throw new Error('无法定位 wechat_chatter 图片 Hook')
-  source = source.replace(uploadHook, patchedUploadHook)
-
-  const downloadHook = `            downloadGlobalX0 = this.context.x0;`
-  const patchedDownloadHook = `            downloadGlobalX0 = this.context.x0;
-            if (uploadGlobalX0.equals(ptr(0)) && !downloadGlobalX0.equals(ptr(0))) {
-                uploadGlobalX0 = downloadGlobalX0;
-                console.log("[+] 下载hook回填 uploadGlobalX0: " + uploadGlobalX0);
-            }`
-  if (!source.includes(downloadHook)) throw new Error('无法定位 wechat_chatter 下载 Hook')
-  source = source.replace(downloadHook, patchedDownloadHook)
-
-  const downloadGuard = `function triggerDownload(receiver, cdnUrl, aesKey, filePath, fileType) {
-    if (!downloadGlobalX0) {`
-  const patchedDownloadGuard = `function triggerDownload(receiver, cdnUrl, aesKey, filePath, fileType) {
-    if (!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (!downloadGlobalX0) {`
-  if (!source.includes(downloadGuard)) throw new Error('无法定位 wechat_chatter 媒体下载入口')
-  source = source.replace(downloadGuard, patchedDownloadGuard)
-
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已应用 CdnManager 冷启动解析补丁')
-}
-
-function patchCdnColdStartConfig(configPath) {
-  const source = fs.readFileSync(configPath, 'utf8')
-  let config
-  try {
-    config = JSON.parse(source)
-  } catch {
-    throw new Error('4.1.11.53 版本配置不是有效 JSON')
-  }
-  config.cdnGetServiceAddr = '0x50a15d0'
-  config.cdnManagerGetterAddr = '0x5259290'
-  fs.writeFileSync(configPath, `${JSON.stringify(config, null, 2)}\n`)
-  console.log('[wechat-personal] 已写入 4.1.11.53 CdnManager 地址')
-}
-
-function patchWechatCoreModuleBase(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (source.includes('WeChat core module base:')) return
-  const initMarker = 'function initAddresses() {'
-  const initIndex = source.indexOf(initMarker)
-  if (initIndex < 0 || !source.startsWith('var targetPath = ')) {
-    throw new Error('无法定位 wechat_chatter 基址扫描逻辑')
-  }
-  const patchedHeader = `var targetPath = "/Applications/WeChat.app/Contents/Resources/wechat.dylib";
-var module = Process.enumerateModules().find(function(m) {
-    return m.path === targetPath || m.path.endsWith("/Contents/Resources/wechat.dylib");
-});
-if (!module) {
-    throw new Error("[-] Cannot find WeChat core module: " + targetPath);
-}
-var moduleBase = module.base;
-var baseAddr = moduleBase;
-console.log("[+] WeChat core module base: " + baseAddr + " path=" + module.path);
-setImmediate(initAddresses);
-
-`
-  source = patchedHeader + source.slice(initIndex)
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已应用微信核心模块基址补丁')
-}
-
-function addModifiedWorkNotice(scriptPath) {
-  let source = fs.readFileSync(scriptPath, 'utf8')
-  if (source.includes('TraceMemo wechat_chatter compatibility modifications')) return
-
-  const notice = `/*
- * TraceMemo wechat_chatter compatibility modifications
- * Modified: 2026-08-17
- * Upstream: https://github.com/yincongcyincong/wechat_chatter
- * Runtime version: v0.0.18
- * License: GNU General Public License version 3 (GPL-3.0)
- * Changes: WeChat module discovery, per-send payload isolation, dynamic voice upload buffers,
- * CdnManager cold-start resolution, media hook backfill, and image Hook readiness logging.
- * These modifications are not provided by the upstream author.
- */
-
-`
-  source = notice + source
-  fs.writeFileSync(scriptPath, source)
-  console.log('[wechat-personal] 已写入 GPL 修改声明')
-}
-
-patchWechatCoreModuleBase(script)
-patchPerSendPayload(script)
-patchSendContextCapture(script)
-patchVoiceAudioBuffer(script)
-patchImageHookReadiness(script)
-patchCdnColdStartConfig(config)
-patchCdnColdStart(script)
-addModifiedWorkNotice(script)
-fs.chmodSync(executable, 0o755)
-console.log('[wechat-personal] 运行时准备完成')
@@ -0,0 +1,201 @@
+#!/usr/bin/env node
+
+/* eslint-disable @typescript-eslint/explicit-function-return-type */
+/* eslint-disable @typescript-eslint/no-require-imports */
+const crypto = require('node:crypto')
+const fs = require('node:fs')
+const path = require('node:path')
+
+const REQUIRED_FILES = ['tm-wechat-host', 'libtmwechat.dylib', 'runtime-manifest.json']
+
+function parseArgs(argv) {
+  const result = {}
+  for (let index = 0; index < argv.length; index += 1) {
+    const value = argv[index]
+    if (value === '--source' || value === '--target') {
+      if (!argv[index + 1]) throw new Error(`${value} requires a path`)
+      result[value.slice(2)] = argv[index + 1]
+      index += 1
+    } else {
+      throw new Error(`Unknown option: ${value}`)
+    }
+  }
+  return result
+}
+
+function sha256(filePath) {
+  return crypto.createHash('sha256').update(fs.readFileSync(filePath)).digest('hex')
+}
+
+function listFiles(root) {
+  const files = []
+  function visit(directory) {
+    for (const entry of fs.readdirSync(directory, { withFileTypes: true })) {
+      const absolute = path.join(directory, entry.name)
+      if (entry.isDirectory()) visit(absolute)
+      else if (entry.isFile()) files.push(path.relative(root, absolute))
+    }
+  }
+  visit(root)
+  return files.sort()
+}
+
+function assertMachOArm64(filePath, label) {
+  const buffer = fs.readFileSync(filePath)
+  if (buffer.length < 8 || buffer.readUInt32LE(0) !== 0xfeedfacf) {
+    throw new Error(`${label} is not a 64-bit Mach-O binary: ${filePath}`)
+  }
+  if (buffer.readUInt32LE(4) !== 0x0100000c) {
+    throw new Error(`${label} is not arm64: ${filePath}`)
+  }
+}
+
+function readAndValidateArtifact(sourceDir) {
+  if (!fs.existsSync(sourceDir)) {
+    throw new Error('native runtime artifact not found: build the macOS runtime artifact first')
+  }
+
+  const actualFiles = listFiles(sourceDir)
+  const expectedFiles = [...REQUIRED_FILES].sort()
+  if (JSON.stringify(actualFiles) !== JSON.stringify(expectedFiles)) {
+    throw new Error(`native runtime artifact has an unexpected tree: ${actualFiles.join(', ')}`)
+  }
+
+  const manifestPath = path.join(sourceDir, 'runtime-manifest.json')
+  const manifest = JSON.parse(fs.readFileSync(manifestPath, 'utf8'))
+  if (manifest.runtime !== 'tm-wechat-native') throw new Error('Unexpected runtime name')
+  if (manifest.platform !== 'darwin-arm64') throw new Error('Runtime platform must be darwin-arm64')
+  if (manifest.architecture !== 'arm64') throw new Error('Runtime architecture must be arm64')
+  if (manifest.protocolVersion !== 1) throw new Error('Unsupported runtime protocolVersion')
+  if (!/^[a-f0-9]{64}$/.test(manifest.tmSendSourceSha256 || '')) {
+    throw new Error('Runtime manifest has no valid tmSendSourceSha256')
+  }
+  if (!/^[a-f0-9]{64}$/.test(manifest.addressProfileSha256 || '')) {
+    throw new Error('Runtime manifest has no valid addressProfileSha256')
+  }
+  const capabilities = new Set(manifest.capabilities || [])
+  for (const capability of ['text', 'image', 'voice']) {
+    if (!capabilities.has(capability)) throw new Error(`Runtime capability missing: ${capability}`)
+  }
+  if (
+    !Array.isArray(manifest.supportedWechatVersions) ||
+    manifest.supportedWechatVersions.length === 0
+  ) {
+    throw new Error('Runtime manifest has no supportedWechatVersions')
+  }
+
+  for (const relative of REQUIRED_FILES) {
+    const absolute = path.join(sourceDir, relative)
+    if (!fs.statSync(absolute).isFile()) throw new Error(`Runtime file missing: ${relative}`)
+  }
+  const hostPath = path.join(sourceDir, 'tm-wechat-host')
+  const dylibPath = path.join(sourceDir, 'libtmwechat.dylib')
+  assertMachOArm64(hostPath, 'tm-wechat-host')
+  assertMachOArm64(dylibPath, 'libtmwechat.dylib')
+  if (!fs.readFileSync(dylibPath).includes(Buffer.from(manifest.tmSendSourceSha256, 'utf8'))) {
+    throw new Error('Embedded agent hash is not present in libtmwechat.dylib')
+  }
+  if (!fs.readFileSync(dylibPath).includes(Buffer.from(manifest.addressProfileSha256, 'utf8'))) {
+    throw new Error('Embedded address profile hash is not present in libtmwechat.dylib')
+  }
+
+  return {
+    manifest,
+    hashes: Object.fromEntries(
+      REQUIRED_FILES.map((file) => [file, sha256(path.join(sourceDir, file))])
+    )
+  }
+}
+
+function syncArtifact(sourceDir, targetDir) {
+  const source = readAndValidateArtifact(sourceDir)
+  const parent = path.dirname(targetDir)
+  const name = path.basename(targetDir)
+  const staging = path.join(parent, `.${name}.prepare-${process.pid}`)
+  const backup = path.join(parent, `.${name}.backup-${process.pid}`)
+
+  fs.mkdirSync(parent, { recursive: true })
+  fs.rmSync(staging, { recursive: true, force: true })
+  fs.rmSync(backup, { recursive: true, force: true })
+  fs.mkdirSync(staging, { recursive: true })
+
+  try {
+    for (const relative of REQUIRED_FILES) {
+      const destination = path.join(staging, relative)
+      fs.mkdirSync(path.dirname(destination), { recursive: true })
+      fs.copyFileSync(path.join(sourceDir, relative), destination)
+    }
+    fs.chmodSync(path.join(staging, 'tm-wechat-host'), 0o755)
+
+    const staged = readAndValidateArtifact(staging)
+    for (const relative of REQUIRED_FILES) {
+      if (source.hashes[relative] !== staged.hashes[relative]) {
+        throw new Error(`Packaged runtime hash mismatch: ${relative}`)
+      }
+    }
+
+    if (fs.existsSync(targetDir)) fs.renameSync(targetDir, backup)
+    try {
+      fs.renameSync(staging, targetDir)
+    } catch (error) {
+      if (fs.existsSync(backup)) fs.renameSync(backup, targetDir)
+      throw error
+    }
+    fs.rmSync(backup, { recursive: true, force: true })
+    return staged
+  } finally {
+    fs.rmSync(staging, { recursive: true, force: true })
+  }
+}
+
+function resolveSource(explicit) {
+  if (explicit) return path.resolve(explicit)
+  const fromEnv = String(process.env.TM_NATIVE_RUNTIME_DIR || '').trim()
+  if (fromEnv) return path.resolve(fromEnv)
+  // 本机路径写在这里即可，该文件不进仓库。
+  const localConfig = path.join(__dirname, '..', '.native-runtime-source')
+  if (fs.existsSync(localConfig)) {
+    const configured = fs.readFileSync(localConfig, 'utf8').trim()
+    if (configured) return path.resolve(configured)
+  }
+  throw new Error(
+    'runtime artifact source is required: pass --source <dir>, set TM_NATIVE_RUNTIME_DIR, ' +
+      'or write the path into .native-runtime-source'
+  )
+}
+
+function main() {
+  const projectRoot = path.resolve(__dirname, '..')
+  const args = parseArgs(process.argv.slice(2))
+  const sourceDir = resolveSource(args.source)
+  const targetDir = path.resolve(
+    args.target || path.join(projectRoot, 'resources', 'runtime', 'darwin-arm64')
+  )
+  const result = syncArtifact(sourceDir, targetDir)
+  console.log(`[prepare-wechat-native] source: ${sourceDir}`)
+  console.log(`[prepare-wechat-native] target: ${targetDir}`)
+  console.log(
+    `[prepare-wechat-native] runtime: ${result.manifest.runtime}/${result.manifest.version}`
+  )
+  console.log(`[prepare-wechat-native] agent: ${result.manifest.tmSendSourceSha256}`)
+  console.log(`[prepare-wechat-native] address profile: ${result.manifest.addressProfileSha256}`)
+  console.log('[prepare-wechat-native] files: 3')
+}
+
+module.exports = {
+  REQUIRED_FILES,
+  assertMachOArm64,
+  readAndValidateArtifact,
+  syncArtifact
+}
+
+if (require.main === module) {
+  try {
+    main()
+  } catch (error) {
+    console.error(
+      `[prepare-wechat-native] ${error instanceof Error ? error.message : String(error)}`
+    )
+    process.exitCode = 1
+  }
+}
@@ -1,8 +1,10 @@
 import crypto from 'crypto'
 import http, { IncomingMessage, ServerResponse, Server } from 'http'
+import { app } from 'electron'
 import {
  isReady,
  listContacts,
+  listContactsAsync,
  listMessages,
  getGroupSnapshot,
  listRecentChat,
@@ -27,6 +29,11 @@ import { safeError, safeLog, safeWarn } from './safe-log'
 import { apiTokenStore } from './api-token-store'
 import { HttpMediaError, readImageMedia, type HttpImageResult } from './http-media-service'
 import { LocalQueryApiService } from './services/local-query-api-service'
+import { automationRuleStore, AutomationRulePersistenceError } from './services/automation-rule-store'
+import { automationExecutionLogService } from './services/automation-execution-log-service'
+import { groupExitMonitorService } from './services/group-exit-monitor-service'
+import { LocalAgentApiError, LocalAgentApiService } from './services/local-agent-api-service'
+import type { GroupStatsService } from './services/group-stats-service'

 export const DEFAULT_HTTP_HOST = '127.0.0.1'
 export const DEFAULT_HTTP_PORT = 6131
@@ -44,6 +51,11 @@ interface RouteContext {
  body?: unknown
 }

+type HttpMethod = 'GET' | 'HEAD' | 'POST' | 'PATCH' | 'DELETE'
+type RouteHandler = ((ctx: RouteContext) => void | Promise<void>) & {
+  allowedMethods: readonly HttpMethod[]
+}
+
 export interface HttpServerOptions {
  tokenProvider?: () => string | null
  mediaProvider?: (messageId: string) => Promise<HttpImageResult>
@@ -53,21 +65,56 @@ export interface HttpServerOptions {
  scheduledReportDatabaseReadyProvider?: ScheduledReportApiDependencies['isDatabaseReady']
  scheduledReportPlatform?: NodeJS.Platform
  queryApiService?: LocalQueryApiService
+  agentApiService?: LocalAgentApiService
+  groupStatsService?: Pick<GroupStatsService, 'getMemberStats'>
+  appVersionProvider?: () => string
 }

 let configuredQueryApiService: LocalQueryApiService | undefined
+let configuredGroupStatsService: Pick<GroupStatsService, 'getMemberStats'> | undefined
 export function setLocalQueryApiService(service: LocalQueryApiService | undefined): void {
  configuredQueryApiService = service
 }

-type RouteHandler = (ctx: RouteContext) => void | Promise<void>
+export function setLocalGroupStatsService(
+  service: Pick<GroupStatsService, 'getMemberStats'> | undefined
+): void {
+  configuredGroupStatsService = service
+}
+
+const MAX_JSON_BODY_BYTES = 1024 * 1024
+
+class RequestBodyTooLargeError extends Error {
+  constructor() {
+    super('Request body exceeds the maximum size')
+    this.name = 'RequestBodyTooLargeError'
+  }
+}
+
+function withMethods(
+  methods: readonly HttpMethod[],
+  handler: (ctx: RouteContext) => void | Promise<void>
+): RouteHandler {
+  return Object.assign(handler, { allowedMethods: methods })
+}
+
+function sendMethodNotAllowed(res: ServerResponse, methods: readonly HttpMethod[]): void {
+  res.setHeader('Allow', methods.join(', '))
+  sendError(res, 405, `请求方法不受支持；允许的方法：${methods.join(', ')}`)
+}

 function sendJson(res: ServerResponse, status: number, payload: unknown): void {
-  const body = JSON.stringify(payload, null, 2)
+  const requestId = String(res.getHeader('X-Request-Id') || '')
+  let responsePayload = payload
+  if (status >= 400 && payload && typeof payload === 'object' && !('requestId' in payload)) {
+    responsePayload = { ...(payload as Record<string, unknown>), requestId }
+  }
+  const body = JSON.stringify(responsePayload, null, 2)
  res.writeHead(status, {
    'Content-Type': 'application/json; charset=utf-8',
    'Content-Length': Buffer.byteLength(body),
-    'Cache-Control': 'no-store'
+    'Cache-Control': 'no-store',
+    ...(requestId ? { 'X-Request-Id': requestId } : {})
  })
  res.end(body)
 }
@@ -90,8 +137,8 @@ function applyCorsHeaders(req: IncomingMessage, res: ServerResponse): boolean {
  if (!isAllowedCorsOrigin(origin)) return false
  res.setHeader('Access-Control-Allow-Origin', origin)
  res.setHeader('Vary', 'Origin')
-  res.setHeader('Access-Control-Allow-Methods', 'GET, POST, PATCH, DELETE, OPTIONS')
-  res.setHeader('Access-Control-Allow-Headers', 'Content-Type, Authorization')
+  res.setHeader('Access-Control-Allow-Methods', 'GET, HEAD, POST, PATCH, DELETE, OPTIONS')
+  res.setHeader('Access-Control-Allow-Headers', 'Content-Type, Authorization, X-Request-Id')
  return true
 }

@@ -111,6 +158,20 @@ function sendUnauthorized(res: ServerResponse): void {
  })
 }

+function sendAgentHttpError(
+  res: ServerResponse,
+  status: number,
+  code: string,
+  message: string,
+  details?: unknown
+): void {
+  const requestId = String(res.getHeader('X-Request-Id') || '')
+  sendJson(res, status, {
+    error: { code, message, ...(details !== undefined ? { details } : {}) },
+    requestId
+  })
+}
+
 function sendError(res: ServerResponse, status: number, message: string, extra?: unknown): void {
  sendJson(res, status, { error: message, status, ...(extra ? { details: extra } : {}) })
 }
@@ -120,7 +181,8 @@ function sendBinary(res: ServerResponse, status: number, result: HttpImageResult
    'Content-Type': result.mimeType,
    'Content-Length': result.buffer.length,
    'Cache-Control': 'private, no-store',
-    'X-Content-Type-Options': 'nosniff'
+    'X-Content-Type-Options': 'nosniff',
+    'X-Request-Id': String(res.getHeader('X-Request-Id') || '')
  })
  res.end(result.buffer)
 }
@@ -135,10 +197,26 @@ function sanitizeChatlogMessage(message: Record<string, unknown>): Record<string
  return { ...message, contentData: safeContentData }
 }

-function readBody(req: IncomingMessage): Promise<string> {
+function readBody(req: IncomingMessage, maxBytes = MAX_JSON_BODY_BYTES): Promise<string> {
  return new Promise((resolve, reject) => {
+    const contentLength = Number(req.headers['content-length'])
+    if (Number.isFinite(contentLength) && contentLength > maxBytes) {
+      req.pause()
+      reject(new RequestBodyTooLargeError())
+      return
+    }
    const chunks: Buffer[] = []
-    req.on('data', (chunk: Buffer) => chunks.push(chunk))
+    let size = 0
+    req.on('data', (chunk: Buffer) => {
+      size += chunk.length
+      if (size > maxBytes) {
+        chunks.length = 0
+        req.pause()
+        reject(new RequestBodyTooLargeError())
+        return
+      }
+      chunks.push(chunk)
+    })
    req.on('end', () => resolve(Buffer.concat(chunks).toString('utf-8')))
    req.on('error', reject)
  })
@@ -207,18 +285,26 @@ function parseNumeric(value: string | null, fallback: number): number {
  return Number.isFinite(n) ? n : fallback
 }

+function getApplicationVersion(): string {
+  try {
+    return typeof app.getVersion === 'function' ? app.getVersion() : 'unknown'
+  } catch {
+    return 'unknown'
+  }
+}
+
 const routes: Record<string, RouteHandler> = {
-  '/api/v1/health': ({ res }) => {
+  '/api/v1/health': withMethods(['GET'], ({ res }) => {
    sendJson(res, 200, {
      ok: true,
      ready: isReady(),
      service: 'TraceMemo Reader',
-      version: '1.0.0',
+      version: getApplicationVersion(),
      timestamp: new Date().toISOString()
    })
-  },
+  }),

-  '/api/v1/current_time': ({ res }) => {
+  '/api/v1/current_time': withMethods(['GET'], ({ res }) => {
    const now = new Date()
    sendJson(res, 200, {
      time: now.toISOString(),
@@ -228,23 +314,28 @@ const routes: Record<string, RouteHandler> = {
        now.getDate()
      ).padStart(2, '0')}`
    })
-  },
+  }),

-  '/api/v1/contact': ({ res, url }) => {
+  '/api/v1/contact': withMethods(['GET'], async ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const filter = url.searchParams.get('filter') || undefined
    const type = url.searchParams.get('type') || undefined
-    let contacts = listContacts(filter)
+    // Use the hydrated data source: on macOS the session cache only carries raw
+    // ids until display names / contact identities are hydrated, so the sync
+    // `listContacts` would miss nickname and remark matches (Issue #51).
+    let contacts = await listContactsAsync(filter)
    if (type === 'user' || type === 'group') {
      contacts = contacts.filter((c) => c.type === type)
    }
    sendJson(res, 200, { count: contacts.length, contacts })
-  },
+  }),

-  '/api/v1/chatroom': ({ res, url }) => {
+  '/api/v1/chatroom': withMethods(['GET'], async ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const keyword = url.searchParams.get('keyword') || ''
-    let groups = listContacts().filter((c) => c.type === 'group')
+    // Same hydration requirement as /api/v1/contact: group display names are
+    // exactly the fields that stay un-hydrated on macOS.
+    let groups = (await listContactsAsync()).filter((c) => c.type === 'group')
    if (keyword) {
      const lower = keyword.toLowerCase()
      groups = groups.filter(
@@ -254,16 +345,16 @@ const routes: Record<string, RouteHandler> = {
      )
    }
    sendJson(res, 200, { count: groups.length, chatrooms: groups })
-  },
+  }),

-  '/api/v1/recent_chat': ({ res, url }) => {
+  '/api/v1/recent_chat': withMethods(['GET'], ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const limit = parseNumeric(url.searchParams.get('limit'), 50)
    const items = listRecentChat(limit)
    sendJson(res, 200, { count: items.length, items })
-  },
+  }),

-  '/api/v1/chatlog': ({ res, url }) => {
+  '/api/v1/chatlog': withMethods(['GET'], ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const talker = url.searchParams.get('talker')
    if (!talker) return sendError(res, 400, '缺少必要参数 talker')
@@ -301,27 +392,27 @@ const routes: Record<string, RouteHandler> = {
        sanitizeChatlogMessage(message as unknown as Record<string, unknown>)
      )
    })
-  },
+  }),

-  '/api/v1/group_snapshot': ({ res, url }) => {
+  '/api/v1/group_snapshot': withMethods(['GET'], ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const md5 = url.searchParams.get('md5')
    if (!md5) return sendError(res, 400, '缺少必要参数 md5')
    const snapshot = getGroupSnapshot(md5)
    if (!snapshot) return sendError(res, 404, `未找到群聊: ${md5}`)
    sendJson(res, 200, snapshot)
-  },
+  }),

-  '/api/v1/resolve': ({ res, url }) => {
+  '/api/v1/resolve': withMethods(['GET'], ({ res, url }) => {
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    const q = url.searchParams.get('q')
    if (!q) return sendError(res, 400, '缺少必要参数 q')
    const contact = resolveMd5(q)
    if (!contact) return sendError(res, 404, `未匹配到联系人: ${q}`)
    sendJson(res, 200, contact)
-  },
+  }),

-  '/api/v1/report': async ({ req, res, body }) => {
+  '/api/v1/report': withMethods(['POST'], async ({ req, res, body }) => {
    if (req.method !== 'POST') return sendError(res, 405, '需要 POST 请求')
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    if (typeof body !== 'string' || !body.trim()) {
@@ -348,9 +439,9 @@ const routes: Record<string, RouteHandler> = {
    }
    const result = await exportGroupReport(request)
    sendJson(res, result.success ? 200 : 500, result)
-  },
+  }),

-  '/api/v1/agent/group-report': async ({ req, res, body }) => {
+  '/api/v1/agent/group-report': withMethods(['POST'], async ({ req, res, body }) => {
    if (req.method !== 'POST') return sendError(res, 405, '需要 POST 请求')
    if (!isReady()) return sendError(res, 503, 'TraceMemo 数据库未初始化')
    let request: { group?: string; range?: 'today' | 'yesterday' | '7days' }
@@ -364,9 +455,9 @@ const routes: Record<string, RouteHandler> = {
      range: request.range
    })
    sendJson(res, result.success ? 200 : 400, result)
-  },
+  }),

-  '/api/v1/agent/status': ({ res }) => {
+  '/api/v1/agent/status': withMethods(['GET'], ({ res }) => {
    const status = agentHubService.getStatus()
    sendJson(res, 200, {
      ok: status.hub === 'online' && status.connector === 'online',
@@ -376,9 +467,9 @@ const routes: Record<string, RouteHandler> = {
      databaseReady: status.databaseReady,
      accountId: status.accountId
    })
-  },
+  }),

-  '/api/v1/agent/send': async ({ req, res, body }) => {
+  '/api/v1/agent/send': withMethods(['POST'], async ({ req, res, body }) => {
    if (req.method !== 'POST') return sendError(res, 405, '需要 POST 请求')
    let request: { to?: string; text?: string; media_url?: string }
    try {
@@ -392,7 +483,7 @@ const routes: Record<string, RouteHandler> = {
      mediaUrl: request.media_url
    })
    sendJson(res, result.success ? 200 : result.status === 'token_expired' ? 401 : 503, result)
-  }
+  })
 }

 const SCHEDULED_REPORTS_ROUTE = '/api/v1/scheduled-reports'
@@ -442,18 +533,18 @@ function createScheduledReportRoute(
  api: ScheduledReportApiService
 ): RouteHandler | undefined {
  if (pathname === WECHAT_SEND_CAPABILITY_ROUTE) {
-    return async ({ req, res }) => {
+    return withMethods(['GET'], async ({ req, res }) => {
      if (req.method !== 'GET') return sendError(res, 405, '需要 GET 请求')
      try {
        sendJson(res, 200, { capability: await api.getCapability() })
      } catch (error) {
        sendScheduledError(res, error)
      }
-    }
+    })
  }

  if (pathname === SCHEDULED_REPORTS_ROUTE) {
-    return async ({ req, res, body }) => {
+    return withMethods(['GET', 'POST'], async ({ req, res, body }) => {
      try {
        if (req.method === 'GET') {
          const tasks = await api.list()
@@ -469,7 +560,7 @@ function createScheduledReportRoute(
      } catch (error) {
        sendScheduledError(res, error)
      }
-    }
+    })
  }

  const retryPrefix = `${SCHEDULED_REPORTS_ROUTE}/executions/`
@@ -482,7 +573,7 @@ function createScheduledReportRoute(
    } catch {
      return undefined
    }
-    return async ({ req, res }) => {
+    return withMethods(['POST'], async ({ req, res }) => {
      if (req.method !== 'POST') return sendError(res, 405, '需要 POST 请求')
      try {
        const execution = await api.retrySend(executionId)
@@ -490,7 +581,7 @@ function createScheduledReportRoute(
      } catch (error) {
        sendScheduledError(res, error)
      }
-    }
+    })
  }

  const prefix = `${SCHEDULED_REPORTS_ROUTE}/`
@@ -504,8 +595,15 @@ function createScheduledReportRoute(
    return undefined
  }
  const action = segments[1]
+  if (action && !['enable', 'disable', 'run', 'executions'].includes(action)) return undefined

-  return async ({ req, res, body }) => {
+  const methods: HttpMethod[] = !action
+    ? ['GET', 'PATCH', 'DELETE']
+    : action === 'executions'
+      ? ['GET']
+      : ['POST']
+
+  return withMethods(methods, async ({ req, res, body }) => {
    try {
      if (!action && req.method === 'GET') {
        sendJson(res, 200, { task: await api.get(taskId) })
@@ -545,7 +643,7 @@ function createScheduledReportRoute(
    } catch (error) {
      sendScheduledError(res, error)
    }
-  }
+  })
 }

 const MEDIA_ROUTE_PREFIX = '/api/v1/media/'
@@ -560,42 +658,46 @@ function queryStatusCode(status: string): number {
  return 400
 }

-function createQueryRoute(api: LocalQueryApiService): RouteHandler | undefined {
-  return async ({ req, res, body }) => {
-    const pathname = new URL(req.url || '/', 'http://localhost').pathname
-    if (pathname === '/api/v1/query/capabilities') {
-      if (req.method !== 'GET') return sendError(res, 405, '需要 GET 请求')
+function createQueryRoute(pathname: string, api: LocalQueryApiService): RouteHandler | undefined {
+  if (pathname === '/api/v1/query/capabilities') {
+    return withMethods(['GET'], ({ res }) => {
      return sendJson(res, 200, api.capabilities())
-    }
-    if (req.method !== 'POST') return sendError(res, 405, '需要 POST 请求')
-    let payload: any
+    })
+  }
+  type QueryOperationResult =
+    | Awaited<ReturnType<LocalQueryApiService['messages']>>
+    | Awaited<ReturnType<LocalQueryApiService['search']>>
+    | Awaited<ReturnType<LocalQueryApiService['context']>>
+    | Awaited<ReturnType<LocalQueryApiService['overview']>>
+  const operations: Record<string, (payload: unknown) => Promise<QueryOperationResult>> = {
+    '/api/v1/query/messages': (payload) =>
+      api.messages(payload as Parameters<LocalQueryApiService['messages']>[0]),
+    '/api/v1/query/search': (payload) =>
+      api.search(payload as Parameters<LocalQueryApiService['search']>[0]),
+    '/api/v1/query/message-context': (payload) =>
+      api.context(payload as Parameters<LocalQueryApiService['context']>[0]),
+    '/api/v1/query/conversation-overview': (payload) =>
+      api.overview(payload as Parameters<LocalQueryApiService['overview']>[0])
+  }
+  const operation = operations[pathname]
+  if (!operation) return undefined
+  return withMethods(['POST'], async ({ res, body }) => {
+    let payload: unknown
    try { payload = JSON.parse(typeof body === 'string' ? body : '') } catch { return sendError(res, 400, 'invalid_request') }
    if (!payload || typeof payload !== 'object') return sendError(res, 400, 'invalid_request')
    try {
-      const result = pathname === '/api/v1/query/messages'
-        ? await api.messages(payload)
-        : pathname === '/api/v1/query/search'
-          ? await api.search(payload)
-          : pathname === '/api/v1/query/message-context'
-            ? await api.context(payload)
-            : pathname === '/api/v1/query/conversation-overview'
-              ? await api.overview(payload)
-              : undefined
-      if (!result) return sendError(res, 404, `端点不存在: ${pathname}`)
+      const result = await operation(payload)
      return sendJson(res, queryStatusCode(result.status), result)
    } catch (error) {
      return sendError(res, 400, error instanceof Error ? error.message : 'invalid_request')
    }
-  }
+  })
 }

 function createMediaRoute(
  mediaProvider: (messageId: string) => Promise<HttpImageResult>
 ): RouteHandler {
-  return async ({ req, res, url }) => {
-    if (req.method !== 'GET' && req.method !== 'HEAD') {
-      return sendError(res, 405, '需要 GET 请求')
-    }
+  return withMethods(['GET', 'HEAD'], async ({ req, res, url }) => {
    const encodedMessageId = url.pathname.slice(MEDIA_ROUTE_PREFIX.length)
    let messageId: string
    try {
@@ -634,7 +736,218 @@ function createMediaRoute(
      safeError('[HttpServer] media request failed:', error)
      return sendError(res, 500, '图片读取失败')
    }
+  })
+}
+
+function createAgentApiService(options: HttpServerOptions): LocalAgentApiService {
+  const groupStats = options.groupStatsService || configuredGroupStatsService
+  return options.agentApiService || new LocalAgentApiService({
+    automationRuleStore,
+    automationExecutionLogService,
+    listContacts: listContactsAsync,
+    isDatabaseReady: isReady,
+    getVersion: options.appVersionProvider || getApplicationVersion,
+    getPersonalWechatCapability: () =>
+      personalWechatCapabilityService.getPersonalWechatSendCapability(),
+    getAgentHubStatus: () => agentHubService.getStatus(),
+    getGroupExitMonitorState: () => {
+      const state = groupExitMonitorService.getState()
+      return { ...state, monitoredRoomIds: state.monitoredRoomIds || [] }
+    },
+    configureGroupExitMonitor: (configuration) =>
+      groupExitMonitorService.configure(configuration),
+    setGroupExitMonitorRoomIds: (roomIds) => groupExitMonitorService.setMonitoredRoomIds(roomIds),
+    setGroupExitMonitorEnabled: (enabled) => groupExitMonitorService.setEnabled(enabled),
+    listGroupExitMonitorEvents: (query) => groupExitMonitorService.listEvents(query),
+    ...(groupStats ? { getGroupMemberStats: (query) => groupStats.getMemberStats(query) } : {})
+  })
+}
+
+function sendAgentApiError(res: ServerResponse, error: unknown): void {
+  const requestId = String(res.getHeader('X-Request-Id') || '')
+  if (error instanceof LocalAgentApiError) {
+    sendAgentHttpError(res, error.status, error.code, error.message, error.details)
+    return
  }
+  if (error instanceof AutomationRulePersistenceError) {
+    sendAgentHttpError(res, 500, 'PERSISTENCE_FAILED', '自动化规则未能保存到本地')
+    return
+  }
+  safeError(`[HttpServer requestId=${requestId}] Agent API request failed:`, error)
+  sendAgentHttpError(res, 500, 'INTERNAL_ERROR', 'Agent API 请求失败')
+}
+
+async function handleAgentApi<T>(
+  res: ServerResponse,
+  operation: () => Promise<T> | T,
+  respond: (value: T) => void
+): Promise<void> {
+  try {
+    respond(await operation())
+  } catch (error) {
+    sendAgentApiError(res, error)
+  }
+}
+
+function parseAgentJson(body: unknown): unknown {
+  if (typeof body !== 'string' || !body.trim()) {
+    throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '请求体不能为空')
+  }
+  try {
+    return JSON.parse(body)
+  } catch {
+    throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '请求体 JSON 格式无效')
+  }
+}
+
+function createAgentApiRoute(
+  pathname: string,
+  api: LocalAgentApiService
+): RouteHandler | undefined {
+  if (pathname === '/api/v1/capabilities') {
+    return withMethods(['GET'], ({ res }) =>
+      void handleAgentApi(res, () => api.getCapabilities(), (capabilities) =>
+        sendJson(res, 200, capabilities)
+      )
+    )
+  }
+
+  const groupExitMonitorPath = '/api/v1/monitors/group-exits'
+  if (pathname === groupExitMonitorPath) {
+    return withMethods(['GET', 'PATCH'], async ({ req, res, body }) => {
+      if (req.method === 'GET') {
+        await handleAgentApi(res, () => api.getGroupExitMonitorState(), (state) =>
+          sendJson(res, 200, state)
+        )
+        return
+      }
+      await handleAgentApi(res, () => api.updateGroupExitMonitor(parseAgentJson(body)), (state) =>
+        sendJson(res, 200, state)
+      )
+    })
+  }
+
+  if (pathname === `${groupExitMonitorPath}/events`) {
+    return withMethods(['GET'], ({ res, url }) =>
+      void handleAgentApi(res, () => api.listGroupExitMonitorEvents(url.searchParams), (result) =>
+        sendJson(res, 200, result)
+      )
+    )
+  }
+
+  const groupStatsPrefix = '/api/v1/groups/'
+  if (pathname.startsWith(groupStatsPrefix)) {
+    const segments = pathname.slice(groupStatsPrefix.length).split('/')
+    if (segments.length === 2 && segments[1] === 'member-stats' && segments[0]) {
+      let conversationId: string
+      try {
+        conversationId = decodeURIComponent(segments[0])
+      } catch {
+        return undefined
+      }
+      return withMethods(['GET'], ({ res, url }) =>
+        void handleAgentApi(
+          res,
+          () => api.getGroupMemberStats(conversationId, url.searchParams),
+          (result) => sendJson(res, 200, result)
+        )
+      )
+    }
+  }
+
+  const automationCollection = '/api/v1/automations'
+  if (pathname === automationCollection) {
+    return withMethods(['GET', 'POST'], async ({ req, res, url, body }) => {
+      if (req.method === 'GET') {
+        await handleAgentApi(
+          res,
+          () => api.listAutomations({
+            type: url.searchParams.get('type'),
+            enabled: url.searchParams.get('enabled')
+          }),
+          (rules) => sendJson(res, 200, { count: rules.length, rules })
+        )
+        return
+      }
+      await handleAgentApi(res, () => api.createAutomation(parseAgentJson(body)), (rule) =>
+        sendJson(res, 201, { created: true, rule })
+      )
+    })
+  }
+
+  if (pathname === `${automationCollection}/validate`) {
+    return withMethods(['POST'], async ({ res, body }) => {
+      await handleAgentApi(res, () => api.validateAutomation(parseAgentJson(body)), (result) =>
+        sendJson(res, 200, result)
+      )
+    })
+  }
+
+  if (pathname === `${automationCollection}/executions`) {
+    return withMethods(['GET'], ({ res, url }) =>
+      void handleAgentApi(res, () => api.listExecutions(url.searchParams), (result) =>
+        sendJson(res, 200, result)
+      )
+    )
+  }
+
+  const prefix = `${automationCollection}/`
+  if (!pathname.startsWith(prefix)) return undefined
+  const segments = pathname.slice(prefix.length).split('/').filter(Boolean)
+  if (segments.length < 1 || segments.length > 2) return undefined
+  let id: string
+  try {
+    id = decodeURIComponent(segments[0])
+  } catch {
+    return undefined
+  }
+  if (!id || id.includes('/') || id.includes('\\')) return undefined
+  const action = segments[1]
+  if (action && action !== 'enable' && action !== 'disable') return undefined
+
+  if (action) {
+    return withMethods(['POST'], ({ res }) =>
+      void handleAgentApi(res, () => api.setAutomationEnabled(id, action === 'enable'), (rule) =>
+        sendJson(res, 200, { updated: true, rule })
+      )
+    )
+  }
+
+  return withMethods(['GET', 'PATCH', 'DELETE'], async ({ req, res, body }) => {
+    if (req.method === 'GET') {
+      await handleAgentApi(res, () => api.getAutomation(id), (rule) =>
+        sendJson(res, 200, { rule })
+      )
+      return
+    }
+    if (req.method === 'PATCH') {
+      await handleAgentApi(
+        res,
+        () => api.updateAutomation(id, parseAgentJson(body)),
+        (rule) => sendJson(res, 200, { updated: true, rule })
+      )
+      return
+    }
+    await handleAgentApi(res, () => api.deleteAutomation(id), (result) =>
+      sendJson(res, 200, { deleted: true, ...result })
+    )
+  })
+}
+
+function requestIdFor(req: IncomingMessage): string {
+  const incoming = req.headers['x-request-id']
+  return typeof incoming === 'string' && /^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$/.test(incoming)
+    ? incoming
+    : crypto.randomUUID()
+}
+
+function isAgentApiPath(pathname: string): boolean {
+  return pathname === '/api/v1/capabilities' ||
+    pathname === '/api/v1/monitors/group-exits' ||
+    pathname.startsWith('/api/v1/monitors/group-exits/') ||
+    pathname.startsWith('/api/v1/groups/') ||
+    pathname === '/api/v1/automations' ||
+    pathname.startsWith('/api/v1/automations/')
 }

 export function startHttpServer(
@@ -646,11 +959,19 @@ export function startHttpServer(
  const mediaProvider = options.mediaProvider || readImageMedia
  const scheduledReportApi = createScheduledReportApi(options)
  const queryApi = options.queryApiService || configuredQueryApiService || new LocalQueryApiService()
+  const agentApi = createAgentApiService(options)
  return new Promise((resolve, reject) => {
    const server: Server = http.createServer(async (req, res) => {
+      const requestId = requestIdFor(req)
+      res.setHeader('X-Request-Id', requestId)
+      let agentApiRequest = false
      try {
        const url = new URL(req.url || '/', `http://${host}:${port}`)
+        agentApiRequest = isAgentApiPath(url.pathname)
        if (!applyCorsHeaders(req, res)) {
+          if (agentApiRequest) {
+            return sendAgentHttpError(res, 403, 'FORBIDDEN', 'Origin 不允许访问本地 API')
+          }
          return sendError(res, 403, 'Origin 不允许访问本地 API')
        }
        if (req.method === 'OPTIONS') {
@@ -659,15 +980,37 @@ export function startHttpServer(
        }
        const handler =
          routes[url.pathname] ||
+          createAgentApiRoute(url.pathname, agentApi) ||
          createScheduledReportRoute(url.pathname, scheduledReportApi) ||
-          (url.pathname.startsWith(QUERY_ROUTE_PREFIX) ? createQueryRoute(queryApi) : undefined) ||
+          (url.pathname.startsWith(QUERY_ROUTE_PREFIX)
+            ? createQueryRoute(url.pathname, queryApi)
+            : undefined) ||
          (url.pathname.startsWith(MEDIA_ROUTE_PREFIX)
            ? createMediaRoute(mediaProvider)
            : undefined)
        if (!handler) {
+          if (agentApiRequest) {
+            return sendAgentHttpError(res, 404, 'NOT_FOUND', `端点不存在: ${url.pathname}`)
+          }
          return sendError(res, 404, `端点不存在: ${url.pathname}`)
        }
+        if (!handler.allowedMethods.includes(req.method as HttpMethod)) {
+          if (agentApiRequest) {
+            res.setHeader('Allow', handler.allowedMethods.join(', '))
+            return sendAgentHttpError(
+              res,
+              405,
+              'METHOD_NOT_ALLOWED',
+              `请求方法不受支持；允许的方法：${handler.allowedMethods.join(', ')}`,
+              { allowedMethods: handler.allowedMethods }
+            )
+          }
+          return sendMethodNotAllowed(res, handler.allowedMethods)
+        }
        if (url.pathname !== '/api/v1/health' && !isAuthorized(req, tokenProvider())) {
+          if (agentApiRequest) {
+            return sendAgentHttpError(res, 401, 'UNAUTHORIZED', 'Valid API token required')
+          }
          return sendUnauthorized(res)
        }
        let body: string | undefined
@@ -677,8 +1020,22 @@ export function startHttpServer(
        const ctx: RouteContext = { req, res, url, body }
        await handler(ctx)
      } catch (error) {
-        safeError('[HttpServer] 请求处理失败:', error)
+        if (error instanceof RequestBodyTooLargeError) {
+          res.setHeader('Connection', 'close')
+          res.shouldKeepAlive = false
+          if (agentApiRequest) {
+            sendAgentHttpError(res, 413, 'PAYLOAD_TOO_LARGE', '请求体不能超过 1 MiB')
+            return
+          }
+          sendError(res, 413, '请求体不能超过 1 MiB')
+          return
+        }
+        safeError(`[HttpServer requestId=${requestId}] 请求处理失败:`, error)
        if (!res.headersSent) {
+          if (agentApiRequest) {
+            sendAgentHttpError(res, 500, 'INTERNAL_ERROR', 'Agent API 请求失败')
+            return
+          }
          sendError(res, 500, error instanceof Error ? error.message : String(error))
        }
      }
@@ -14,13 +14,17 @@ import {
  Menu,
  Tray,
  dialog,
-  protocol
+  protocol,
+  screen
 } from 'electron'
 import { dirname, extname, join } from 'path'
 import { existsSync, promises as fsPromises } from 'fs'
 import { electronApp, optimizer, is } from '@electron-toolkit/utils'
 import icon from '../../resources/icon.png?asset'
+import trayTemplateIcon from '../../resources/trayTemplate.png?asset'
+import trayTemplateRetinaIcon from '../../resources/trayTemplate@2x.png?asset'
 import { WechatDb } from './wechat-db'
+import { createTrayImage } from './tray-icon'
 import { bootstrapWcdbNativeAsync, Wcdb4Client } from './wcdb4-client'
 import { VoiceService } from './voice-service'
 import { StickerService } from './sticker-service'
@@ -72,7 +76,7 @@ import type { SystemOcrCapability, SystemOcrRequest, SystemOcrResult } from '../
 import { KeyServiceMac } from './key-service-mac'
 import { KeyService as KeyServiceWin } from './key-service-win'
 import * as chat from './services/chat-service'
-import { apiServer, setLocalQueryApiService } from './http-server'
+import { apiServer, setLocalGroupStatsService, setLocalQueryApiService } from './http-server'
 import { skillResourceService } from './services/skill-resource-service'
 import { buildLocalApiCurlCommand, testLocalApiRequest } from './services/local-api-test-service'
 import { isWechatRunning } from './services/wechat-process-status'
@@ -112,22 +116,28 @@ import { agentHubService } from './services/agent-hub-service'
 import { WechatConnectorService } from './services/wechat-ilink'
 import { wechatSendGateway } from './services/wechat-send-gateway'
 import { groupExitMonitorService } from './services/group-exit-monitor-service'
+import {
+  filterSendableFriendContacts,
+  leaveNotificationContactDisplayName
+} from './services/leave-notification-target'
+import { MessageListenerService } from './services/message-listener-service'
+import { automationRuleStore } from './services/automation-rule-store'
+import { automationExecutionLogService } from './services/automation-execution-log-service'
+import { initAutomationService, getAutomationService } from './services/automation-service'
+import { GroupStatsService } from './services/group-stats-service'
 import { wechatActionLogService } from './services/wechat-action-log-service'
-import { wechatActionGateway } from './services/wechat-action-gateway'
+import { toPersonalWechatSendResult, wechatActionGateway } from './services/wechat-action-gateway'
 import { personalWechatSendService } from './services/personal-wechat-send-service'
+import { macWechatRuntimeManager } from './services/mac-wechat-runtime-manager'
 import { getPersonalWechatSendCapability } from './services/personal-wechat-capability-service'
 import { scheduledReportService } from './services/scheduled-report-service'
-import { PersonalWechatRuntimeManager } from './services/personal-wechat-runtime-manager'
 import { personalWechatVoiceEnvironmentService } from './services/personal-wechat-voice-environment-service'
 import type {
  PersonalWechatGeneratedTtsVoiceRequest,
  PersonalWechatSendRequest,
-  PersonalWechatSendResult
+  PersonalWechatSendResult,
+  PersonalWechatSenderStatus
 } from '../shared/personal-wechat'
-import type {
-  ScheduledReportCreateInput,
-  ScheduledReportUpdateInput
-} from '../shared/scheduled-report'
 import { isTruthyDebugFlag } from '../shared/debug-flags'
 import { TextToSpeechSettingsService } from './services/text-to-speech-settings-service'
 import type {
@@ -196,6 +206,12 @@ import type {
  WechatShareServiceConfig
 } from '../shared/wechat-share-card'

+async function currentPersonalWechatSenderStatus(): Promise<PersonalWechatSenderStatus> {
+  return process.platform === 'darwin'
+    ? macWechatRuntimeManager.buildSenderStatus()
+    : personalWechatSendService.getStatus()
+}
+
 // electron-vite can close the child's stdout/stderr after spawning Electron.
 // Plain console.error then throws EPIPE on a closed pipe and crashes the IPC
 // handler. Wrap console.* before any other module logs anything.
@@ -211,6 +227,7 @@ let voiceService: VoiceService | null = null
 let voiceRecognition: VoiceRecognitionUseCase | null = null
 let voiceBatchService: VoiceBatchService | null = null
 let knowledgeSearchService: KnowledgeSearchService | null = null
+let groupStatsService: GroupStatsService | null = null
 let localQueryApiService: LocalQueryApiService | null = null
 let aiSearchPipelineService: AiSearchPipelineService | null = null
 let queryAgentService: QueryAgentService | null = null
@@ -222,7 +239,6 @@ const databaseKeyStore = new DatabaseKeyStore()
 const imageKeyConfigService = new ImageKeyConfigService()
 const aiProviderService = new AIProviderService()
 const textToSpeechSettingsService = new TextToSpeechSettingsService()
-const personalWechatRuntimeManager = new PersonalWechatRuntimeManager()
 const keyServiceMac = new KeyServiceMac()
 const keyServiceWin = new KeyServiceWin()
 const wechatShareConfigStore = new WechatShareConfigStore()
@@ -357,6 +373,18 @@ function configureRecallProtection(

 const packagedIconPath = join(process.resourcesPath, 'resources', 'icon.png')
 const appIconPath = existsSync(packagedIconPath) ? packagedIconPath : icon
+const packagedTrayTemplateIconPath = join(process.resourcesPath, 'resources', 'trayTemplate.png')
+const packagedTrayTemplateRetinaIconPath = join(
+  process.resourcesPath,
+  'resources',
+  'trayTemplate@2x.png'
+)
+const trayTemplateIconPath = existsSync(packagedTrayTemplateIconPath)
+  ? packagedTrayTemplateIconPath
+  : trayTemplateIcon
+const trayTemplateRetinaIconPath = existsSync(packagedTrayTemplateRetinaIconPath)
+  ? packagedTrayTemplateRetinaIconPath
+  : trayTemplateRetinaIcon

 protocol.registerSchemesAsPrivileged([
  {
@@ -492,9 +520,29 @@ async function createLocalMediaResponse(request: Request, filePath: string): Pro

 function createWindow(): void {
  // 创建浏览器窗口
+  /*
+   * 初始尺寸按当前屏幕工作区的 60% 计算，不再用固定 1400×800。
+   * 两个夹逼都是必要的：
+   * - 下限：小屏上 60% 会算出比原固定值更小的窗口（1920×1080 的工作区高度
+   *   只有 ~985，60% 才 591 高），反而比改动前更糟；高度下限取 900，
+   *   即「高度拉大一点」的诉求。
+   * - 上限：不能超过工作区本身，否则初始尺寸会把标题栏顶出屏幕。
+   */
+  const { workAreaSize } = screen.getPrimaryDisplay()
+  const initialWidth = Math.min(
+    Math.max(Math.round(workAreaSize.width * 0.6), 1400),
+    workAreaSize.width
+  )
+  const initialHeight = Math.min(
+    Math.max(Math.round(workAreaSize.height * 0.6), 900),
+    workAreaSize.height
+  )
  const mainWindow = new BrowserWindow({
-    width: 1400,
-    height: 800,
+    width: initialWidth,
+    height: initialHeight,
+    minWidth: 960,
+    minHeight: 640,
+    center: true,
    show: false,
    autoHideMenuBar: true,
    icon: appIconPath,
@@ -754,6 +802,10 @@ app.whenReady().then(async () => {
  })
  aiSearchPipelineService = new AiSearchPipelineService(knowledgeSearchService, aiProviderService)
  localQueryApiService = new LocalQueryApiService(knowledgeSearchService)
+  // 群员统计复用同一个 Knowledge 实例：它只是「读派生库 + 读成员名单」的编排，
+  // 不持有自己的数据库，也不新建索引。
+  groupStatsService = new GroupStatsService(knowledgeSearchService)
+  setLocalGroupStatsService(groupStatsService)
  // 图片文字索引覆盖度是**独立覆盖维度**：接到 search_messages 的 tool result 上，
  // 让 Query Agent 在图片索引没做完时不能凭 0 条证据断言"没有"。
  localQueryApiService.setImageTextCoverageProvider(() =>
@@ -830,11 +882,6 @@ app.whenReady().then(async () => {
      if (!window.isDestroyed()) window.webContents.send('voice:modelProgress', status)
    }
  })
-  personalWechatRuntimeManager.setProgressListener((status) => {
-    for (const window of BrowserWindow.getAllWindows()) {
-      if (!window.isDestroyed()) window.webContents.send('wechat-personal:runtimeProgress', status)
-    }
-  })
  protocol.handle('wxe-media', async (request) => {
    const filePath = videoAssetService?.pathForUrl(request.url)
    if (!filePath) return new Response('Not found', { status: 404 })
@@ -890,8 +937,6 @@ app.whenReady().then(async () => {
  // 设置应用程序用户模型 ID
  electronApp.setAppUserModelId('com.tracememo.app')

-  if (process.platform === 'darwin') app.dock?.setIcon(appIconPath)
-
  // 开发环境中默认使用 F12 打开或关闭 DevTools
  // 生产环境中忽略 CommandOrControl + R
  // 参见 https://github.com/alex8088/electron-toolkit/tree/master/packages/utils
@@ -997,20 +1042,120 @@ app.whenReady().then(async () => {
          return { success: false, error: '应用正在退出，数据库连接已取消', monitoring: false }
        }
        const wcdb4Client = nextWechatDb.getWcdb4Client()
+        /**
+         * 正式 MessageListener（Observation Mode 接入）。
+         *
+         * 本轮**只监听、回读、规范化、去重、统计**，不触发任何业务：
+         * 不匹配关键词、不判断 @我、不出日报、不回复、不调 AI / Agent、不发送。
+         * 下一层的 Trigger 由后续任务接入。
+         */
+        const messageListener = new MessageListenerService(wcdb4Client)
+        /**
+         * Automation v1（@我生成日报）。
+         *
+         * `isListening` 要等下面 `startMonitor` 有结果才知道，所以先用闭包变量占位。
+         */
+        let automationListening = false
+        initAutomationService(wcdb4Client, { isListening: () => automationListening })
+
+        /**
+         * 旧「定时日报任务」迁移所需的会话解析器。
+         *
+         * **必须在数据库就绪之后注入**：旧 `task.group` 可能是会话 md5、群名或 roomId
+         * 三种形态，只有 `resolveMd5` 能收敛成稳定会话 id。
+         * 注入前若渲染层已经读过一次规则（`automation:listRules`），迁移会**整体推迟**
+         * 而不落盘 —— 注入这一步会立刻补跑（见 `AutomationRuleStore.setLegacyConversationResolver`）。
+         *
+         * 这样设计的原因：拿不到解析器时如果把旧数据判成「无法无损映射」，
+         * 用户的定时日报会被**无辜停用**，而事实只是"数据库还没打开"。
+         */
+        automationRuleStore.setLegacyConversationResolver((raw) => {
+          const key = String(raw || '').trim()
+          if (!key) return undefined
+          // 已经是稳定会话 id 就直接用，不查库。
+          if (key.endsWith('@chatroom')) return key
+          try {
+            const username = String(chat.resolveMd5(key)?.m_nsUsrName || '').trim()
+            return username || undefined
+          } catch {
+            return undefined
+          }
+        })
+        /**
+         * 退群通知的唯一通路：**退群监控只负责产生事件，自动化负责发送**。
+         *
+         * `handleGroupExit` 自己吞掉异常，这里再兜一层 `.catch` 是防御性的 ——
+         * 未捕获的 rejection 会污染监控循环，而退群事实早就已经记录成功了。
+         */
+        groupExitMonitorService.setGroupExitHandler((event) => {
+          try {
+            return getAutomationService()
+              .handleGroupExit(event)
+              .catch((error) => {
+                console.warn(
+                  `[Automation] handleGroupExit failed: ${
+                    error instanceof Error ? error.message : String(error)
+                  }`
+                )
+              })
+          } catch (error) {
+            console.warn(
+              `[Automation] 退群通知未初始化，已跳过本次事件: ${
+                error instanceof Error ? error.message : String(error)
+              }`
+            )
+            return undefined
+          }
+        })
+        messageListener.onMessage((message) => {
+          // 日志只允许出现「类型 / 群或私聊 / 是否自己发 / @ 数量」这类不可逆标识，
+          // 不含 wxid、昵称、群名、正文与 source。
+          console.log(
+            `[MessageListener] incoming messageType=${message.messageType}` +
+              ` group=${message.isGroup} self=${message.isSelf}` +
+              ` mentions=${message.mentionTargets.length}`
+          )
+          // Automation 自带 isSelf / cooldown / 同消息幂等三重闸，不会形成回复循环。
+          // 用 try 包住同步那一段：`getAutomationService()` 在未初始化时会抛，
+          // 而 `.catch()` 只能接住异步拒绝 —— 漏了这层就会把异常抛进 MessageListener 的投递循环。
+          try {
+            void getAutomationService()
+              .handleMessage(message)
+              .catch((error) => {
+                console.warn(
+                  `[Automation] handleMessage failed: ${
+                    error instanceof Error ? error.message : String(error)
+                  }`
+                )
+              })
+          } catch (error) {
+            console.warn(
+              `[Automation] 未初始化，已跳过本次消息: ${
+                error instanceof Error ? error.message : String(error)
+              }`
+            )
+          }
+        })
        const sessions = await wcdb4Client.getSessionsAsync({ hydrateDisplayNames: false })
        configureRecallProtection(wcdb4Client, resolvedRoot, settings.recallProtectionEnabled)
        voiceService = new VoiceService(wcdb4Client, resolvedRoot)
        voiceRecognition?.connect(voiceService, resolvedRoot)
        stickerService = new StickerService(wcdb4Client)
        videoAssetService = new VideoAssetService(wcdb4Client)
-        const monitoring = await wcdb4Client.startMonitor((type, json) => {
+        const monitoring = await wcdb4Client.startMonitor((event) => {
          wcdb4Client.invalidateSessionCache()
-          groupExitMonitorService.notifyDatabaseChanged(json)
-          recallArchiveMonitor?.handleDatabaseChange(json)
+          // 正式 MessageListener：只做 coalesce + 有界回读 + dedup + 投递，不触发业务。
+          // v2 事件带 sessionId ⇒ 回读不再依赖「最近活跃会话」。
+          messageListener.handleNativeChange(event)
+          groupExitMonitorService.notifyDatabaseChanged(event.raw)
+          recallArchiveMonitor?.handleDatabaseChange(event.raw)
          for (const window of BrowserWindow.getAllWindows()) {
-            if (!window.isDestroyed()) window.webContents.send('wcdb-change', { type, json })
+            if (!window.isDestroyed()) {
+              window.webContents.send('wcdb-change', { type: event.type, json: event.raw })
+            }
          }
        })
+        automationListening = monitoring === true
        void groupExitMonitorService.start(monitoring)
        const recentSession = sessions[0]
        if (recentSession?.username) {
@@ -1152,7 +1297,54 @@ app.whenReady().then(async () => {
  )
  // 日报等系统动作仍复用现有发送服务；普通聊天不再暴露这个入口。
  ipcMain.handle('wechat-personal:send', async (_, request: PersonalWechatSendRequest) => {
-    if (request.type !== 'voice' || String(request.fromId || '').trim()) {
+    // 项目规则：**所有发送都必须经过 WechatActionGateway**（审计 + 幂等 + 同一个 Send Log）。
+    // 手动发送在改造前从这里直连 `PersonalWechatSendService`，于是完全不留痕 ——
+    // 排查「消息到底发没发出去」时恰好缺的就是那份证据。
+    // 普通手动发送保持 triggerType=user；带 postfixText 的日报图片由 Gateway 编排成
+    // 两个 automation action，以复用同一条 3s 队列，并严格在图片 sent 后才发后置词。
+    if (request.type === 'text' || request.type === 'image') {
+      const to = String(request.to || '').trim()
+      const recipient = {
+        type:
+          request.isGroup || to.endsWith('@chatroom') ? ('group' as const) : ('contact' as const),
+        id: to
+      }
+      if (request.type === 'image' && request.postfixText !== undefined) {
+        const sequence = await wechatActionGateway.executeReportImageSequence({
+          recipient,
+          imagePath: String(request.filePath || ''),
+          postfixText: request.postfixText
+        })
+        const imageResult = toPersonalWechatSendResult(
+          sequence.image,
+          await currentPersonalWechatSenderStatus()
+        )
+        if (!imageResult.success || !sequence.postfix) return imageResult
+        return {
+          ...imageResult,
+          postfixSent: sequence.postfix.status === 'sent',
+          ...(sequence.postfix.status === 'sent'
+            ? {}
+            : { postfixError: sequence.postfix.reason || '后置词发送失败' })
+        }
+      }
+      const action = await wechatActionGateway.execute({
+        origin: 'user_manual',
+        purpose: request.type === 'text' ? 'manual_text' : 'manual_image',
+        triggerType: 'user',
+        recipient,
+        content:
+          request.type === 'text'
+            ? { type: 'text', text: String(request.text || '') }
+            : { type: 'image', path: String(request.filePath || '') }
+      })
+      // 返回契约保持 `PersonalWechatSendResult`，界面判读不用改。
+      return toPersonalWechatSendResult(action, await currentPersonalWechatSenderStatus())
+    }
+
+    // 语音仍走既有分支：它有自己的网关入口 `wechat-personal:sendGeneratedTtsVoice`，
+    // 且需要先解析当前账号 wxid 才能编码。
+    if (String(request.fromId || '').trim()) {
      return personalWechatSendService.send(request)
    }
    let fromId = ''
@@ -1469,31 +1661,153 @@ app.whenReady().then(async () => {
    return snapshot
  })

+  /**
+   * 群员统计（单群）。
+   *
+   * 时间单位刻意用 epoch **毫秒**：这一路完全走 Knowledge，而 Knowledge 内部口径就是毫秒。
+   * 沿用聊天消息的秒级口径会在 service 内部凭空多出一次换算 —— 而单位换错是**静默读 0 条**，
+   * 不会报错。
+   */
+  ipcMain.handle(
+    'group-stats:getMemberStats',
+    async (_, request: { userMd5?: unknown; startTime?: unknown; endTime?: unknown }) => {
+      if (!groupStatsService) throw new Error('群员统计服务尚未就绪')
+      const userMd5 = String(request?.userMd5 || '').trim()
+      if (!userMd5) throw new Error('缺少会话标识')
+      const startTime = Number(request?.startTime)
+      const endTime = Number(request?.endTime)
+      if (!Number.isFinite(startTime) || !Number.isFinite(endTime) || endTime < startTime) {
+        throw new Error('统计时间范围无效')
+      }
+      return groupStatsService.getMemberStats({ userMd5, startTime, endTime })
+    }
+  )
+
  ipcMain.handle('group-exit-monitor:getState', () => groupExitMonitorService.getState())
  ipcMain.handle('group-exit-monitor:setEnabled', (_, enabled: boolean) =>
    groupExitMonitorService.setEnabled(enabled === true)
  )
-  ipcMain.handle(
-    'group-exit-monitor:setGroups',
-    (_, roomIds: string[], notificationRoomIds?: string[]) =>
-      groupExitMonitorService.setMonitoredRoomIds(
-        Array.isArray(roomIds) ? roomIds : [],
-        Array.isArray(notificationRoomIds) ? notificationRoomIds : []
-      )
-  )
-  ipcMain.handle('group-exit-monitor:setTemplate', (_, template: unknown) =>
-    groupExitMonitorService.setNotificationTemplate(template)
+  /**
+   * 保存**监控范围**。
+   *
+   * 只有一个参数：旧版第二个参数（通知群聊）已经迁到「自动化 → 退群通知」，
+   * 退群监控不再持有任何通知配置。
+   */
+  ipcMain.handle('group-exit-monitor:setGroups', (_, roomIds: string[]) =>
+    groupExitMonitorService.setMonitoredRoomIds(Array.isArray(roomIds) ? roomIds : [])
  )
  ipcMain.handle('group-exit-monitor:checkNow', () => groupExitMonitorService.checkNow())
+  /**
+   * 按群查退群事件（档案合并展示用）。
+   *
+   * 时间参数是 epoch **毫秒**，与事件的 `detectedAt` 同口径。
+   */
+  ipcMain.handle('group-exit-monitor:listEvents', (_, query: unknown) => {
+    const input = (query || {}) as {
+      roomId?: unknown
+      sinceMs?: unknown
+      untilMs?: unknown
+      limit?: unknown
+    }
+    return groupExitMonitorService.listEvents({
+      roomId: typeof input.roomId === 'string' ? input.roomId : undefined,
+      sinceMs: Number(input.sinceMs),
+      untilMs: Number(input.untilMs),
+      limit: Number(input.limit)
+    })
+  })
  ipcMain.handle('group-exit-monitor:clearEvents', () => groupExitMonitorService.clearEvents())
-  ipcMain.handle('group-exit-monitor:resendEvent', (_, eventId: string) =>
-    groupExitMonitorService.resendEvent(eventId)
-  )
  ipcMain.handle('group-exit-monitor:markRead', (_, readAt?: number) =>
    groupExitMonitorService.markRead(readAt)
  )
  ipcMain.handle('wechat-action-log:list', () => wechatActionLogService.list())

+  /**
+   * Automation v1（@我生成日报）。
+   *
+   * 规则与执行日志都是纯文件存储，**不依赖数据库**，所以这两组 handler 在数据库
+   * 解锁前也可以安全调用；只有 `getStatus` / `listGroups` 需要会话数据，因此用
+   * `tryAutomationService()` 兜底，避免渲染层在启动阶段拿到一个 rejected promise。
+   */
+  const tryAutomationService = (): ReturnType<typeof getAutomationService> | null => {
+    try {
+      return getAutomationService()
+    } catch {
+      return null
+    }
+  }
+  ipcMain.handle('automation:listRules', () => automationRuleStore.listRules())
+  ipcMain.handle('automation:createRule', (_, draft: unknown) =>
+    automationRuleStore.createRule(draft)
+  )
+  ipcMain.handle('automation:updateRule', (_, input: unknown) => {
+    const payload = (input || {}) as { id?: unknown; draft?: unknown }
+    return automationRuleStore.updateRule(String(payload.id || ''), payload.draft) ?? null
+  })
+  ipcMain.handle('automation:deleteRule', (_, id: string) => automationRuleStore.deleteRule(id))
+  ipcMain.handle('automation:setRuleEnabled', (_, input: unknown) => {
+    const payload = (input || {}) as { id?: unknown; enabled?: unknown }
+    return (
+      automationRuleStore.setRuleEnabled(String(payload.id || ''), payload.enabled === true) ?? null
+    )
+  })
+  ipcMain.handle('automation:listExecutions', (_, query: unknown) => {
+    const input = (query || {}) as { limit?: unknown }
+    return automationExecutionLogService.list({ limit: Number(input.limit) })
+  })
+  ipcMain.handle('automation:clearExecutions', () => automationExecutionLogService.clear())
+  ipcMain.handle('automation:listGroups', () => tryAutomationService()?.listGroups() ?? [])
+  /**
+   * 保存「退群通知」规则（singleton upsert）。
+   *
+   * 单独一个通道而不是复用 `updateRule`：这条规则是系统规则，
+   * 保存语义是"存在即更新、不存在即创建"，且 id 固定 —— 不允许渲染层自己拼 id。
+   */
+  ipcMain.handle('automation:saveLeaveNotificationRule', (_, draft: unknown) =>
+    automationRuleStore.saveLeaveNotificationRule(draft)
+  )
+  /**
+   * 「指定好友」的可选项。
+   *
+   * 过滤（群聊 / 公众号 / 文件传输助手 / 自己）在 main 侧完成 ——
+   * 只有这里知道当前登录账号是谁，渲染层再筛一遍必然会漂移。
+   */
+  ipcMain.handle('automation:listSendableContacts', () => {
+    try {
+      const selfWxid = String(chat.getSelfAccountInfo()?.wxid || '')
+      return filterSendableFriendContacts(chat.listContacts(), selfWxid).map((contact) => ({
+        id: String(contact.m_nsUsrName || ''),
+        name: leaveNotificationContactDisplayName(contact)
+      }))
+    } catch (error) {
+      console.warn(
+        `[Automation] 读取可选联系人失败: ${error instanceof Error ? error.message : String(error)}`
+      )
+      return []
+    }
+  })
+  ipcMain.handle('automation:getStatus', async () => {
+    const service = tryAutomationService()
+    if (service) return service.getStatus()
+    // 数据库尚未就绪：如实返回「未监听 + 能力未知」，而不是假装一切正常。
+    const todayStart = new Date()
+    todayStart.setHours(0, 0, 0, 0)
+    const counts = automationExecutionLogService.countSince(todayStart.getTime())
+    return {
+      listening: false,
+      listeningDegraded: true,
+      todayExecutions: counts.total,
+      todaySuccesses: counts.success,
+      sendCapability: {
+        supported: false,
+        ready: false,
+        canSendText: false,
+        canSendImage: false,
+        message: '数据库尚未就绪，暂时无法获知微信发送能力'
+      }
+    }
+  })
+
  ipcMain.handle('db:search', (_, keyword: string) => chat.searchMessages(keyword))
  ipcMain.handle(
    'knowledge:search',
@@ -1713,44 +2027,58 @@ app.whenReady().then(async () => {
  })

  ipcMain.handle('wechat-personal:getSendCapability', () => getPersonalWechatSendCapability())
-  ipcMain.handle('scheduled-report:list', () => scheduledReportService.listTasks())
-  ipcMain.handle('scheduled-report:listExecutions', (_, taskId?: string) =>
-    scheduledReportService.listExecutions(taskId)
+
+  /**
+   * 定时日报（Automation 的 `scheduled_report` 规则类型）。
+   *
+   * 规则本身的 CRUD 走上面的 `automation:*`；这里只补三块**专属**能力：
+   * 1. 「立即执行」—— 走 `AutomationService.executeScheduledRule(ruleId,'manual')`，
+   *    与 scheduler 触发**同一条链路**（生成 → 落库 → 解析目标 → Gateway 发图）；
+   * 2. 微信异常通知 —— 定时日报失败时的 Agent Hub 推送，随功能保留；
+   * 3. 旧执行记录**只读存档** —— 旧记录无法无损转换，原样保留给 UI 展示。
+   */
+  ipcMain.handle('automation:runScheduledReportRule', async (_, ruleId: string) => {
+    const service = tryAutomationService()
+    if (!service) return { success: false, error: '数据库尚未就绪，暂时无法执行定时日报' }
+    const outcome = await service.executeScheduledRule(String(ruleId || ''), { trigger: 'manual' })
+    if (outcome.executed) {
+      return { success: outcome.status !== 'failed', data: outcome }
+    }
+    // 「没执行」的每一种原因都要能翻译成用户看得懂的一句话 —— 否则点了按钮没反应，无从排查。
+    const skipMessages: Record<string, string> = {
+      missing_rule: '规则 id 无效',
+      rule_not_found: '未找到这条定时日报规则',
+      disabled: '这条定时日报已停用，请先启用再执行',
+      target_needs_review: '这条规则的发送目标需要重新选择后才能执行',
+      in_flight: '这条规则正在执行中，请稍后再试',
+      store_error: '读取规则失败，请稍后再试'
+    }
+    return {
+      success: false,
+      error: skipMessages[outcome.reason || ''] || '定时日报未执行',
+      data: outcome
+    }
+  })
+  ipcMain.handle('automation:listScheduledReportLegacyExecutions', (_, ruleId?: string) =>
+    scheduledReportService.listLegacyExecutions(ruleId)
  )
-  ipcMain.handle('scheduled-report:getNotificationSettings', () =>
+  ipcMain.handle('automation:getScheduledReportNotificationSettings', () =>
    scheduledReportService.getNotificationSettings()
  )
-  ipcMain.handle('scheduled-report:setNotificationEnabled', (_, enabled: boolean) =>
+  ipcMain.handle('automation:getScheduledReportNotificationCapability', () =>
+    scheduledReportService.checkNotificationCapability()
+  )
+  ipcMain.handle('automation:setScheduledReportNotificationEnabled', (_, enabled: boolean) =>
    scheduledReportService.setNotificationEnabled(Boolean(enabled))
  )
-  ipcMain.handle('scheduled-report:create', (_, request: ScheduledReportCreateInput) =>
-    scheduledReportService.createTask(request)
-  )
-  ipcMain.handle(
-    'scheduled-report:update',
-    (_, taskId: string, request: ScheduledReportUpdateInput) =>
-      scheduledReportService.updateTask(taskId, request)
-  )
-  ipcMain.handle('scheduled-report:delete', (_, taskId: string) =>
-    scheduledReportService.deleteTask(taskId)
-  )
-  ipcMain.handle('scheduled-report:setEnabled', (_, taskId: string, enabled: boolean) =>
-    scheduledReportService.setTaskEnabled(taskId, Boolean(enabled))
-  )
-  ipcMain.handle('scheduled-report:runNow', (_, taskId: string) =>
-    scheduledReportService.runScheduledReportNow(taskId)
-  )
-  ipcMain.handle('scheduled-report:retrySend', (_, executionId: string) =>
-    scheduledReportService.retryScheduledReportSend(executionId)
-  )
-  ipcMain.handle('scheduled-report:testErrorNotification', (_, taskId: string) => {
+  ipcMain.handle('automation:testScheduledReportErrorNotification', (_, ruleId: string) => {
    if (!isTruthyDebugFlag(import.meta.env.VITE_SCHEDULED_REPORT_DEBUG)) {
      return Promise.resolve({
        success: false,
        error: '调试测试按钮未开启，请在 .env 中设置 VITE_SCHEDULED_REPORT_DEBUG=true。'
      })
    }
-    return scheduledReportService.testScheduledReportErrorNotification(taskId)
+    return scheduledReportService.testScheduledReportErrorNotification(ruleId)
  })

  ipcMain.handle('report:reveal', async (_, filePath: string) => {
@@ -2216,11 +2544,13 @@ app.whenReady().then(async () => {
    if (client) {
      voiceService = new VoiceService(client, client.getAccountRoot())
      voiceRecognition?.connect(voiceService, client.getAccountRoot())
-      const monitoring = await client.startMonitor((type, json) => {
+      const monitoring = await client.startMonitor((event) => {
        client.invalidateSessionCache()
-        groupExitMonitorService.notifyDatabaseChanged(json)
+        groupExitMonitorService.notifyDatabaseChanged(event.raw)
        for (const window of BrowserWindow.getAllWindows()) {
-          if (!window.isDestroyed()) window.webContents.send('wcdb-change', { type, json })
+          if (!window.isDestroyed()) {
+            window.webContents.send('wcdb-change', { type: event.type, json: event.raw })
+          }
        }
      })
      void groupExitMonitorService.start(monitoring)
@@ -2345,42 +2675,16 @@ app.whenReady().then(async () => {
    agentHubService.clearConversations()
    return { success: true }
  })
-  ipcMain.handle('wechat-personal:getStatus', () => personalWechatSendService.getStatus())
-  ipcMain.handle('wechat-personal:getKeepProcess', () =>
-    personalWechatSendService.getKeepOneBotProcess()
-  )
-  ipcMain.handle('wechat-personal:setKeepProcess', (_, keep: boolean) =>
-    personalWechatSendService.setKeepOneBotProcess(Boolean(keep))
-  )
+  ipcMain.handle('wechat-personal:getStatus', () => currentPersonalWechatSenderStatus())
  ipcMain.handle('wechat-personal:checkStatus', (_, port?: string) =>
    personalWechatSendService.checkWindowsStatus(port)
  )
  ipcMain.handle('wechat-personal:checkVoiceEnvironment', () =>
    personalWechatVoiceEnvironmentService.check()
  )
-  ipcMain.handle('wechat-personal:installPilk', async () => {
-    const result = await personalWechatVoiceEnvironmentService.installPilk()
-    if (!result.success || process.platform !== 'darwin') return result
-
-    const senderStatus = await personalWechatSendService.getStatus()
-    if (!senderStatus.oneBotPid) return result
-    try {
-      const restartedStatus = await personalWechatSendService.restartRuntime()
-      return {
-        ...result,
-        restarted: restartedStatus.state !== 'error',
-        ...(restartedStatus.state === 'error'
-          ? { restartError: restartedStatus.error || restartedStatus.message }
-          : {})
-      }
-    } catch (error) {
-      return {
-        ...result,
-        restarted: false,
-        restartError: error instanceof Error ? error.message : String(error)
-      }
-    }
-  })
+  ipcMain.handle('wechat-personal:installPilk', () =>
+    personalWechatVoiceEnvironmentService.installPilk()
+  )
  ipcMain.handle('wechat-personal:openVoicePythonDownload', async () => {
    try {
      await shell.openExternal('https://www.python.org/downloads/macos/')
@@ -2397,23 +2701,20 @@ app.whenReady().then(async () => {
      return { success: false, error: '无法打开 FFmpeg 安装页面' }
    }
  })
-  ipcMain.handle('wechat-personal:getRuntimeStatus', () => personalWechatRuntimeManager.getStatus())
-  ipcMain.handle('wechat-personal:downloadRuntime', () => personalWechatRuntimeManager.download())
-  ipcMain.handle('wechat-personal:cancelRuntimeDownload', () => ({
-    success: personalWechatRuntimeManager.cancelDownload()
-  }))
-  ipcMain.handle('wechat-personal:removeRuntime', async () => {
-    await personalWechatSendService.terminate(true)
-    return personalWechatRuntimeManager.remove()
+  ipcMain.handle('wechat-personal:rebind', async () => {
+    if (process.platform !== 'darwin') return personalWechatSendService.rebind()
+
+    const result = await macWechatRuntimeManager.bind()
+    const status = await macWechatRuntimeManager.buildSenderStatus()
+    if (result.ok) return status
+
+    return {
+      ...status,
+      state: 'error',
+      message: result.message,
+      error: result.message
+    }
  })
-  ipcMain.handle('wechat-personal:openRuntimeDirectory', async () => {
-    const status = await personalWechatRuntimeManager.getStatus()
-    const directory = status.directory || personalWechatRuntimeManager.directory
-    await fsPromises.mkdir(directory, { recursive: true })
-    const error = await shell.openPath(directory)
-    return error ? { success: false, error } : { success: true }
-  })
-  ipcMain.handle('wechat-personal:rebind', () => personalWechatSendService.rebind())
  ipcMain.handle(
    'wechat-personal:sendGeneratedTtsVoice',
    async (_, request: PersonalWechatGeneratedTtsVoiceRequest) => {
@@ -2441,7 +2742,7 @@ app.whenReady().then(async () => {
        action.sendResult && typeof action.sendResult === 'object'
          ? (action.sendResult as PersonalWechatSendResult)
          : undefined
-      const status = sendResult?.status || (await personalWechatSendService.getStatus())
+      const status = sendResult?.status || (await currentPersonalWechatSenderStatus())
      return { action, status }
    }
  )
@@ -2516,9 +2817,21 @@ app.on('before-quit', (event) => {
      chat.closeChatDbForQuit().catch(() => false),
      voiceRecognition?.dispose().catch(() => undefined),
      knowledgeSearchService?.dispose().catch(() => undefined),
-      personalWechatSendService.terminate().catch((error) => {
-        console.warn('[Shutdown] personal WeChat sender cleanup failed:', error)
-      })
+      /*
+       * Deliberately do NOT stop the macOS native runtime here.
+       *
+       * The host is designed to outlive TraceMemo ("保留发送能力进程"): keeping
+       * it alive preserves the bound frida session, so relaunching the app —
+       * including every dev-mode restart — re-adopts it and works immediately
+       * instead of forcing the user to bind WeChat again.
+       *
+       * Stopping it on quit threw that away and, worse, gave teardown a chance
+       * to hang mid-unload while the app was already exiting. The host has its
+       * own lifecycle instead: it exits itself once WeChat is gone ("stale" ->
+       * self-shutdown), and it can be restarted explicitly from the settings
+       * card ("重新加载组件").
+       */
+      Promise.resolve()
    ])
    if (!nativeCallsDrained) {
      console.warn('[Shutdown] WCDB async calls did not fully drain before quit')
@@ -2576,11 +2889,12 @@ function buildTrayMenu(): Menu {
 function setupTray(): void {
  if (tray) return
  try {
-    const image = nativeImage.createFromPath(appIconPath)
-    const traySize = process.platform === 'darwin' ? 20 : 24
-    const trayImage = image.isEmpty()
-      ? nativeImage.createEmpty()
-      : image.resize({ width: traySize, height: traySize, quality: 'best' })
+    const trayImage = createTrayImage(
+      process.platform,
+      appIconPath,
+      { oneX: trayTemplateIconPath, twoX: trayTemplateRetinaIconPath },
+      nativeImage
+    )
    tray = new Tray(trayImage)
    tray.setToolTip('TraceMemo')
    // macOS may show a Tray context menu on a primary click when it is set
@@ -4,6 +4,7 @@ import type {
  KnowledgeImageOcrState,
  KnowledgeAttachmentMetadata,
  KnowledgeEvidence,
+  KnowledgeMemberStatsResult,
  KnowledgeMessageKind,
  KnowledgePassProgress,
  KnowledgeRuntimeState,
@@ -498,6 +499,37 @@ export class KnowledgeSearchService {
    }
  }

+  /**
+   * 单群发言聚合（群员统计的数据来源）。
+   *
+   * 这一层只**如实**返回引擎能给出的东西（含 `indexLatestAt`），不做追赶决策 ——
+   * 「要不要等索引、要不要把结果标成不完整」是产品判断，属于 GroupStatsService。
+   * 同时返回一次 `sourceLatestAt`（同步、零额外 WCDB 调用），让调用方一次拿到
+   * freshness 的两个口径，不必再发一次 status 请求。
+   */
+  async memberStats(request: {
+    conversationId: string
+    startTime: number
+    endTime: number
+  }): Promise<{ result: KnowledgeMemberStatsResult | null; sourceLatestAt: number | null }> {
+    const sourceLatestAt = this.sourceLatestAt()
+    const accountId = this.currentAccountId()
+    if (!accountId) return { result: null, sourceLatestAt }
+    try {
+      const result = await this.service.memberStats({
+        accountId,
+        fts: DEFAULT_KNOWLEDGE_FTS_CONFIG,
+        conversationId: request.conversationId,
+        startTime: request.startTime,
+        endTime: request.endTime
+      })
+      return { result, sourceLatestAt }
+    } catch (error) {
+      console.warn('[Knowledge] member stats failed:', error)
+      return { result: null, sourceLatestAt }
+    }
+  }
+
  /**
   * 源数据最新活跃时间（epoch ms）。派生索引看到不源数据，freshness 判定由它 + `indexLatestAt` 组成。
   */
@@ -5,6 +5,8 @@ import type {
  KnowledgeIndexProgress,
  KnowledgeIndexRequest,
  KnowledgeIndexResult,
+  KnowledgeMemberStatsRequest,
+  KnowledgeMemberStatsResult,
  KnowledgeRuntimeStatus,
  KnowledgeSearchRequest,
  KnowledgeSearchResult,
@@ -53,6 +55,13 @@ export class KnowledgeService {
    return this.worker.highWaterMarks({ ...request, databaseRoot: this.databaseRoot })
  }

+  /** 单个会话内「按发送者聚合」的发言统计（群员统计用）。 */
+  memberStats(
+    request: Omit<KnowledgeMemberStatsRequest, 'databaseRoot'>
+  ): Promise<KnowledgeMemberStatsResult> {
+    return this.worker.memberStats({ ...request, databaseRoot: this.databaseRoot })
+  }
+
  /** 只中止正在跑的索引任务；查询请求不受影响。 */
  cancelIndex(): Promise<boolean> {
    return this.worker.cancelActiveIndex()
@@ -13,6 +13,7 @@ import type {
  KnowledgeIndexProgress,
  KnowledgeIndexRequest,
  KnowledgeIndexResult,
+  KnowledgeMemberStatsResult,
  KnowledgeRuntimeStatus,
  KnowledgeNormalizedMessage,
  KnowledgeQuery,
@@ -317,6 +318,93 @@ export class KnowledgeStore {
    }
  }

+  /**
+   * 单个会话在时间窗内的「按发送者聚合」统计。
+   *
+   * 只回聚合结果、不回消息正文：群员统计只需要「谁说了几条、最后一条是什么时候」，
+   * 把消息逐条搬到主进程再统计会把几十万行推过 IPC 边界。
+   *
+   * 三条硬规则**全部由 SQL 保证**，不指望调用方记得：
+   * 1. `kind <> 'system'`：系统消息不是任何成员的发言（微信侧 10000/10002 在建库时
+   *    已归一为 `kind = 'system'`，见 `knowledge-search-service` 的 kind 映射）；
+   * 2. `sender_id IS NULL` 的行不归给任何人，只计入 `unattributedMessages` ——
+   *    硬塞给某个成员会让「未发言」名单出现错误否定；
+   * 3. 时间窗口是**闭区间**，单位 **epoch 毫秒**，与 knowledge 内部口径一致，
+   *    不经过 WCDB 的秒级边界（跨错单位会静默读到 0 条）。
+   */
+  memberStats(request: {
+    conversationId: string
+    startTime: number
+    endTime: number
+  }): KnowledgeMemberStatsResult {
+    const { conversationId, startTime, endTime } = request
+    const indexLatestAt = this.readIndexLatestAt()
+    const empty: KnowledgeMemberStatsResult = {
+      conversationId,
+      totalMessages: 0,
+      senders: [],
+      unattributedMessages: 0,
+      excludedSystemMessages: 0,
+      earliestMessageTime: null,
+      indexLatestAt
+    }
+    if (!conversationId) return empty
+
+    // 用 `(conversation_id, create_time)` 索引直接命中：这是本查询唯一的访问路径，
+    // 写成全表扫描等价于把单群统计的 26ms 变成 10s。
+    const scope = 'conversation_id = ? AND create_time >= ? AND create_time <= ?'
+    const args = [conversationId, startTime, endTime]
+
+    const count = (extra: string): number => {
+      const row = this.database
+        .prepare(`SELECT COUNT(*) AS n FROM knowledge_messages WHERE ${scope} AND ${extra}`)
+        .get(...args) as DbRow | undefined
+      return Number(row?.n) || 0
+    }
+
+    const rows = asRows(
+      this.database
+        .prepare(
+          `SELECT sender_id,
+                  COUNT(*) AS message_count,
+                  MAX(create_time) AS last_message_time
+             FROM knowledge_messages
+            WHERE ${scope} AND kind <> 'system' AND sender_id IS NOT NULL
+            GROUP BY sender_id
+            ORDER BY message_count DESC, last_message_time DESC`
+        )
+        .all(...args)
+    )
+
+    const senders: KnowledgeMemberStatsResult['senders'] = []
+    for (const row of rows) {
+      const senderId = String(row.sender_id ?? '').trim()
+      if (!senderId) continue
+      senders.push({
+        senderId,
+        messageCount: Number(row.message_count) || 0,
+        lastMessageTime: Number(row.last_message_time) || 0
+      })
+    }
+
+    // 窗口内最早一条消息：**不过滤 kind** —— 群的第一条常常是建群通知，
+    // 那才是用户认知里的「这个群第一条消息」。选「全部」时用它显示真实起点。
+    const earliestRow = this.database
+      .prepare(`SELECT MIN(create_time) AS m FROM knowledge_messages WHERE ${scope}`)
+      .get(...args) as DbRow | undefined
+    const earliestValue = Number(earliestRow?.m)
+
+    return {
+      conversationId,
+      totalMessages: count("kind <> 'system'"),
+      senders,
+      unattributedMessages: count("kind <> 'system' AND sender_id IS NULL"),
+      excludedSystemMessages: count("kind = 'system'"),
+      earliestMessageTime: Number.isFinite(earliestValue) && earliestValue > 0 ? earliestValue : null,
+      indexLatestAt
+    }
+  }
+
  /**
   * 索引已经覆盖到的源数据时间（epoch ms）——「索引更新到哪」的权威口径。
   *
@@ -6,6 +6,8 @@ import type {
  KnowledgeIndexProgress,
  KnowledgeIndexRequest,
  KnowledgeIndexResult,
+  KnowledgeMemberStatsRequest,
+  KnowledgeMemberStatsResult,
  KnowledgeRuntimeStatus,
  KnowledgeSearchRequest,
  KnowledgeSearchResult,
@@ -19,6 +21,7 @@ type WorkerResult =
  | KnowledgeCapacityPreflight
  | KnowledgeSearchResult
  | KnowledgeRuntimeStatus
+  | KnowledgeMemberStatsResult
  | { marks: Record<string, number> }
  | { removed: true }
 type PendingRequest = {
@@ -76,6 +79,11 @@ export class KnowledgeWorkerHost {
    )
  }

+  /** 单个会话内「按发送者聚合」的发言统计（群员统计用，只回聚合不回正文）。 */
+  memberStats(payload: KnowledgeMemberStatsRequest): Promise<KnowledgeMemberStatsResult> {
+    return this.request('memberStats', payload) as Promise<KnowledgeMemberStatsResult>
+  }
+
  /** 只中止正在跑的索引任务，返回是否真的有任务被中止。 */
  async cancelActiveIndex(): Promise<boolean> {
    const target = this.activeIndexRequestId
@@ -1,6 +1,8 @@
 import type {
  KnowledgeCapacityPreflightRequest,
  KnowledgeIndexRequest,
+  KnowledgeMemberStatsRequest,
+  KnowledgeMemberStatsResult,
  KnowledgeRuntimeStatus,
  KnowledgeSearchRequest,
  KnowledgeStatusRequest,
@@ -187,6 +189,38 @@ async function handleHighWater(
  })
 }

+/**
+ * 群员统计的按发送者聚合。
+ *
+ * 与 `handleSearch` 同样先判库是否存在：**「还没建索引」是正常状态，不是故障**，
+ * 返回空结果而不是抛错，让上层能稳定地区分「没人发言」与「索引不存在」。
+ */
+async function handleMemberStats(
+  request: KnowledgeWorkerRequest,
+  payload: KnowledgeMemberStatsRequest
+): Promise<void> {
+  const path = getKnowledgeDatabasePath(payload.databaseRoot, payload.accountId)
+  if (!existsSync(path)) {
+    const unavailable: KnowledgeMemberStatsResult = {
+      conversationId: payload.conversationId,
+      totalMessages: 0,
+      senders: [],
+      unattributedMessages: 0,
+      excludedSystemMessages: 0,
+      earliestMessageTime: null,
+      indexLatestAt: null
+    }
+    send({ version: 1, type: 'result', requestId: request.requestId, payload: unavailable })
+    return
+  }
+  send({
+    version: 1,
+    type: 'result',
+    requestId: request.requestId,
+    payload: getStore(payload).memberStats(payload)
+  })
+}
+
 async function handle(request: KnowledgeWorkerRequest, messageReceivedAt: number): Promise<void> {
  try {
    if (request.type === 'cancel') {
@@ -226,6 +260,10 @@ async function handle(request: KnowledgeWorkerRequest, messageReceivedAt: number
      await handleHighWater(request, request.payload as KnowledgeStatusRequest)
      return
    }
+    if (request.type === 'memberStats') {
+      await handleMemberStats(request, request.payload as KnowledgeMemberStatsRequest)
+      return
+    }
    if (request.type === 'index') {
      await handleIndex(request, request.payload as KnowledgeIndexRequest)
      return
@@ -0,0 +1,731 @@
+import {
+  AUTOMATION_STEP_LABELS,
+  AUTOMATION_SEND_ORIGIN,
+  AUTOMATION_SEND_PURPOSE,
+  DEFAULT_REPLY_TEXT,
+  automationIdempotencyKey,
+  normalizeReplyDelaySeconds,
+  scheduledReportRangeLabel,
+  type AutomationAction,
+  type AutomationExecutionTrigger,
+  type AutomationRule,
+  type AutomationStep,
+  type AutomationStepKey,
+  type LeaveNotificationConfig,
+  type ScheduledReportAutomationConfig
+} from '../../shared/automation'
+import {
+  renderLeaveNotificationText,
+  type GroupMemberExitedEvent
+} from '../../shared/group-exit-event'
+import type { WechatActionContent, WechatActionRequest, WechatActionResult } from '../../shared/wechat-action'
+import type { SaveGeneratedReportRequest, SaveGeneratedReportResult } from '../../shared/report-history'
+import {
+  generateAgentGroupReport,
+  type AgentGroupReportRequest,
+  type AgentGroupReportResult
+} from './agent-group-report-service'
+import { wechatActionGateway } from './wechat-action-gateway'
+import { saveGeneratedReport } from '../report-history-service'
+import { getContactAvatars, resolveMd5 } from './chat-service'
+import type { AutomationTargetResolution } from './automation-wechat-target'
+
+/**
+ * AutomationActionRunner —— 把一次命中跑成**步骤序列**。
+ *
+ * 两条硬要求：
+ *
+ * 1. **每一步都有独立的状态、起止时间与耗时。** 用户层日志的全部价值就在于此：
+ *    只告诉用户「失败了」等于没说，得告诉他是「生成日报」那步坏了。
+ * 2. **失败必须切断后续。** 日报没生成出来就绝不能继续发图 —— 那会发出一条
+ *    空的 / 过期的 / 上一条的消息，比不发更糟。所以失败后剩下的步骤一律 `skipped`。
+ */
+
+/** 一步执行完毕后，后续步骤的处置方式。 */
+const STEP_ORDER: AutomationStepKey[] = ['received', 'matched', 'reply', 'report', 'send']
+
+/** 退群通知的步骤顺序（与消息型规则**不共用**）。 */
+const LEAVE_NOTIFICATION_STEP_ORDER: AutomationStepKey[] = [
+  'exit_received',
+  'exit_matched',
+  'exit_target',
+  'exit_send'
+]
+
+/**
+ * 定时日报的步骤顺序（与另外两族**不共用**）。
+ *
+ * 「生成中 / 已生成」与「确定目标 / 已发送」**刻意拆成四步**：
+ * 用户必须能一眼看出"日报确实生成了，只是没发出去" —— 合成两步就表达不了。
+ */
+const SCHEDULED_REPORT_STEP_ORDER: AutomationStepKey[] = [
+  'schedule_triggered',
+  'report_generating',
+  'report_generated',
+  'send_resolved',
+  'report_sent'
+]
+
+/**
+ * 前置步骤失败时，后续步骤的 `skipReason`。
+ *
+ * 必须写清「是因为前面那步没成」，否则用户看到一连串「已跳过」会以为是规则没配好。
+ */
+const SKIPPED_AFTER_REPLY_FAILURE = '前置步骤失败（回复确认未成功），本次不再继续。'
+const SKIPPED_AFTER_REPORT_FAILURE = '前置步骤失败（日报未生成），没有图片可发送。'
+
+/**
+ * 定时日报「生成 / 落库」失败时，后续步骤的 `skipReason`。
+ *
+ * 与 `SKIPPED_AFTER_REPORT_FAILURE` 的区别：那条说的是「@我日报没图可发」，
+ * 这条说的是「定时日报这一步就没跑起来」—— 文案分开，避免用户混淆两条链路。
+ */
+const SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE =
+  '前置步骤失败（定时日报未生成或未保存），本次未发送。'
+
+/**
+ * 规则启用了「发送日报图片」，但手上没有图片文件。
+ *
+ * 这不是「正常跳过」，也不允许退而求其次去发空路径 / 上一张旧图 / 不存在的文件 ——
+ * 发错东西比不发更糟，所以直接判失败。
+ */
+const SEND_WITHOUT_IMAGE_ERROR = '日报图片未生成，无法发送。'
+
+export interface AutomationRunInput {
+  executionId: string
+  rule: AutomationRule
+  /** 会话标识：群为 `xxx@chatroom`，私聊为 wxid。同时也是发送对象。 */
+  conversationId: string
+  isGroup: boolean
+  /** 用户可读的来源名（群名 / 昵称）。**不是** id。 */
+  sourceDisplayName: string
+}
+
+export interface AutomationRunResult {
+  steps: AutomationStep[]
+  status: 'success' | 'failed'
+  errorSummary?: string
+  pngPath?: string
+}
+
+/**
+ * 退群通知的执行输入。
+ *
+ * 目标解析与发送能力预检都由 `AutomationService` 在调用前完成，
+ * 这里只负责"把它跑成步骤序列" —— 于是**所有步骤构造只在一处**，
+ * 不会出现"服务拼一半、runner 拼一半"的裂口。
+ */
+export interface LeaveNotificationRunInput {
+  executionId: string
+  event: GroupMemberExitedEvent
+  config: LeaveNotificationConfig
+  resolution: AutomationTargetResolution
+  /** 事件来源的显示名（群名 / 「群聊」）。**不含 wxid**。 */
+  sourceDisplayName: string
+  /** 发送能力缺失时的一句话说明；有值时 `exit_send` 直接判失败。 */
+  sendBlockedReason?: string
+}
+
+export interface LeaveNotificationRunResult {
+  steps: AutomationStep[]
+  status: 'success' | 'failed'
+  errorSummary?: string
+}
+
+/**
+ * 定时日报的执行输入。
+ *
+ * 与退群通知同构：目标解析、来源显示名、发送能力预检都在 `AutomationService`
+ * 调用前算好，runner 只负责"跑成步骤序列"。
+ */
+export interface ScheduledReportRunInput {
+  executionId: string
+  rule: AutomationRule
+  config: ScheduledReportAutomationConfig
+  resolution: AutomationTargetResolution
+  /** 日报来源群的显示名（群名 / 「日报来源群」）。**不含 roomId**。 */
+  sourceDisplayName: string
+  /**
+   * 本次触发方式。
+   *
+   * 只影响**发送节流口径**：定时触发算 `automation`（纳入发送节流），
+   * 用户在页面上点「立即执行」算 `user`（与旧定时日报的手动执行一致）。
+   * 执行日志里的 `trigger` 由 `AutomationService` 单独记录，不从这里读。
+   */
+  trigger: Extract<AutomationExecutionTrigger, 'schedule' | 'manual'>
+  /** 发送能力缺失时的一句话说明；有值时 `report_sent` 直接判失败。 */
+  sendBlockedReason?: string
+}
+
+export interface ScheduledReportRunResult {
+  steps: AutomationStep[]
+  status: 'success' | 'failed'
+  errorSummary?: string
+  /**
+   * 日报是否**已经生成并落库**。
+   *
+   * 这是「生成成功但发送失败」的判据：`reportGenerated === true && status === 'failed'`
+   * 就是"日报在，只是没发出去"。日报本身已进日报历史，不会被丢掉。
+   */
+  reportGenerated: boolean
+  /** 生成出来的 PNG（仅在内存里传递，**不落执行日志**）。 */
+  pngPath?: string
+  /**
+   * 底层生成错误的**机器可读码**（例如 `NO_MESSAGES`）。
+   *
+   * 用途只有一个：让上层区分「真的失败了」和「这一天没有消息可生成」——
+   * 后者不该给用户推微信异常通知。
+   */
+  errorCode?: string
+}
+
+export interface AutomationActionRunnerDependencies {
+  generateReport?: (request: AgentGroupReportRequest) => Promise<AgentGroupReportResult>
+  executeAction?: (request: WechatActionRequest) => Promise<WechatActionResult>
+  now?: () => number
+  /** 延迟实现。默认真 sleep；单测注入即时 resolve 的假实现，避免真的等 2 秒。 */
+  delay?: (ms: number) => Promise<void>
+  /** 日报落库（日报历史）。与旧定时日报**复用同一个**实现，不复制。 */
+  saveGeneratedReport?: (request: SaveGeneratedReportRequest) => Promise<SaveGeneratedReportResult>
+  /** 把日报来源群标识解析成联系人（拿头像 / 显示名）。 */
+  resolveReportContact?: (raw: string) => {
+    md5?: string
+    m_nsUsrName?: string
+    m_nsNickName?: string
+    avatar?: string
+  } | null
+  getContactAvatars?: (usernames: string[]) => Promise<Record<string, string>>
+}
+
+/** 策略层的错误码 → 用户可读短句。UI 直接展示这些文案，不做二次翻译。 */
+const ACTION_ERROR_MESSAGES: Record<string, string> = {
+  INVALID_REQUEST: '发送请求不合法',
+  INVALID_RECIPIENT: '找不到有效的发送对象',
+  ACTION_NOT_ALLOWED: '该自动化动作未被允许执行',
+  RECIPIENT_SCOPE_VIOLATION: '发送对象与触发来源不一致',
+  SEND_CAPABILITY_UNAVAILABLE: '当前环境没有可用的微信发送能力',
+  SEND_NOT_READY: '微信发送能力尚未就绪，请先绑定个人微信',
+  SEND_FAILED: '微信发送失败',
+  POLICY_BLOCKED: '该发送动作未通过策略检查',
+  UNKNOWN: '发送失败（未知原因）'
+}
+
+function createStep(key: AutomationStepKey): AutomationStep {
+  return { key, label: AUTOMATION_STEP_LABELS[key], status: 'pending' }
+}
+
+function markSuccess(step: AutomationStep, at: number): void {
+  step.status = 'success'
+  step.startedAt = step.startedAt ?? at
+  step.finishedAt = at
+  step.durationMs = Math.max(0, at - step.startedAt)
+}
+
+function markFailed(step: AutomationStep, at: number, error: string): void {
+  step.status = 'failed'
+  step.startedAt = step.startedAt ?? at
+  step.finishedAt = at
+  step.durationMs = Math.max(0, at - step.startedAt)
+  step.error = error
+}
+
+function markSkipped(step: AutomationStep, skipReason?: string): void {
+  step.status = 'skipped'
+  if (skipReason) step.skipReason = skipReason
+}
+
+function actionErrorMessage(code: string | undefined, fallback: string | undefined): string {
+  if (fallback && fallback.trim()) {
+    // 策略层给的 reason 已是中文短句；直接用，避免二次包装丢信息。
+    return fallback.trim()
+  }
+  return (code && ACTION_ERROR_MESSAGES[code]) || ACTION_ERROR_MESSAGES.UNKNOWN
+}
+
+export class AutomationActionRunner {
+  private readonly generateReport: (request: AgentGroupReportRequest) => Promise<AgentGroupReportResult>
+  private readonly executeAction: (request: WechatActionRequest) => Promise<WechatActionResult>
+  private readonly now: () => number
+  private readonly delay: (ms: number) => Promise<void>
+  private readonly saveGeneratedReport: (
+    request: SaveGeneratedReportRequest
+  ) => Promise<SaveGeneratedReportResult>
+  private readonly resolveReportContact: NonNullable<
+    AutomationActionRunnerDependencies['resolveReportContact']
+  >
+  private readonly getContactAvatars: (usernames: string[]) => Promise<Record<string, string>>
+
+  constructor(dependencies: AutomationActionRunnerDependencies = {}) {
+    this.generateReport = dependencies.generateReport ?? generateAgentGroupReport
+    this.executeAction = dependencies.executeAction ?? ((request) => wechatActionGateway.execute(request))
+    this.now = dependencies.now ?? (() => Date.now())
+    this.delay =
+      dependencies.delay ??
+      ((ms) => new Promise<void>((resolve) => setTimeout(resolve, ms)))
+    this.saveGeneratedReport = dependencies.saveGeneratedReport ?? saveGeneratedReport
+    this.resolveReportContact = dependencies.resolveReportContact ?? ((raw) => resolveMd5(raw))
+    this.getContactAvatars = dependencies.getContactAvatars ?? ((ids) => getContactAvatars(ids))
+  }
+
+  async run(input: AutomationRunInput): Promise<AutomationRunResult> {
+    const steps = STEP_ORDER.map((key) => createStep(key))
+    const stepAt = (key: AutomationStepKey): AutomationStep =>
+      steps.find((step) => step.key === key) as AutomationStep
+
+    markSuccess(stepAt('received'), this.now())
+    markSuccess(stepAt('matched'), this.now())
+
+    const replyAction = findAction(input.rule, 'replyText')
+    const reportAction = findAction(input.rule, 'generateReport')
+    const sendAction = findAction(input.rule, 'sendReportImage')
+
+    // ---- 步骤 3：回复确认 ----
+    //
+    // 回复等待：规则一命中就秒回，看起来就是个机器人（消息刚到、回复就到）。
+    // 等待时长是**规则自己的一项执行参数**（`rule.replyDelaySeconds`，在
+    // 「编辑自动化 → 3 · 触发后执行」里配），所以不同规则可以不一样。
+    //
+    // 等待刻意放在 `reply` 步骤计时**之外** —— `reply.durationMs` 只应该反映发送本身，
+    // 否则用户看到「回复确认 2000ms」会误以为是发送慢。
+    // 已经命中就不再回头重判规则：等待窗口里规则被停用/删掉也不中断本次执行，
+    // 与 cooldown、同消息幂等的口径一致（都是「命中那一刻」的快照）。
+    if (replyAction) {
+      // 用共享的归一化函数，而不是 `Number(...) || 0`：旧版 rules.json 里没有这个字段，
+      // 那应该按**默认 2 秒**处理（否则「默认 2 秒」要等用户手动进编辑页才会生效）。
+      const replyDelayMs = normalizeReplyDelaySeconds(input.rule.replyDelaySeconds) * 1_000
+      if (replyDelayMs > 0) await this.delay(replyDelayMs)
+    }
+
+    if (!replyAction) {
+      markSkipped(stepAt('reply'))
+    } else {
+      const replyStep = stepAt('reply')
+      replyStep.status = 'running'
+      replyStep.startedAt = this.now()
+      const sent = await this.sendThroughGateway(input, 'reply', {
+        type: 'text',
+        text: replyAction.text?.trim() || DEFAULT_REPLY_TEXT
+      })
+      if (!sent.ok) {
+        markFailed(replyStep, this.now(), sent.error || '回复确认失败')
+        markRemainingSkipped(steps, 'reply', SKIPPED_AFTER_REPLY_FAILURE)
+        return { steps, status: 'failed', errorSummary: replyStep.error }
+      }
+      markSuccess(replyStep, this.now())
+    }
+
+    // ---- 步骤 4：生成日报 ----
+    let pngPath: string | undefined
+    if (!reportAction) {
+      markSkipped(stepAt('report'))
+    } else {
+      const reportStep = stepAt('report')
+      reportStep.status = 'running'
+      reportStep.startedAt = this.now()
+      let result: AgentGroupReportResult
+      try {
+        result = await this.generateReport({ group: input.conversationId, range: 'today' })
+      } catch (error) {
+        result = {
+          success: false,
+          error: error instanceof Error ? error.message : String(error)
+        }
+      }
+      if (!result.success || !result.pngPath) {
+        markFailed(reportStep, this.now(), result.error || '日报生成失败')
+        markRemainingSkipped(steps, 'report', SKIPPED_AFTER_REPORT_FAILURE)
+        return { steps, status: 'failed', errorSummary: reportStep.error }
+      }
+      pngPath = result.pngPath
+      markSuccess(reportStep, this.now())
+    }
+
+    // ---- 步骤 5：发送日报图片 ----
+    //
+    // 三种情况必须分开判断 —— 合并成 `!sendAction || !pngPath` 会把
+    // 「规则要求发图、但图根本没生成」当成正常跳过，execution 还记成 success：
+    //
+    // A. 规则**本来就没有启用**这个动作 → skipped，这是正常的，不影响整体结果；
+    // B. 启用了，但要发的东西不存在 → **failed**，不能假装成功，
+    //    更不能退而求其次去发空路径 / 上一次的旧图 / 不存在的文件；
+    // C. 前置（生成日报）已经失败 → 上面就 return 了，走不到这里。
+    if (!sendAction) {
+      markSkipped(stepAt('send'))
+      return { steps, status: 'success', ...(pngPath ? { pngPath } : {}) }
+    }
+    const sendStep = stepAt('send')
+    if (!pngPath) {
+      markFailed(sendStep, this.now(), SEND_WITHOUT_IMAGE_ERROR)
+      return { steps, status: 'failed', errorSummary: sendStep.error }
+    }
+    sendStep.status = 'running'
+    sendStep.startedAt = this.now()
+    const sent = await this.sendThroughGateway(input, 'report', { type: 'image', path: pngPath })
+    if (!sent.ok) {
+      markFailed(sendStep, this.now(), sent.error || '发送日报图片失败')
+      return { steps, status: 'failed', errorSummary: sendStep.error, pngPath }
+    }
+    markSuccess(sendStep, this.now())
+    return { steps, status: 'success', pngPath }
+  }
+
+  /**
+   * 定时日报：把一次「到点触发 / 手动立即执行」跑成一次生成 + 一次发送。
+   *
+   * 与旧 `ScheduledReportService.executeTask` 的行为对齐（**不重写日报能力**）：
+   * 1. 先生成（含落库进日报历史）—— 发送能力不足**也照常生成**（旧语义如此）；
+   * 2. 再解析目标 —— 解析失败**不 fallback**，直接判失败；
+   * 3. 最后经 `WechatActionGateway` 发图片 —— 仍然是**发图片**，不退化成纯文本。
+   *
+   * 「生成成功但发送失败」是可表达的：`report_generated` 为 success、
+   * `report_sent` 为 failed，整体 `failed`，且 `reportGenerated === true`。
+   */
+  async runScheduledReport(input: ScheduledReportRunInput): Promise<ScheduledReportRunResult> {
+    const steps = SCHEDULED_REPORT_STEP_ORDER.map((key) => createStep(key))
+    const stepAt = (key: AutomationStepKey): AutomationStep =>
+      steps.find((step) => step.key === key) as AutomationStep
+
+    markSuccess(stepAt('schedule_triggered'), this.now())
+
+    // ---- 生成日报（含落进日报历史）----
+    const generatingStep = stepAt('report_generating')
+    generatingStep.status = 'running'
+    generatingStep.startedAt = this.now()
+    let generated: AgentGroupReportResult
+    try {
+      generated = await this.generateReport({
+        group: input.config.report.sourceConversationId,
+        range: input.config.report.range,
+        messageTypes: input.config.report.messageTypes,
+        templateId: input.config.report.templateId,
+        memberNameMode: input.config.report.memberNameMode,
+        timeoutSeconds: input.config.report.timeoutSeconds
+      })
+    } catch (error) {
+      generated = {
+        success: false,
+        error: error instanceof Error ? error.message : String(error)
+      }
+    }
+    if (!generated.success || !generated.pngPath) {
+      // 生成失败就**不发**：绝不生成空图片、也绝不退而求其次发上一张旧图。
+      const reason = generated.error || '日报生成失败'
+      markFailed(generatingStep, this.now(), reason)
+      markSkipped(stepAt('report_generated'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      markSkipped(stepAt('send_resolved'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      markSkipped(stepAt('report_sent'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      return {
+        steps,
+        status: 'failed',
+        errorSummary: reason,
+        reportGenerated: false,
+        ...(generated.errorCode ? { errorCode: generated.errorCode } : {})
+      }
+    }
+
+    let pngPath = generated.pngPath
+    try {
+      const reportContact = this.resolveReportContact(input.config.report.sourceConversationId)
+      let contactAvatar = reportContact?.avatar
+      if (!contactAvatar && reportContact?.m_nsUsrName) {
+        try {
+          const avatars = await this.getContactAvatars([reportContact.m_nsUsrName])
+          contactAvatar = avatars[reportContact.m_nsUsrName]
+        } catch (error) {
+          console.warn('[Automation] 日报群头像补全失败:', error)
+        }
+      }
+      const savedHistory = await this.saveGeneratedReport({
+        contactId: reportContact?.md5 || input.config.report.sourceConversationId,
+        contactName:
+          generated.groupName || reportContact?.m_nsNickName || input.sourceDisplayName || '群聊',
+        contactAvatar,
+        source: 'scheduled',
+        dateRange: generated.reportMetadata?.dateRange || scheduledReportRangeLabel(input.config.report.range),
+        reportDate: generated.reportMetadata?.reportDate,
+        messageCount: generated.messageCount ?? generated.reportMetadata?.messageCount ?? 0,
+        generatedAt: new Date(this.now()).toISOString(),
+        htmlPath: generated.htmlPath,
+        pngPath: generated.pngPath,
+        duration: generated.duration,
+        modelName: generated.modelName,
+        tokenUsage: generated.tokenUsage,
+        reportSnapshot: generated.reportSnapshot,
+        reportMetadata: generated.reportMetadata,
+        templateId: input.config.report.templateId
+      })
+      if (!savedHistory.success) {
+        throw new Error(savedHistory.error || '日报历史保存失败')
+      }
+      const recordPath = savedHistory.record?.pngPath || generated.pngPath
+      if (!recordPath) throw new Error('日报历史未返回可发送的 PNG 文件')
+      pngPath = recordPath
+    } catch (error) {
+      const reason = error instanceof Error ? error.message : String(error)
+      markFailed(generatingStep, this.now(), reason)
+      markSkipped(stepAt('report_generated'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      markSkipped(stepAt('send_resolved'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      markSkipped(stepAt('report_sent'), SKIPPED_AFTER_SCHEDULED_GENERATION_FAILURE)
+      return { steps, status: 'failed', errorSummary: reason, reportGenerated: false }
+    }
+    markSuccess(generatingStep, this.now())
+
+    const generatedStep = stepAt('report_generated')
+    markSuccess(generatedStep, this.now())
+    const messageCount = generated.messageCount ?? 0
+    if (messageCount > 0) generatedStep.detail = `共 ${messageCount} 条消息`
+
+    // ---- 目标解析失败：不许 fallback 到任何地方，直接判失败 ----
+    const targetStep = stepAt('send_resolved')
+    if (!input.resolution.ok) {
+      markFailed(targetStep, this.now(), input.resolution.error)
+      markSkipped(stepAt('report_sent'), '没有可用的发送目标，本次未发送。')
+      return {
+        steps,
+        status: 'failed',
+        errorSummary: input.resolution.error,
+        reportGenerated: true,
+        pngPath
+      }
+    }
+    targetStep.status = 'success'
+    targetStep.startedAt = this.now()
+    targetStep.finishedAt = targetStep.startedAt
+    targetStep.durationMs = 0
+    // 用户可读的目标名（例如「文件传输助手」「张三」「我」）—— 不带任何 id。
+    targetStep.detail = input.resolution.target.displayName
+
+    const sendStep = stepAt('report_sent')
+
+    // ---- 发送能力缺失：日报已生成，如实判失败（不回滚、不隐藏）----
+    if (input.sendBlockedReason) {
+      markFailed(sendStep, this.now(), input.sendBlockedReason)
+      return {
+        steps,
+        status: 'failed',
+        errorSummary: input.sendBlockedReason,
+        reportGenerated: true,
+        pngPath
+      }
+    }
+
+    sendStep.status = 'running'
+    sendStep.startedAt = this.now()
+    const sent = await this.sendScheduledReportImage(input, pngPath)
+    if (!sent.ok) {
+      markFailed(sendStep, this.now(), sent.error || '发送日报失败')
+      return {
+        steps,
+        status: 'failed',
+        errorSummary: sendStep.error,
+        reportGenerated: true,
+        pngPath
+      }
+    }
+    markSuccess(sendStep, this.now())
+    return { steps, status: 'success', reportGenerated: true, pngPath }
+  }
+
+  /**
+   * 定时日报的统一发送出口。
+   *
+   * **必须走 `WechatActionGateway`**：幂等 + 3 秒发送间隔 + 审计落盘都在那里。
+   * Automation 层不允许知道 OneBot / WCHook / native host / Windows hook 的存在。
+   */
+  private async sendScheduledReportImage(
+    input: ScheduledReportRunInput,
+    pngPath: string
+  ): Promise<{ ok: boolean; error?: string }> {
+    if (!input.resolution.ok) return { ok: false, error: input.resolution.error }
+    // 定时触发走 automation（纳入发送节流）；用户手动执行走 user（不纳入节流）。
+    const triggerType = input.trigger === 'manual' ? 'user' : 'automation'
+    try {
+      const result = await this.executeAction({
+        idempotencyKey: `${AUTOMATION_SEND_PURPOSE.scheduledReport}:${input.executionId}`,
+        origin: AUTOMATION_SEND_ORIGIN,
+        purpose: AUTOMATION_SEND_PURPOSE.scheduledReport,
+        triggerType,
+        executionId: input.executionId,
+        recipient: input.resolution.target.recipient,
+        content: { type: 'image', path: pngPath }
+      })
+      if (result.status !== 'sent') {
+        return { ok: false, error: actionErrorMessage(result.errorCode, result.reason) }
+      }
+
+      /*
+       * 后置词：**只在图片明确 sent 之后**才发，图片失败时严格短路 ——
+       * 与 `WechatActionGateway.executeReportImageSequence`（手动发送）同一口径。
+       * 空字符串 = 用户只要图片，什么都不补发。
+       *
+       * 独立 purpose + 独立幂等位：共用一个 key 会让「图片发成功、后置词被
+       * 幂等短路」变成常态。
+       */
+      const postfixText = String(input.config.postfixText || '').trim()
+      if (!postfixText) return { ok: true }
+
+      const postfix = await this.executeAction({
+        idempotencyKey: `${AUTOMATION_SEND_PURPOSE.scheduledReportPostfix}:${input.executionId}`,
+        origin: AUTOMATION_SEND_ORIGIN,
+        purpose: AUTOMATION_SEND_PURPOSE.scheduledReportPostfix,
+        triggerType,
+        executionId: input.executionId,
+        recipient: input.resolution.target.recipient,
+        content: { type: 'text', text: postfixText }
+      })
+      if (postfix.status === 'sent') return { ok: true }
+      // 绝不吞掉：图片确实发出去了，但这一次执行**没有完成**，必须如实报出来。
+      return {
+        ok: false,
+        error: `日报图片已发送，但后置词发送失败：${actionErrorMessage(
+          postfix.errorCode,
+          postfix.reason
+        )}`
+      }
+    } catch (error) {
+      return { ok: false, error: error instanceof Error ? error.message : String(error) }
+    }
+  }
+
+  /**
+   * 退群通知：把「检测到成员退出」这件事跑成一次发送。
+   *
+   * 与 `run()` 的差别：没有动作链、没有回复等待、没有 cooldown
+   * （退群是低频事件，且**不能被时间窗合并** —— 两个成员先后退出就是两条通知）。
+   */
+  async runLeaveNotification(input: LeaveNotificationRunInput): Promise<LeaveNotificationRunResult> {
+    const steps = LEAVE_NOTIFICATION_STEP_ORDER.map((key) => createStep(key))
+    const stepAt = (key: AutomationStepKey): AutomationStep =>
+      steps.find((step) => step.key === key) as AutomationStep
+
+    markSuccess(stepAt('exit_received'), this.now())
+    markSuccess(stepAt('exit_matched'), this.now())
+
+    // ---- 目标解析失败：不许 fallback 到任何地方，直接判失败 ----
+    if (!input.resolution.ok) {
+      markFailed(stepAt('exit_target'), this.now(), input.resolution.error)
+      markSkipped(stepAt('exit_send'), '没有可用的通知目标，未发送。')
+      return { steps, status: 'failed', errorSummary: input.resolution.error }
+    }
+    const targetStep = stepAt('exit_target')
+    targetStep.status = 'success'
+    targetStep.startedAt = this.now()
+    targetStep.finishedAt = targetStep.startedAt
+    targetStep.durationMs = 0
+    // 用户可读的目标名（例如「文件传输助手」「张三」「我」）—— 不带任何 id。
+    targetStep.detail = input.resolution.target.displayName
+
+    const sendStep = stepAt('exit_send')
+
+    // ---- 发送能力缺失：如实判失败（退群事实本身仍然是成功的） ----
+    if (input.sendBlockedReason) {
+      markFailed(sendStep, this.now(), input.sendBlockedReason)
+      return { steps, status: 'failed', errorSummary: input.sendBlockedReason }
+    }
+
+    const text = renderLeaveNotificationText(input.event, input.config.template).trim()
+    if (!text) {
+      const error = '通知内容为空，未发送。'
+      markFailed(sendStep, this.now(), error)
+      return { steps, status: 'failed', errorSummary: error }
+    }
+
+    sendStep.status = 'running'
+    sendStep.startedAt = this.now()
+    const sent = await this.sendLeaveNotification(input, text)
+    if (!sent.ok) {
+      markFailed(sendStep, this.now(), sent.error || '发送退群通知失败')
+      return { steps, status: 'failed', errorSummary: sendStep.error }
+    }
+    markSuccess(sendStep, this.now())
+    return { steps, status: 'success' }
+  }
+
+  /**
+   * 退群通知的统一发送出口。
+   *
+   * **必须走 `WechatActionGateway`**：幂等 + 3 秒发送间隔 + 审计落盘都在那里，
+   * 直接调个人微信发送会绕过全部三样。Automation 层也不允许知道
+   * OneBot / WCHook / native host / Windows hook 的存在。
+   *
+   * 幂等键由 **eventId** 派生（不是 executionId）：审计是落盘的，
+   * 于是"重启后重复投递同一退群事件"会被持久层直接短路。
+   */
+  private async sendLeaveNotification(
+    input: LeaveNotificationRunInput,
+    text: string
+  ): Promise<{ ok: boolean; error?: string }> {
+    if (!input.resolution.ok) return { ok: false, error: input.resolution.error }
+    try {
+      const result = await this.executeAction({
+        idempotencyKey: `${AUTOMATION_SEND_PURPOSE.leaveNotification}:${input.event.eventId}`,
+        origin: AUTOMATION_SEND_ORIGIN,
+        purpose: AUTOMATION_SEND_PURPOSE.leaveNotification,
+        triggerType: 'automation',
+        executionId: input.executionId,
+        sourceId: input.event.eventId,
+        recipient: input.resolution.target.recipient,
+        content: { type: 'text', text }
+      })
+      if (result.status === 'sent') return { ok: true }
+      return { ok: false, error: actionErrorMessage(result.errorCode, result.reason) }
+    } catch (error) {
+      return { ok: false, error: error instanceof Error ? error.message : String(error) }
+    }
+  }
+
+  /**
+   * 统一发送出口。
+   *
+   * **刻意不直接调 `WechatSendGateway`，而是走 `WechatActionGateway`**：后者在
+   * `WechatSendGateway` 之上多给了三样本功能必须的东西 ——
+   * 幂等（同一 executionId 不会重复发）、自动化发送节流（3s 间隔，防刷屏）、
+   * 审计落盘。底层实际发送仍然经由 `WechatSendGateway.sendPersonal`，
+   * 所以「所有发送统一走 WechatSendGateway」这条约束依然成立。
+   */
+  private async sendThroughGateway(
+    input: AutomationRunInput,
+    kind: 'reply' | 'report',
+    content: WechatActionContent
+  ): Promise<{ ok: boolean; error?: string }> {
+    try {
+      const result = await this.executeAction({
+        idempotencyKey: automationIdempotencyKey(kind, input.executionId),
+        origin: AUTOMATION_SEND_ORIGIN,
+        purpose: AUTOMATION_SEND_PURPOSE[kind],
+        triggerType: 'automation',
+        executionId: input.executionId,
+        recipient: {
+          type: input.isGroup ? 'group' : 'contact',
+          id: input.conversationId,
+          name: input.sourceDisplayName
+        },
+        content
+      })
+      if (result.status === 'sent') return { ok: true }
+      return { ok: false, error: actionErrorMessage(result.errorCode, result.reason) }
+    } catch (error) {
+      // execute() 本身刻意不抛，这里兜的是注入实现或意外异常。
+      return { ok: false, error: error instanceof Error ? error.message : String(error) }
+    }
+  }
+}
+
+function findAction(rule: AutomationRule, type: AutomationAction['type']): AutomationAction | undefined {
+  return rule.actions.find((action) => action.type === type && action.enabled)
+}
+
+/** 把 `after` 之后的步骤全部标成 `skipped`（前一步挂了，后面的不许再动）。 */
+function markRemainingSkipped(
+  steps: AutomationStep[],
+  after: AutomationStepKey,
+  skipReason?: string
+): void {
+  const from = STEP_ORDER.indexOf(after) + 1
+  for (const key of STEP_ORDER.slice(from)) {
+    const step = steps.find((item) => item.key === key)
+    if (step) markSkipped(step, skipReason)
+  }
+}
+
+export const automationActionRunner = new AutomationActionRunner()
@@ -0,0 +1,153 @@
+import path from 'node:path'
+import { app } from 'electron'
+import fs from 'fs-extra'
+import type {
+  AutomationExecution,
+  AutomationExecutionTrigger
+} from '../../shared/automation'
+
+/**
+ * AutomationExecutionLogService —— **用户层**执行日志。
+ *
+ * 与 debug 日志的区别：这份记录是给用户看「哪一步坏了」的，所以
+ * - 只保留用户可读字段（规则名、来源显示名、步骤状态、耗时、错误）；
+ * - **不含** wxid / localId / serverId / source XML / raw payload / 图片绝对路径；
+ * - 必须落盘，重启后还能看到近期记录。
+ *
+ * 容量有界（`MAX_RECORDS`），否则长期运行会把文件撑到几十兆。
+ */
+
+const STORAGE_DIR = 'automation'
+const EXECUTIONS_FILE = 'executions.json'
+const MAX_RECORDS = 200
+
+export interface AutomationExecutionLogDependencies {
+  userDataPath?: () => string
+}
+
+const EXECUTION_STATUSES: AutomationExecution['status'][] = ['running', 'success', 'failed']
+
+const EXECUTION_TRIGGERS: AutomationExecutionTrigger[] = ['message', 'exit', 'schedule', 'manual']
+
+function normalizeExecution(value: unknown): AutomationExecution | null {
+  if (!value || typeof value !== 'object') return null
+  const record = value as Partial<AutomationExecution>
+  const executionId = String(record.executionId || '').trim()
+  if (!executionId) return null
+  return {
+    executionId,
+    ruleId: String(record.ruleId || ''),
+    ruleName: String(record.ruleName || ''),
+    triggerTime: Number(record.triggerTime) || 0,
+    /*
+     * `trigger` 必须在这里**显式透传**。
+     *
+     * 这个归一化是白名单式的：没列出来的字段会被静默丢掉。定时日报的
+     * 「本次是定时跑的还是用户点的立即执行」全靠这个字段区分，
+     * 漏掉它就会变成"写的时候有、读出来永远没有"—— 而且没有任何报错。
+     * 旧记录没有这个字段，保持 `undefined`（读盘不猜测）。
+     */
+    ...(EXECUTION_TRIGGERS.includes(record.trigger as AutomationExecutionTrigger)
+      ? { trigger: record.trigger as AutomationExecutionTrigger }
+      : {}),
+    sourceDisplayName: String(record.sourceDisplayName || ''),
+    // 不认识的 status（含历史遗留值）一律降级成 `running`，绝不凭空造出成功/失败。
+    status: EXECUTION_STATUSES.includes(record.status as AutomationExecution['status'])
+      ? (record.status as AutomationExecution['status'])
+      : 'running',
+    durationMs: Number(record.durationMs) || 0,
+    steps: Array.isArray(record.steps) ? record.steps : [],
+    ...(record.errorSummary ? { errorSummary: String(record.errorSummary) } : {})
+  }
+}
+
+export class AutomationExecutionLogService {
+  private readonly userDataPath: () => string
+  private records: AutomationExecution[] = []
+  private loaded = false
+
+  constructor(dependencies: AutomationExecutionLogDependencies = {}) {
+    this.userDataPath = dependencies.userDataPath ?? (() => app.getPath('userData'))
+  }
+
+  list(query: { limit?: number } = {}): AutomationExecution[] {
+    this.ensureLoaded()
+    const requested = Number(query?.limit)
+    const limit = Number.isFinite(requested) && requested > 0 ? Math.floor(requested) : MAX_RECORDS
+    return this.records.slice(0, Math.min(limit, MAX_RECORDS)).map((record) => structuredClone(record))
+  }
+
+  /**
+   * 写入（或按 `executionId` 覆盖）一条执行记录。
+   *
+   * 覆盖语义是必需的：执行是「先建 running、跑完再回落终态」，
+   * 中间态与终态共用同一个 `executionId`。
+   */
+  record(execution: AutomationExecution): void {
+    this.ensureLoaded()
+    const normalized = normalizeExecution(execution)
+    if (!normalized) return
+    const withoutSame = this.records.filter((item) => item.executionId !== normalized.executionId)
+    this.records = [normalized, ...withoutSame].slice(0, MAX_RECORDS)
+    this.persist()
+  }
+
+  clear(): boolean {
+    this.ensureLoaded()
+    this.records = []
+    this.persist()
+    return true
+  }
+
+  /**
+   * 统计 `sinceMs` 之后（含）的记录数与成功数。用于顶部「今日执行」。
+   *
+   * 每条记录都对应一次**真正跑过**的执行（gate 拦下的消息不会产生记录），
+   * 所以这里直接计数即可。
+   */
+  countSince(sinceMs: number): { total: number; success: number } {
+    this.ensureLoaded()
+    const from = Number(sinceMs) || 0
+    let total = 0
+    let success = 0
+    for (const record of this.records) {
+      if (record.triggerTime < from) continue
+      total += 1
+      if (record.status === 'success') success += 1
+    }
+    return { total, success }
+  }
+
+  private ensureLoaded(): void {
+    if (this.loaded) return
+    this.loaded = true
+    try {
+      const raw = fs.readJsonSync(this.executionsFilePath()) as unknown
+      const values = Array.isArray(raw) ? raw : []
+      this.records = values
+        .map((value) => normalizeExecution(value))
+        .filter((value): value is AutomationExecution => value !== null)
+        .slice(0, MAX_RECORDS)
+    } catch {
+      this.records = []
+    }
+  }
+
+  private executionsFilePath(): string {
+    return path.join(this.userDataPath(), STORAGE_DIR, EXECUTIONS_FILE)
+  }
+
+  private persist(): void {
+    try {
+      const filePath = this.executionsFilePath()
+      fs.ensureDirSync(path.dirname(filePath))
+      fs.writeJsonSync(filePath, this.records, { spaces: 2 })
+    } catch (error) {
+      console.warn(
+        `[Automation] 保存执行日志失败: ${error instanceof Error ? error.message : String(error)}`
+      )
+    }
+  }
+}
+
+export const automationExecutionLogService = new AutomationExecutionLogService()
@@ -0,0 +1,676 @@
+import { randomUUID } from 'node:crypto'
+import path from 'node:path'
+import { app } from 'electron'
+import fs from 'fs-extra'
+import {
+  BUILTIN_LEAVE_NOTIFICATION_RULE_ID,
+  createDefaultDailyReportRule,
+  createDefaultLeaveNotificationRule,
+  createDefaultScheduledReportRule,
+  normalizeRuleDraft,
+  type AutomationRule,
+  type AutomationRuleType,
+  type ScheduledReportAutomationConfig
+} from '../../shared/automation'
+import {
+  planLeaveNotificationMigration,
+  type LeaveNotificationMigrationPlan,
+  type LegacyLeaveNotificationState
+} from './leave-notification-migration'
+import {
+  planScheduledReportMigrationBatch,
+  type LegacyScheduledReportTask,
+  type ScheduledReportMigrationSummary
+} from './scheduled-report-migration'
+import {
+  GROUP_EXIT_NOTIFICATION_TEMPLATE_MAX_LENGTH,
+  insertGroupNamePlaceholder
+} from '../../shared/group-exit-monitor'
+
+/**
+ * AutomationRuleStore —— 自动化规则的持久化与增删改查。
+ *
+ * 存储形态刻意做到最简：一个 JSON 文件（`{userData}/automation/rules.json`）。
+ * 规则总量是个位数到几十条，引入数据库只会增加迁移负担。
+ *
+ * **一次性标记全部单独存**（`builtinSeeded` / `leaveNotificationMigrated` /
+ * `notificationTemplateUpgraded` / `scheduledReportMigrated`）：
+ * 如果靠"文件里有没有那条内置规则"来判断是否播种，用户一旦删掉它，
+ * 下次启动就会被重新塞回来 —— 用户会认为删除功能坏了。迁移同理。
+ */
+
+const STORAGE_DIR = 'automation'
+const RULES_FILE = 'rules.json'
+const MIGRATION_BACKUP_FILE = 'leave-notification-migration-backup.json'
+/** 旧退群监控的状态文件（迁移时读一次，之后运行期不再读）。 */
+const LEGACY_MONITOR_FILE = 'group-exit-monitor.json'
+/** 旧定时日报的**任务**文件（迁移时读一次，之后运行期不再读）。 */
+const LEGACY_SCHEDULED_DIR = 'scheduled-reports'
+const LEGACY_SCHEDULED_TASKS_FILE = 'tasks.json'
+const SCHEDULED_MIGRATION_BACKUP_FILE = 'scheduled-report-migration-backup.json'
+const CURRENT_VERSION = 4
+
+export class AutomationRulePersistenceError extends Error {
+  constructor() {
+    super('自动化规则保存失败')
+    this.name = 'AutomationRulePersistenceError'
+  }
+}
+
+interface StoredRules {
+  version: number
+  /** 内置「@我生成日报」是否已经播种过。**即使随后被删除也保持 true**。 */
+  builtinSeeded: boolean
+  /** 旧退群通知配置是否已经迁移过。**即使随后被删除也保持 true**。 */
+  leaveNotificationMigrated: boolean
+  /**
+   * 是否已经替用户往模板里补过 `{groupName}`。
+   *
+   * **必须有这个标记**：没有它就得靠"模板里有没有 `{groupName}`"来判断，
+   * 于是用户主动删掉那一行后、下次启动又会被补回来 —— 删不掉的东西最烦人。
+   */
+  notificationTemplateUpgraded: boolean
+  /**
+   * 旧「定时日报」任务是否已经迁移过。
+   *
+   * ⚠️ 与其它标记有一点不同：迁移需要**把旧群标识解析成稳定会话 id**，
+   * 而那需要数据库。数据库没就绪时这一位**保持 false**，等解析器注入后再补跑 ——
+   * 否则会把每一条规则都误判成"目标无法确认"。
+   */
+  scheduledReportMigrated: boolean
+  rules: AutomationRule[]
+}
+
+export interface AutomationRuleStoreDependencies {
+  userDataPath?: () => string
+  now?: () => number
+  /**
+   * 迁移旧定时日报时，把 legacy 群标识（会话 md5 / 群名 / roomId）解析成
+   * **稳定会话 id**（`xxx@chatroom`）。解析不到返回 `undefined`。
+   *
+   * 不注入 ⇒ 迁移**推迟**（不写任何规则、不置标记）。
+   */
+  resolveLegacyConversationId?: (raw: string) => string | undefined
+}
+
+function emptyState(): StoredRules {
+  return {
+    version: CURRENT_VERSION,
+    builtinSeeded: false,
+    leaveNotificationMigrated: false,
+    notificationTemplateUpgraded: false,
+    scheduledReportMigrated: false,
+    rules: []
+  }
+}
+
+function normalizeRuleType(value: unknown): AutomationRuleType {
+  return value === 'leave_notification'
+    ? 'leave_notification'
+    : value === 'scheduled_report'
+      ? 'scheduled_report'
+      : 'daily_report'
+}
+
+/**
+ * 单条规则的读盘归一化。
+ *
+ * 走 `normalizeRuleDraft` —— 它覆盖了 `AutomationRule` 除 id / 时间戳之外的**全部**字段，
+ * 所以这是无损的，同时自动补上历史 rules.json 缺失的 `ruleType` / 各类 config。
+ */
+function normalizeStoredRule(value: unknown): AutomationRule | null {
+  if (!value || typeof value !== 'object') return null
+  const raw = value as Partial<AutomationRule>
+  const id = String(raw.id || '').trim()
+  if (!id) return null
+  const ruleType = normalizeRuleType(raw.ruleType)
+  const normalized = normalizeRuleDraft(
+    {
+      ...raw,
+      ruleType,
+      // 历史规则没有这些字段；反过来也要避免脏字段落盘。
+      ...(ruleType === 'leave_notification' ? { leaveNotification: raw.leaveNotification } : {}),
+      ...(ruleType === 'scheduled_report' ? { scheduledReport: raw.scheduledReport } : {})
+    },
+    String(raw.name || '').trim() || '未命名自动化'
+  )
+  const createdAt = Number(raw.createdAt) || 0
+  const updatedAt = Number(raw.updatedAt) || createdAt
+  return {
+    ...normalized,
+    id,
+    createdAt,
+    updatedAt,
+    ...(normalized.ruleType === 'leave_notification' && normalized.leaveNotification
+      ? { leaveNotification: normalized.leaveNotification }
+      : {}),
+    ...(normalized.ruleType === 'scheduled_report' && normalized.scheduledReport
+      ? { scheduledReport: normalized.scheduledReport }
+      : {})
+  }
+}
+
+/**
+ * 一次性替用户把 `{groupName}` 补进退群通知模板。
+ *
+ * **为什么替他填**：模板里原先根本没有群名变量（`{groupRemark}` 是成员在本群的昵称）。
+ * 目标一旦不是「当前群聊」，通知就等于「张三退群了」—— 收件人不知道是哪个群。
+ *
+ * **为什么只跑一次**：靠"模板里有没有 `{groupName}`"判断会导致用户删掉后被反复补回来。
+ * 所以由 `notificationTemplateUpgraded` 标记保证一次性；用户之后删掉不会回来。
+ */
+function upgradeLeaveNotificationTemplate(rules: AutomationRule[]): {
+  rules: AutomationRule[]
+  count: number
+} {
+  let count = 0
+  const next = rules.map((rule) => {
+    if (rule.ruleType !== 'leave_notification' || !rule.leaveNotification) return rule
+    const current = rule.leaveNotification.template
+    const upgraded = insertGroupNamePlaceholder(current)
+    // 补完不能超长：超了下次读盘会被校验拦下、整条模板被重置成默认，反而更糟。
+    if (upgraded === current || upgraded.length > GROUP_EXIT_NOTIFICATION_TEMPLATE_MAX_LENGTH) {
+      return rule
+    }
+    count += 1
+    return { ...rule, leaveNotification: { ...rule.leaveNotification, template: upgraded } }
+  })
+  return { rules: next, count }
+}
+
+/**
+ * 调度游标（`lastRunAt` / `lastScheduledSlot`）**由 scheduler 拥有**，草稿改不动它们。
+ *
+ * 保存规则时一律从 `current` 取，而不是从草稿取 —— 否则用户进一次编辑页保存，
+ * 就会把"已消费的槽位"抹掉，导致当天重复补跑一次日报。
+ */
+function pickScheduledReportRuntime(
+  current: ScheduledReportAutomationConfig | undefined
+): Partial<ScheduledReportAutomationConfig> {
+  if (!current) return {}
+  return {
+    ...(current.lastRunAt ? { lastRunAt: current.lastRunAt } : {}),
+    ...(current.lastScheduledSlot ? { lastScheduledSlot: current.lastScheduledSlot } : {})
+  }
+}
+
+/** 读盘容错：任何字段可疑都降级成安全值，绝不因为一个坏文件让功能整体不可用。 */
+function normalizeStored(value: unknown): StoredRules {
+  if (!value || typeof value !== 'object') return emptyState()
+  const input = value as Partial<StoredRules>
+  const rules = Array.isArray(input.rules)
+    ? input.rules
+        .map((rule) => normalizeStoredRule(rule))
+        .filter((rule): rule is AutomationRule => rule !== null)
+    : []
+  return {
+    version: Number(input.version) || 1,
+    builtinSeeded: input.builtinSeeded === true,
+    leaveNotificationMigrated: input.leaveNotificationMigrated === true,
+    notificationTemplateUpgraded: input.notificationTemplateUpgraded === true,
+    scheduledReportMigrated: input.scheduledReportMigrated === true,
+    rules
+  }
+}
+
+export class AutomationRuleStore {
+  private readonly userDataPath: () => string
+  private readonly now: () => number
+  private state: StoredRules = emptyState()
+  private loaded = false
+  /** 迁移结果，供启动日志/报告读取（不含任何 id）。 */
+  private lastMigrationPlan: LeaveNotificationMigrationPlan | null = null
+  private lastScheduledMigrationSummary: ScheduledReportMigrationSummary | null = null
+  private legacyConversationResolver: ((raw: string) => string | undefined) | null
+
+  constructor(dependencies: AutomationRuleStoreDependencies = {}) {
+    this.userDataPath = dependencies.userDataPath ?? (() => app.getPath('userData'))
+    this.now = dependencies.now ?? (() => Date.now())
+    this.legacyConversationResolver = dependencies.resolveLegacyConversationId ?? null
+  }
+
+  /**
+   * 注入 / 替换旧定时日报迁移所需的会话解析器（数据库就绪后由 main 调用）。
+   *
+   * 若之前因为"解析器不可用"推迟了迁移，这里会**立刻补跑**并落盘。
+   */
+  setLegacyConversationResolver(resolver: (raw: string) => string | undefined): void {
+    this.legacyConversationResolver = resolver
+    if (!this.loaded) return
+    if (this.state.scheduledReportMigrated) return
+    const next = structuredClone(this.state)
+    if (this.runScheduledReportMigration(next)) this.commitState(next)
+  }
+
+  /** 旧定时日报迁移是否仍在等待解析器（仅用于启动日志与测试断言）。 */
+  hasPendingScheduledReportMigration(): boolean {
+    this.ensureLoaded()
+    return !this.state.scheduledReportMigrated && this.readLegacyScheduledTasks().length > 0
+  }
+
+  listRules(): AutomationRule[] {
+    this.ensureLoaded()
+    return this.state.rules.map((rule) => structuredClone(rule))
+  }
+
+  getRule(id: string): AutomationRule | undefined {
+    const key = String(id || '').trim()
+    if (!key) return undefined
+    const found = this.listRules().find((rule) => rule.id === key)
+    return found
+  }
+
+  /** 上一次迁移的判定结果（未迁移时为 null）。 */
+  getLastLeaveNotificationMigration(): LeaveNotificationMigrationPlan | null {
+    this.ensureLoaded()
+    return this.lastMigrationPlan ? { ...this.lastMigrationPlan } : null
+  }
+
+  /** 上一次旧定时日报迁移的统计（未迁移时为 null）。 */
+  getLastScheduledReportMigration(): ScheduledReportMigrationSummary | null {
+    this.ensureLoaded()
+    return this.lastScheduledMigrationSummary ? { ...this.lastScheduledMigrationSummary } : null
+  }
+
+  createRule(draft: unknown): AutomationRule {
+    this.ensureLoaded()
+    const timestamp = this.now()
+    const normalized = normalizeRuleDraft(draft)
+    const rule: AutomationRule = {
+      ...normalized,
+      id: randomUUID(),
+      createdAt: timestamp,
+      updatedAt: timestamp
+    }
+    this.commitRules([...this.state.rules, rule])
+    return structuredClone(rule)
+  }
+
+  updateRule(id: string, draft: unknown): AutomationRule | undefined {
+    this.ensureLoaded()
+    const key = String(id || '').trim()
+    const index = this.state.rules.findIndex((rule) => rule.id === key)
+    if (index < 0) return undefined
+    const current = this.state.rules[index]
+    // 名字留空时沿用原名，而不是变成「未命名自动化」—— 编辑页只改开关时不该改名。
+    const normalized = normalizeRuleDraft(draft, current.name)
+    const rule: AutomationRule = {
+      ...current,
+      ...normalized,
+      id: current.id,
+      // ruleType 不允许被草稿改掉：它是规则的**身份**，不是可编辑字段。
+      ruleType: current.ruleType,
+      ...(current.ruleType === 'leave_notification' && normalized.leaveNotification
+        ? { leaveNotification: normalized.leaveNotification }
+        : {}),
+      ...(current.ruleType === 'scheduled_report' && normalized.scheduledReport
+        ? {
+            scheduledReport: {
+              ...normalized.scheduledReport,
+              ...pickScheduledReportRuntime(current.scheduledReport)
+            }
+          }
+        : {}),
+      createdAt: current.createdAt,
+      updatedAt: this.now()
+    }
+    // 切类型时不能留下另一种类型才认识的字段。
+    if (rule.ruleType === 'daily_report') {
+      delete rule.leaveNotification
+      delete rule.scheduledReport
+    } else if (rule.ruleType === 'leave_notification') {
+      delete rule.scheduledReport
+    } else {
+      delete rule.leaveNotification
+    }
+    this.commitRules(this.state.rules.map((item, at) => (at === index ? rule : item)))
+    return structuredClone(rule)
+  }
+
+  /**
+   * 保存「退群通知」规则（**singleton upsert**）。
+   *
+   * 不存在则创建（固定 id），存在则更新 —— 所以点多少次保存都只有一条规则。
+   */
+  saveLeaveNotificationRule(draft: unknown): AutomationRule {
+    this.ensureLoaded()
+    const existing = this.state.rules.find(
+      (rule) => rule.id === BUILTIN_LEAVE_NOTIFICATION_RULE_ID
+    )
+    if (existing) {
+      const updated = this.updateRule(existing.id, { ...(draft as object), ruleType: 'leave_notification' })
+      if (updated) return updated
+    }
+    const timestamp = this.now()
+    const normalized = normalizeRuleDraft({ ...(draft as object), ruleType: 'leave_notification' })
+    const rule: AutomationRule = {
+      ...createDefaultLeaveNotificationRule(timestamp),
+      ...normalized,
+      id: BUILTIN_LEAVE_NOTIFICATION_RULE_ID,
+      ruleType: 'leave_notification',
+      name: normalized.name || '退群通知',
+      createdAt: timestamp,
+      updatedAt: timestamp
+    }
+    this.commitRules([...this.state.rules, rule])
+    return structuredClone(rule)
+  }
+
+  deleteRule(id: string): boolean {
+    this.ensureLoaded()
+    const key = String(id || '').trim()
+    const before = this.state.rules.length
+    const rules = this.state.rules.filter((rule) => rule.id !== key)
+    if (rules.length === before) return false
+    this.commitRules(rules)
+    return true
+  }
+
+  setRuleEnabled(id: string, enabled: boolean): AutomationRule | undefined {
+    const rule = this.getRule(id)
+    if (!rule) return undefined
+    return this.updateRule(id, { ...rule, enabled: enabled === true })
+  }
+
+  /**
+   * 写**调度游标**（`lastRunAt` / `lastScheduledSlot`）。
+   *
+   * 刻意不走 `updateRule`：这不是用户配置编辑，走草稿归一化会顺带触碰别的字段，
+   * 也可能被"草稿里没这个字段"给抹掉。这里只允许改这两个键。
+   */
+  setScheduledReportRuntime(
+    id: string,
+    patch: { lastRunAt?: string; lastScheduledSlot?: string }
+  ): AutomationRule | undefined {
+    this.ensureLoaded()
+    const key = String(id || '').trim()
+    const index = this.state.rules.findIndex((rule) => rule.id === key)
+    if (index < 0) return undefined
+    const current = this.state.rules[index]
+    if (current.ruleType !== 'scheduled_report' || !current.scheduledReport) return undefined
+    const next: AutomationRule = {
+      ...current,
+      scheduledReport: {
+        ...current.scheduledReport,
+        ...(patch.lastRunAt !== undefined ? { lastRunAt: patch.lastRunAt } : {}),
+        ...(patch.lastScheduledSlot !== undefined
+          ? { lastScheduledSlot: patch.lastScheduledSlot }
+          : {})
+      },
+      updatedAt: this.now()
+    }
+    this.commitRules(this.state.rules.map((item, at) => (at === index ? next : item)))
+    return structuredClone(next)
+  }
+
+  private ensureLoaded(): void {
+    if (this.loaded) return
+    this.loaded = true
+    let stored: StoredRules
+    try {
+      stored = normalizeStored(fs.readJsonSync(this.rulesFilePath()) as unknown)
+    } catch {
+      stored = emptyState()
+    }
+    if (!stored.builtinSeeded) {
+      stored.builtinSeeded = true
+      stored.rules = [...stored.rules, createDefaultDailyReportRule(this.now())]
+    }
+    if (!stored.leaveNotificationMigrated) {
+      stored.leaveNotificationMigrated = true
+      stored.rules = this.migrateLeaveNotification(stored.rules)
+    }
+    if (!stored.notificationTemplateUpgraded) {
+      stored.notificationTemplateUpgraded = true
+      const upgraded = upgradeLeaveNotificationTemplate(stored.rules)
+      stored.rules = upgraded.rules
+      // 只记条数，**不记模板正文**。
+      if (upgraded.count > 0) {
+        console.log(`[Automation] 退群通知模板已补上群名变量（${upgraded.count} 条）`)
+      }
+    }
+    this.runScheduledReportMigration(stored)
+    stored.version = CURRENT_VERSION
+    this.state = stored
+    try {
+      this.persist()
+    } catch {
+      // Keep the in-memory defaults available; later writes report persistence failures.
+    }
+  }
+
+  /** 旧退群通知配置 → 内置退群通知规则。**只跑一次**，且先备份旧配置。 */
+  private migrateLeaveNotification(rules: AutomationRule[]): AutomationRule[] {
+    if (rules.some((rule) => rule.id === BUILTIN_LEAVE_NOTIFICATION_RULE_ID)) return rules
+    const timestamp = this.now()
+    const legacy = this.readLegacyLeaveNotificationState()
+    if (!legacy) {
+      // 全新安装：没有历史可迁，直接建默认规则（默认目标：文件传输助手）。
+      this.logMigration('fresh_install', null)
+      return [...rules, createDefaultLeaveNotificationRule(timestamp)]
+    }
+    const plan = planLeaveNotificationMigration(legacy)
+    this.lastMigrationPlan = plan
+    this.writeMigrationBackup(legacy, plan)
+    this.logMigration(plan.outcome, plan)
+    return [
+      ...rules,
+      createDefaultLeaveNotificationRule(timestamp, {
+        template: plan.template,
+        target: plan.target,
+        enabled: plan.enabled,
+        // 二次勾选（旧「通知群聊」）逐字保留，迁移不再是有损的。
+        notifyScope: plan.notifyScope,
+        notifyRoomIds: plan.notifyRoomIds
+      })
+    ]
+  }
+
+  /**
+   * 旧「定时日报」任务 → `scheduled_report` 规则。**只跑一次**。
+   *
+   * 返回 `true` 表示本次真的做了迁移并需要落盘。
+   * **解析器不可用时返回 `false` 且不置标记** —— 迁移推迟到数据库就绪后补跑。
+   */
+  private runScheduledReportMigration(stored: StoredRules): boolean {
+    if (stored.scheduledReportMigrated) return false
+    const tasks = this.readLegacyScheduledTasks()
+    if (!tasks.length) {
+      // 全新安装 / 从没用过旧定时日报：没有历史可迁，直接置位。
+      stored.scheduledReportMigrated = true
+      console.log('[Automation] 定时日报迁移 outcome=fresh_install')
+      return true
+    }
+    const resolver = this.legacyConversationResolver
+    if (!resolver) {
+      console.log(
+        `[Automation] 定时日报迁移已推迟：数据库尚未就绪，无法解析旧群标识（待迁 ${tasks.length} 条）`
+      )
+      return false
+    }
+    const { plans, summary } = planScheduledReportMigrationBatch(tasks, resolver, this.now())
+    this.lastScheduledMigrationSummary = summary
+    this.writeScheduledMigrationBackup(tasks, summary)
+    const existing = new Set(stored.rules.map((rule) => rule.id))
+    const migrated: AutomationRule[] = []
+    for (const plan of plans) {
+      // 幂等：同一 id 已存在就不再插入（重复执行迁移也不会多出规则）。
+      if (existing.has(plan.ruleId)) continue
+      existing.add(plan.ruleId)
+      migrated.push({
+        ...createDefaultScheduledReportRule(plan.createdAt, {
+          id: plan.ruleId,
+          name: plan.name,
+          enabled: plan.enabled,
+          config: plan.config
+        }),
+        createdAt: plan.createdAt,
+        updatedAt: plan.updatedAt
+      })
+    }
+    stored.rules = [...stored.rules, ...migrated]
+    stored.scheduledReportMigrated = true
+    // 只记数量与判定分布，**禁止**出现群名 / roomId / 群主昵称。
+    console.log(
+      `[Automation] 定时日报迁移 outcome=done total=${summary.total} migrated=${migrated.length}` +
+        ` lossless=${summary.lossless} needsReview=${summary.needsReview}` +
+        ` duplicatesSkipped=${summary.duplicatesSkipped}`
+    )
+    return true
+  }
+
+  /**
+   * 读旧定时日报**任务**文件。
+   *
+   * 只读、只在这里读一次；运行期其余代码**不再读** legacy（避免双读）。
+   * 读失败视为"没有历史任务"，不抛。
+   */
+  private readLegacyScheduledTasks(): LegacyScheduledReportTask[] {
+    try {
+      const value = fs.readJsonSync(
+        path.join(this.userDataPath(), LEGACY_SCHEDULED_DIR, LEGACY_SCHEDULED_TASKS_FILE)
+      ) as unknown
+      return Array.isArray(value) ? (value as LegacyScheduledReportTask[]) : []
+    } catch {
+      return []
+    }
+  }
+
+  /** 迁移前把旧任务**原样**落一份 backup（不做不可逆覆盖）。 */
+  private writeScheduledMigrationBackup(
+    tasks: LegacyScheduledReportTask[],
+    summary: ScheduledReportMigrationSummary
+  ): void {
+    try {
+      const filePath = path.join(
+        this.userDataPath(),
+        STORAGE_DIR,
+        SCHEDULED_MIGRATION_BACKUP_FILE
+      )
+      fs.ensureDirSync(path.dirname(filePath))
+      fs.writeJsonSync(
+        filePath,
+        {
+          migratedAt: this.now(),
+          summary,
+          // 备份里保留原始字段：这是**用户数据**，不是日志，不进任何用户可见界面。
+          legacyTasks: tasks
+        },
+        { spaces: 2 }
+      )
+    } catch (error) {
+      console.warn(`[Automation] 写入定时日报迁移备份失败: ${errorText(error)}`)
+    }
+  }
+
+  /**
+   * 读旧退群监控状态文件里的通知相关字段。
+   *
+   * 只读、只在这里读一次；运行期其余代码**不再读** legacy（避免双读）。
+   * 读失败视为"没有历史配置"，不抛。
+   */
+  private readLegacyLeaveNotificationState(): LegacyLeaveNotificationState | null {
+    try {
+      const raw = fs.readJsonSync(
+        path.join(this.userDataPath(), LEGACY_MONITOR_FILE)
+      ) as Partial<LegacyLeaveNotificationState>
+      if (!raw || typeof raw !== 'object') return null
+      return {
+        monitoredRoomIds: Array.isArray(raw.monitoredRoomIds) ? raw.monitoredRoomIds : [],
+        notificationRoomIds: Array.isArray(raw.notificationRoomIds) ? raw.notificationRoomIds : [],
+        ...(raw.notificationTemplate !== undefined
+          ? { notificationTemplate: raw.notificationTemplate }
+          : {})
+      }
+    } catch {
+      return null
+    }
+  }
+
+  /**
+   * 迁移前把旧配置原样落一份 backup。
+   *
+   * 不做不可逆覆盖：旧状态文件本身也**不删**，只是不再被运行时代码读取。
+   */
+  private writeMigrationBackup(
+    legacy: LegacyLeaveNotificationState,
+    plan: LeaveNotificationMigrationPlan
+  ): void {
+    try {
+      const filePath = path.join(this.userDataPath(), STORAGE_DIR, MIGRATION_BACKUP_FILE)
+      fs.ensureDirSync(path.dirname(filePath))
+      fs.writeJsonSync(
+        filePath,
+        {
+          migratedAt: this.now(),
+          outcome: plan.outcome,
+          // 备份里保留原始 roomId：这是**用户数据**，不是日志，不进任何用户可见界面。
+          legacy: {
+            monitoredRoomIds: legacy.monitoredRoomIds,
+            notificationRoomIds: legacy.notificationRoomIds,
+            notificationTemplate: legacy.notificationTemplate
+          },
+          applied: {
+            enabled: plan.enabled,
+            targetType: plan.target.type,
+            notifyScope: plan.notifyScope,
+            notifyRoomIds: plan.notifyRoomIds
+          }
+        },
+        { spaces: 2 }
+      )
+    } catch (error) {
+      console.warn(`[Automation] 写入退群通知迁移备份失败: ${errorText(error)}`)
+    }
+  }
+
+  /** 迁移日志：只允许出现判定结果与布尔量，**禁止** roomId / 模板正文。 */
+  private logMigration(outcome: string, plan: LeaveNotificationMigrationPlan | null): void {
+    if (!plan) {
+      console.log(`[Automation] 退群通知迁移 outcome=${outcome}`)
+      return
+    }
+    console.log(
+      `[Automation] 退群通知迁移 outcome=${outcome} enabled=${plan.enabled}` +
+        ` targetType=${plan.target.type} needsReview=${plan.targetNeedsReview}`
+    )
+  }
+
+  private rulesFilePath(): string {
+    return path.join(this.userDataPath(), STORAGE_DIR, RULES_FILE)
+  }
+
+  private persist(): void {
+    try {
+      const filePath = this.rulesFilePath()
+      fs.ensureDirSync(path.dirname(filePath))
+      fs.writeJsonSync(filePath, this.state, { spaces: 2 })
+    } catch (error) {
+      console.warn(
+        `[Automation] 保存规则失败: ${error instanceof Error ? error.message : String(error)}`
+      )
+      throw new AutomationRulePersistenceError()
+    }
+  }
+
+  private commitRules(rules: AutomationRule[]): void {
+    this.commitState({ ...this.state, rules })
+  }
+
+  private commitState(next: StoredRules): void {
+    const previous = this.state
+    this.state = next
+    try {
+      this.persist()
+    } catch (error) {
+      this.state = previous
+      throw error
+    }
+  }
+}
+
+function errorText(error: unknown): string {
+  return error instanceof Error ? error.message : String(error)
+}
+
+export const automationRuleStore = new AutomationRuleStore()
@@ -0,0 +1,266 @@
+import {
+  LEAVE_NOTIFICATION_TARGET_OPTIONS,
+  SCHEDULED_REPORT_TARGET_OPTIONS,
+  leaveNotificationTargetLabel,
+  scheduledReportTargetLabel
+} from '../../shared/automation'
+import { WECHAT_FILE_HELPER_USERNAME, isFileHelperUsername } from '../../shared/wechat-identities'
+
+/**
+ * **中性的微信发送目标解析**。
+ *
+ * 退群通知与定时日报的目标语义完全一致（来源会话 / 自己 / 文件传输助手 / 指定好友），
+ * 差的只是"来源会话"从哪来：
+ * - 退群通知 → 事件所在群（`event.conversationId`）；
+ * - 定时日报 → 该规则配置的日报来源群（`config.report.sourceConversationId`）。
+ *
+ * ⇒ 所以**只允许有一份实现**（只保留一套 resolve）。
+ * 两个业务各自只有一层薄适配（把参数摊平后交给这里），不重复任何判定逻辑。
+ *
+ * 五条规则（两边共用）：
+ * 1. 目标解析失败 → **报错**，绝不 fallback 到别的目标（尤其不许偷偷发文件助手 / 来源群）；
+ * 2. 只允许出现在选项表里的目标类型；
+ * 3. 显示名不带 wxid / roomId：日志与界面只出现用户能看懂的称呼；
+ * 4. 联系人必须**可发送**（排除群聊 / 公众号 / 文件传输助手 / 自己）；
+ * 5. 只有 `source_chat` 用得到来源会话，其余目标完全不看它。
+ */
+
+/** 解析所需的最小联系人结构（main 侧 `FormattedContact` 结构上可赋值给它）。 */
+export interface AutomationTargetContact {
+  m_nsUsrName: string
+  m_nsNickName?: string
+  md5?: string
+  type: 'user' | 'group'
+  isOfficialAccount?: boolean
+  wechatNickname?: string
+  remark?: string
+}
+
+/** 目标类型（两个业务共用同一组字面量）。 */
+export type AutomationTargetType = 'source_chat' | 'self' | 'file_transfer' | 'contact'
+
+export interface ResolvedAutomationTarget {
+  recipient: { type: 'group' | 'contact'; id: string; name: string }
+  /** 用户可读的目标名，直接进执行日志的 `detail`。 */
+  displayName: string
+}
+
+export type AutomationTargetResolution =
+  | { ok: true; target: ResolvedAutomationTarget }
+  | { ok: false; error: string }
+
+/** 联系人的展示名：备注 → 微信昵称 → 会话昵称 → 兜底称呼。**永不回落到 wxid**。 */
+export function automationContactDisplayName(contact: AutomationTargetContact): string {
+  return (
+    contact.remark?.trim() ||
+    contact.wechatNickname?.trim() ||
+    contact.m_nsNickName?.trim() ||
+    '指定好友'
+  )
+}
+
+/**
+ * 这个联系人能不能作为「指定好友」的发送目标。
+ *
+ * 排除：群聊、公众号（`gh_`）、文件传输助手、自己。
+ * UI 的联系人选择器与运行时的目标解析**共用这一条判定**，
+ * 否则会出现"能选但发不出去"。
+ */
+export function isSendableFriendContact(
+  contact: AutomationTargetContact,
+  selfWxid?: string
+): boolean {
+  if (contact.type !== 'user') return false
+  if (contact.isOfficialAccount) return false
+  const username = String(contact.m_nsUsrName || '').trim()
+  if (!username) return false
+  if (isFileHelperUsername(username)) return false
+  const self = String(selfWxid || '').trim()
+  if (self && username === self) return false
+  return true
+}
+
+/** 供 UI 用的「可选好友」过滤（与运行时同一套判定）。 */
+export function filterSendableFriendContacts<T extends AutomationTargetContact>(
+  contacts: T[],
+  selfWxid?: string
+): T[] {
+  return contacts.filter((contact) => isSendableFriendContact(contact, selfWxid))
+}
+
+/** 目标类型是否在（该业务的）选项表里。 */
+function isKnownTargetType(
+  type: string,
+  options: ReadonlyArray<{ type: string }>
+): boolean {
+  return options.some((option) => option.type === type)
+}
+
+/**
+ * 把群标识解析成**稳定会话 id**（`xxx@chatroom`）。
+ *
+ * 兼容三种历史形态：roomId 本身 / 会话 md5 / 群名。
+ * 解析不出来返回 `undefined` —— 调用方必须如实报错，**不许回落**。
+ */
+export function resolveGroupConversationId(
+  raw: string,
+  contacts: AutomationTargetContact[]
+): string | undefined {
+  const value = String(raw || '').trim()
+  if (!value) return undefined
+  if (value.endsWith('@chatroom')) return value
+  const matched = contacts.find(
+    (contact) =>
+      (contact.type === 'group' || contact.m_nsUsrName.endsWith('@chatroom')) &&
+      (contact.md5 === value ||
+        contact.m_nsUsrName === value ||
+        contact.m_nsNickName?.trim() === value)
+  )
+  const conversationId = String(matched?.m_nsUsrName || '').trim()
+  return conversationId.endsWith('@chatroom') ? conversationId : undefined
+}
+
+/** 群标识 → 可读群名（解析不到时返回空串，由调用方决定兜底文案）。 */
+export function resolveGroupDisplayName(
+  raw: string,
+  contacts: AutomationTargetContact[]
+): string {
+  const value = String(raw || '').trim()
+  if (!value) return ''
+  const matched = contacts.find(
+    (contact) =>
+      contact.m_nsUsrName === value ||
+      contact.md5 === value ||
+      contact.m_nsNickName?.trim() === value
+  )
+  return String(matched?.m_nsNickName || '').trim()
+}
+
+export interface ResolveAutomationTargetInput {
+  targetType: AutomationTargetType
+  /**
+   * `source_chat` 时使用：来源会话（退群事件所在群 / 日报来源群）。
+   * 允许是 roomId / 会话 md5 / 群名 —— 一律经 `resolveGroupConversationId` 收敛。
+   */
+  sourceConversationId?: string
+  /** 来源会话的可读名（拿不到时留空，由这里给兜底称呼）。 */
+  sourceDisplayName?: string
+  /** `contact` 时使用：稳定 id（wxid / username）。 */
+  contactId?: string
+  contacts: AutomationTargetContact[]
+  /** 当前登录账号的 wxid；拿不到时传空串。 */
+  selfWxid?: string
+}
+
+/**
+ * 目标解析的**文案与类型表**（按业务注入）。
+ *
+ * 为什么把文案参数化而不是各写一份 switch：两个业务的**判定逻辑必须一模一样**，
+ * 只有"主语"不同。文案集中放在这里，判定仍然只有一份。
+ */
+export interface AutomationTargetMessages {
+  /** 目标类型表的单一来源（与 UI 选项同源）。 */
+  options: ReadonlyArray<{ type: string }>
+  /** 类型非法（伪造值 / 未知值）。 */
+  invalidTarget: string
+  /** `source_chat` 但来源会话拿不到。 */
+  sourceMissing: string
+  /** `source_chat` 且来源会话没有可读名时，`recipient.name` 的兜底。 */
+  sourceRecipientName: string
+  /** `source_chat` 且来源会话没有可读名时，`displayName` 的兜底。 */
+  sourceDisplayName: string
+  /** `self` 但拿不到自身身份。 */
+  selfMissing: string
+  /** `contact` 但没选联系人。 */
+  contactNotChosen: string
+  /** `contact` 但联系人失效 / 不可发送。 */
+  contactUnavailable: string
+}
+
+/**
+ * 中性目标解析。**唯一的实现**。
+ */
+export function resolveAutomationTarget(
+  input: ResolveAutomationTargetInput,
+  messages: AutomationTargetMessages
+): AutomationTargetResolution {
+  const type = input.targetType
+  if (!type || !isKnownTargetType(type, messages.options)) {
+    return { ok: false, error: messages.invalidTarget }
+  }
+
+  switch (type) {
+    case 'source_chat': {
+      const conversationId = resolveGroupConversationId(
+        String(input.sourceConversationId || ''),
+        input.contacts
+      )
+      if (!conversationId) return { ok: false, error: messages.sourceMissing }
+      const explicitName = String(input.sourceDisplayName || '').trim()
+      const resolvedName = resolveGroupDisplayName(conversationId, input.contacts)
+      const name = explicitName || resolvedName || messages.sourceRecipientName
+      const displayName = explicitName || resolvedName || messages.sourceDisplayName
+      return {
+        ok: true,
+        target: {
+          recipient: { type: 'group', id: conversationId, name },
+          displayName
+        }
+      }
+    }
+
+    case 'self': {
+      const selfWxid = String(input.selfWxid || '').trim()
+      // 明确报错，不 fallback。
+      if (!selfWxid) return { ok: false, error: messages.selfMissing }
+      return {
+        ok: true,
+        target: {
+          recipient: { type: 'contact', id: selfWxid, name: '我' },
+          displayName: '我'
+        }
+      }
+    }
+
+    case 'file_transfer':
+      return {
+        ok: true,
+        target: {
+          recipient: {
+            type: 'contact',
+            id: WECHAT_FILE_HELPER_USERNAME,
+            name: '文件传输助手'
+          },
+          displayName: '文件传输助手'
+        }
+      }
+
+    case 'contact': {
+      const contactId = String(input.contactId || '').trim()
+      if (!contactId) return { ok: false, error: messages.contactNotChosen }
+      const contact = input.contacts.find(
+        (item) => String(item.m_nsUsrName || '').trim() === contactId
+      )
+      if (!contact || !isSendableFriendContact(contact, input.selfWxid)) {
+        // 之前选择的联系人被删除 / 不可发送 / 找不到时，规则不偷偷改发别处。
+        return { ok: false, error: messages.contactUnavailable }
+      }
+      const displayName = automationContactDisplayName(contact)
+      return {
+        ok: true,
+        target: {
+          recipient: { type: 'contact', id: contactId, name: displayName },
+          displayName
+        }
+      }
+    }
+  }
+}
+
+/** 退群通知的已知目标类型表（与文案同源）。 */
+export const LEAVE_NOTIFICATION_TARGET_TYPES = LEAVE_NOTIFICATION_TARGET_OPTIONS
+
+/** 定时日报的已知目标类型表（与文案同源）。 */
+export const SCHEDULED_REPORT_TARGET_TYPES = SCHEDULED_REPORT_TARGET_OPTIONS
+
+export { leaveNotificationTargetLabel, scheduledReportTargetLabel }
@@ -1036,6 +1036,20 @@ export async function getGroupSnapshotAsync(userMd5: string): Promise<GroupSnaps
  }
 }

+/**
+ * 把群会话 md5 解析成原生接口真正需要的 roomId（`xxx@chatroom`）。
+ *
+ * 存在的理由：群员统计从界面拿到的是会话 md5（与 `getGroupSnapshot(userMd5)` 同口径），
+ * 而成员相关的原生接口吃的是 username。没有这层导出，调用方就得自己去碰
+ * `dbRef.getWcdb4Client()`，等于让服务层绕开 chat-service 的封装。
+ * 非群会话或解析不到时返回 null。
+ */
+export function resolveGroupRoomId(userMd5: string): string | null {
+  if (!dbRef) return null
+  const roomId = dbRef.getWcdb4Client().getUsernameByMd5(userMd5)
+  return roomId && roomId.endsWith('@chatroom') ? roomId : null
+}
+
 /** 退群检测专用轻量读取，不执行成员名称或头像 hydration。 */
 export async function getGroupMemberIdsAsync(
  roomId: string
@@ -2,19 +2,17 @@ import { app, BrowserWindow } from 'electron'
 import fs from 'fs-extra'
 import path from 'path'
 import * as chat from './chat-service'
-import { wechatActionGateway, type WechatActionGateway } from './wechat-action-gateway'
 import {
-  buildMemberLeftNotification,
  findRemovedGroupMembers,
  groupExitMemberName,
-  normalizeGroupExitNotificationTemplate,
-  validateGroupExitNotificationTemplate,
  type GroupExitMonitorEvent,
  type GroupExitMonitorMember,
-  type GroupExitNotificationStatus,
-  type GroupExitNotificationState,
  type GroupExitMonitorState
 } from '../../shared/group-exit-monitor'
+import {
+  toGroupMemberExitedEvent,
+  type GroupMemberExitedEvent
+} from '../../shared/group-exit-event'

 type StoredState = {
  enabled?: boolean
@@ -23,9 +21,14 @@ type StoredState = {
  lastReadAt?: number
  monitorSelectionConfigured?: boolean
  monitoredRoomIds?: string[]
-  notificationRoomIds?: string[]
-  notificationTemplate?: unknown
  snapshots?: Partial<StoredGroupSnapshot>[]
+  /*
+   * 历史遗留字段（`notificationRoomIds` / `notificationTemplate`）**刻意不再声明**。
+   *
+   * 它们已经迁到「自动化 → 退群通知」规则里，本服务运行期不再读、不再写。
+   * 旧状态文件原样保留在磁盘上（迁移时另有一份 backup），只是没有任何读取方 ——
+   * 这是"不双读"的落地方式。迁移逻辑在 `automation-rule-store.ts`。
+   */
 }

 type GroupSnapshotRecord = {
@@ -57,19 +60,30 @@ type StoredGroupSnapshot = Pick<

 const DB_CHANGE_DEBOUNCE_MS = 350
 const DUPLICATE_WINDOW_MS = 2 * 60 * 1000
+/**
+ * **只约束「一次状态回传带多少条」**，不是存储上限。
+ *
+ * 事件现在写在 append-only 的 JSONL 里（`eventsPath()`），**永久保留、不做截断**。
+ * 之所以必须把「存储」和「回传」分开：状态文件是整体重写的（`save()`），
+ * 事件留在里面时每来一条新事件都要重写整个文件 —— 越写越慢。
+ * 而 `getState()` 每次都把结果推过 IPC，也不能无上限。
+ */
 const MAX_EVENTS = 500

+/**
+ * 退群事件处理器。
+ *
+ * 由主进程注入成 `automationService.handleGroupExit` —— 本服务**不再自己发送**。
+ * 契约：**必须自己吞掉异常**（实现方负责），本服务也会再兜一层。
+ */
+export type GroupExitEventHandler = (event: GroupMemberExitedEvent) => void | Promise<void>
+
 export interface GroupExitMonitorServiceDependencies {
-  actionGateway?: GroupExitActionGateway
+  onGroupExit?: GroupExitEventHandler
 }

-type GroupExitActionGateway = Pick<WechatActionGateway, 'execute'> &
-  Partial<
-    Pick<WechatActionGateway, 'registerMemberEvent' | 'registerMemberEvents' | 'clearMemberEvents'>
-  >
-
 class GroupExitMonitorService {
-  private readonly actionGateway: GroupExitActionGateway
+  private onGroupExit: GroupExitEventHandler | undefined
  private active = false
  private enabled = true
  private nativeMonitorActive = false
@@ -87,7 +101,6 @@ class GroupExitMonitorService {
  private events: GroupExitMonitorEvent[] = []
  private monitorSelectionConfigured = true
  private monitoredRoomIds = new Set<string>()
-  private notificationRoomIds = new Set<string>()
  private loaded = false
  private accountRoot = ''
  private groupNamesByRoomId = new Map<string, string>()
@@ -95,24 +108,34 @@ class GroupExitMonitorService {
  private eventSequence = 0
  private scopeGeneration = 0
  private legacyFallbackLogged = false
-  private notificationTemplate = normalizeGroupExitNotificationTemplate(undefined)

  constructor(deps: GroupExitMonitorServiceDependencies = {}) {
-    this.actionGateway = deps.actionGateway || wechatActionGateway
+    this.onGroupExit = deps.onGroupExit
+  }
+
+  /**
+   * 注入退群事件处理器（主进程在 AutomationService 初始化后调用）。
+   *
+   * 用 setter 而不是构造参数：AutomationService 依赖 `Wcdb4Client`，
+   * 要等数据库解锁后才存在，而本服务是模块级单例。
+   */
+  setGroupExitHandler(handler: GroupExitEventHandler | undefined): void {
+    this.onGroupExit = handler
  }

  getState(): GroupExitMonitorState {
    this.ensureLoaded()
    return {
-      events: [...this.events],
+      // this.events 恒为「新在前」，所以这里取到的就是**最新**的 MAX_EVENTS 条。
+      events: this.events.slice(0, MAX_EVENTS),
+      /** 永久保留的事件总数（`events` 只是最新的一批）。 */
+      totalEventCount: this.events.length,
      enabled: this.enabled,
      running: this.enabled && this.active && chat.isReady(),
      nativeMonitorActive: this.nativeMonitorActive,
      monitoredGroupCount: this.snapshots.size,
      monitorSelectionConfigured: this.monitorSelectionConfigured,
      monitoredRoomIds: Array.from(this.monitoredRoomIds),
-      notificationRoomIds: Array.from(this.notificationRoomIds),
-      notificationTemplate: this.notificationTemplate,
      lastCheckedAt: this.lastCheckedAt,
      lastReadAt: this.lastReadAt,
      unreadCount: this.events.filter((event) => event.detectedAt > this.lastReadAt).length
@@ -132,12 +155,11 @@ class GroupExitMonitorService {
      (this.accountRoot && !sameAccountRoot(currentRoot, this.accountRoot)) ||
      (!this.accountRoot && this.snapshots.size > 0)
    ) {
-      this.actionGateway.clearMemberEvents?.()
      this.events = []
+      this.rewriteEventsToDisk([])
      this.lastReadAt = 0
      this.monitorSelectionConfigured = true
      this.monitoredRoomIds.clear()
-      this.notificationRoomIds.clear()
      this.snapshots.clear()
      this.groupNamesByRoomId.clear()
    }
@@ -208,75 +230,116 @@ class GroupExitMonitorService {
    return this.getState()
  }

-  async setMonitoredRoomIds(
-    roomIds: string[],
-    notificationRoomIds: string[] = []
-  ): Promise<GroupExitMonitorState> {
+  /**
+   * 保存**监控范围**。
+   *
+   * 参数只剩监控目标 —— 旧版第二个参数（通知群聊）已经迁到自动化规则里，
+   * 本服务不再持有任何通知配置。
+   */
+  async setMonitoredRoomIds(roomIds: string[]): Promise<GroupExitMonitorState> {
+    return this.configure({ monitoredRoomIds: roomIds })
+  }
+
+  /**
+   * Atomically update the monitor scope and enabled state.
+   *
+   * Agent API PATCH validates the whole request before calling this method. Keeping the
+   * two mutations in one service operation also means persistence and the renderer broadcast
+   * observe one final configuration rather than an intermediate half-patched state.
+   */
+  async configure(configuration: {
+    monitoredRoomIds?: string[]
+    enabled?: boolean
+  }): Promise<GroupExitMonitorState> {
    this.ensureLoaded()
-    this.eventSequence += 1
-    this.scopeGeneration += 1
-    this.monitorSelectionConfigured = true
-    const nextMonitoredRoomIds = normalizeRoomIds(roomIds)
-    for (const roomId of this.snapshots.keys()) {
-      if (!nextMonitoredRoomIds.has(roomId)) this.snapshots.delete(roomId)
+    const hasScope = configuration.monitoredRoomIds !== undefined
+    const hasEnabled = configuration.enabled !== undefined
+    if (!hasScope && !hasEnabled) return this.getState()
+    if (hasEnabled && !hasScope && this.enabled === configuration.enabled) return this.getState()
+
+    if (hasScope) {
+      this.eventSequence += 1
+      this.scopeGeneration += 1
+      this.monitorSelectionConfigured = true
+      const nextMonitoredRoomIds = normalizeRoomIds(configuration.monitoredRoomIds || [])
+      for (const roomId of this.snapshots.keys()) {
+        if (!nextMonitoredRoomIds.has(roomId)) this.snapshots.delete(roomId)
+      }
+      for (const roomId of this.hydrationQueue) {
+        if (!nextMonitoredRoomIds.has(roomId)) this.hydrationQueue.delete(roomId)
+      }
+      this.monitoredRoomIds = nextMonitoredRoomIds
+      this.groupNamesRefreshPending = true
+      this.lastCheckedAt = undefined
    }
-    for (const roomId of this.hydrationQueue) {
-      if (!nextMonitoredRoomIds.has(roomId)) this.hydrationQueue.delete(roomId)
+
+    if (hasEnabled && this.enabled !== configuration.enabled) {
+      this.enabled = configuration.enabled === true
+      this.eventSequence += 1
+      this.scopeGeneration += 1
+      this.checkQueued = false
+      this.hydrationQueue.clear()
+      if (this.changeTimer) clearTimeout(this.changeTimer)
+      this.changeTimer = null
+
+      if (this.enabled) {
+        // 用户主动暂停期间的成员变化不补报；重新开启后从当前状态建立新基线。
+        this.snapshots.clear()
+        this.lastCheckedAt = undefined
+        this.groupNamesRefreshPending = true
+      }
    }
-    this.monitoredRoomIds = nextMonitoredRoomIds
-    this.groupNamesRefreshPending = true
-    const requestedNotifications = normalizeRoomIds(notificationRoomIds)
-    this.notificationRoomIds = new Set(
-      Array.from(requestedNotifications).filter((roomId) => this.monitoredRoomIds.has(roomId))
-    )
-    this.lastCheckedAt = undefined
+
    this.save()
    this.broadcast()
-    // 建立基线放到后台，保存配置可以立即返回。
-    if (this.enabled && this.active) void this.check()
+    // 仅修改范围时保持原有的后台基线行为；涉及 enabled 的 PATCH 等待一次检查，
+    // 让调用方拿到的是最终运行状态。
+    if (this.enabled && this.active && chat.isReady()) {
+      if (hasEnabled) await this.check()
+      else void this.check()
+    }
    return this.getState()
  }

  async setEnabled(enabled: boolean): Promise<GroupExitMonitorState> {
-    this.ensureLoaded()
-    if (this.enabled === enabled) return this.getState()
-
-    this.enabled = enabled
-    this.eventSequence += 1
-    this.scopeGeneration += 1
-    this.checkQueued = false
-    this.hydrationQueue.clear()
-    if (this.changeTimer) clearTimeout(this.changeTimer)
-    this.changeTimer = null
-
-    if (enabled) {
-      // 用户主动暂停期间的成员变化不补报；重新开启后从当前状态建立新基线。
-      this.snapshots.clear()
-      this.lastCheckedAt = undefined
-      this.groupNamesRefreshPending = true
-    }
-    this.save()
-    this.broadcast()
-    if (enabled && this.active && chat.isReady()) await this.check()
-    return this.getState()
+    return this.configure({ enabled })
  }

-  setNotificationTemplate(value: unknown): GroupExitMonitorState {
+  /**
+   * 按群 / 时间范围查退群事件（档案合并展示用）。
+   *
+   * 与 `getState()` 的分工：后者只带回最近 `MAX_EVENTS` 条、且是**给管理页**看的概览；
+   * 档案要的是「某个群在这段时间里的全部事件」，所以单独开一个查询入口，
+   * 直接打在内存里的完整历史上（事件是永久保留的）。
+   *
+   * 返回**按时间升序**（旧 → 新），与档案消息流的顺序一致。
+   */
+  listEvents(
+    query: { roomId?: string; sinceMs?: number; untilMs?: number; limit?: number } = {}
+  ): GroupExitMonitorEvent[] {
    this.ensureLoaded()
-    const result = validateGroupExitNotificationTemplate(value)
-    if (!result.valid || !result.template) {
-      throw new Error(result.error || '退群监测模板无效')
+    const roomId = String(query.roomId || '').trim()
+    const since = Number(query.sinceMs)
+    const until = Number(query.untilMs)
+    const limit = Number(query.limit)
+
+    // this.events 是倒序（新在前）。
+    let result = [...this.events].reverse()
+    if (roomId) result = result.filter((event) => event.roomId === roomId)
+    if (Number.isFinite(since)) result = result.filter((event) => event.detectedAt >= since)
+    if (Number.isFinite(until)) result = result.filter((event) => event.detectedAt <= until)
+    // 超量时保留**最近**的一批（尾部即最新）。
+    if (Number.isFinite(limit) && limit > 0 && result.length > limit) {
+      result = result.slice(-limit)
    }
-    this.notificationTemplate = result.template
-    this.save()
-    this.broadcast()
-    return this.getState()
+    return result
  }

  clearEvents(): GroupExitMonitorState {
    this.ensureLoaded()
    this.events = []
-    this.actionGateway.clearMemberEvents?.()
+    // 磁盘上的 append-only 历史也要清掉，否则下次启动又读回来了。
+    this.rewriteEventsToDisk([])
    this.lastReadAt = Date.now()
    this.save()
    this.broadcast()
@@ -399,7 +462,7 @@ class GroupExitMonitorService {
    groups: GroupMembershipRecord[],
    scopeGeneration: number
  ): Promise<number> {
-    const notifications: Array<{ group: GroupSnapshotRecord; event: GroupExitMonitorEvent }> = []
+    const exits: GroupExitMonitorEvent[] = []
    let changedGroups = 0
    for (const membership of groups) {
      if (!this.enabled || !this.active || scopeGeneration !== this.scopeGeneration) {
@@ -430,10 +493,9 @@ class GroupExitMonitorService {
      if (previous && next.members.length < previous.members.length) {
        const removed = findRemovedGroupMembers(previous.members, next.members)
        for (const member of removed) {
+          // 一人一条事件、一条通知 —— 保持既有产品语义，不聚合。
          const event = this.recordExit(next, member, previous.members.length, next.members.length)
-          if (event && this.notificationRoomIds.has(next.roomId)) {
-            notifications.push({ group: next, event })
-          }
+          if (event) exits.push(event)
        }
      }

@@ -448,12 +510,11 @@ class GroupExitMonitorService {
      return changedGroups
    }
    this.lastCheckedAt = Date.now()
-    // 先把事件和新基线作为同一检查点落盘，再执行可失败的通知动作。
+    // 先把事件和新基线作为同一检查点落盘，再交给自动化。
+    // 「退群事实已记录」与「通知发送成功」是两件独立的事。
    this.save()
    this.broadcast()
-    if (notifications.length) {
-      await Promise.all(notifications.map(({ group, event }) => this.notifyGroup(group, event)))
-    }
+    for (const event of exits) this.emitGroupExit(event)
    return changedGroups
  }

@@ -589,129 +650,145 @@ class GroupExitMonitorService {
      currentCount,
      delta: currentCount - previousCount,
      message,
-      detectedAt,
-      notificationStatus: 'not_requested'
+      detectedAt
    }
-    this.actionGateway.registerMemberEvent?.(event)
-    this.events = [event, ...this.events].slice(0, MAX_EVENTS)
+    // 内存按时间倒序（新事件在前）；磁盘**只追加这一条**，不重写历史。
+    // 这里不再有 `.slice(0, MAX_EVENTS)` —— 事件是永久保留的。
+    this.events = [event, ...this.events]
+    this.appendEventsToDisk([event])
    console.log(
      `[GroupMonitor] detected member exit roomId=${group.roomId} member=${member.wxid} ${previousCount}->${currentCount}`
    )
    return event
  }

-  private async notifyGroup(
-    group: GroupSnapshotRecord,
-    event: GroupExitMonitorEvent,
-    idempotencyKey = `member_left_notification:${event.id}`
-  ): Promise<void> {
-    event.notificationStatus = 'pending'
-    event.notification = { status: 'pending' }
-    this.save()
-    this.broadcast()
+  /**
+   * 把退群事件交给自动化 —— **不等待**。
+   *
+   * 三条约束：
+   * 1. **绝不 await**：快照扫描与成员 diff 不能被微信发送耗时拖住；
+   * 2. **异常必须被捕获**：`void promise` 漏掉 `.catch` 会变成 unhandled rejection；
+   * 3. **不影响退群事实**：事件与快照在同一检查点已经先落盘，通知失败不回滚记录。
+   */
+  private emitGroupExit(event: GroupExitMonitorEvent): void {
+    const handler = this.onGroupExit
+    if (!handler) return
    try {
-      const result = await this.actionGateway.execute({
-        idempotencyKey,
-        origin: 'member_monitor',
-        purpose: 'member_left_notification',
-        triggerType: 'automation',
-        sourceId: event.id,
-        recipient: {
-          type: 'group',
-          id: group.roomId,
-          name: group.groupName
-        },
-        content: {
-          type: 'text',
-          text: buildMemberLeftNotification(event, this.notificationTemplate)
-        },
-        metadata: {
-          memberId: event.memberWxid,
-          memberName: event.memberName,
-          detectedAt: event.detectedAt,
-          eventType: 'member_left',
-          eventRoomId: event.roomId
-        }
-      })
-      const notification: GroupExitNotificationState = {
-        status: result.status,
-        actionId: result.actionId,
-        decision: result.decision,
-        ...(result.errorCode ? { errorCode: result.errorCode } : {}),
-        ...(result.reason ? { reason: result.reason } : {}),
-        startedAt: result.startedAt,
-        finishedAt: result.finishedAt
-      }
-      event.notificationStatus = result.status
-      event.notification = notification
-      if (result.status !== 'sent') {
+      void Promise.resolve(
+        handler(toGroupMemberExitedEvent(event))
+      ).catch((error) => {
        console.warn(
-          `[GroupMonitor] 群聊通知未发送 roomId=${group.roomId} status=${result.status} code=${result.errorCode || ''}`
+          `[GroupMonitor] 退群通知处理失败 eventId=${event.id}: ${
+            error instanceof Error ? error.message : String(error)
+          }`
        )
-      }
+      })
    } catch (error) {
-      event.notificationStatus = 'failed'
-      event.notification = {
-        status: 'failed',
-        errorCode: 'UNKNOWN',
-        reason: error instanceof Error ? error.message : String(error)
-      }
+      // handler 同步抛出的情况（`handleGroupExit` 本身不抛，这里是防御性兜底）。
      console.warn(
-        `[GroupMonitor] 群聊通知异常 roomId=${group.roomId}:`,
-        error instanceof Error ? error.message : String(error)
+        `[GroupMonitor] 退群通知处理异常 eventId=${event.id}: ${
+          error instanceof Error ? error.message : String(error)
+        }`
      )
-    } finally {
-      this.save()
-      this.broadcast()
    }
  }

-  async resendEvent(eventId: string): Promise<GroupExitMonitorState> {
-    const event = this.events.find((item) => item.id === eventId)
-    if (!event) throw new Error('退群动态不存在')
-    await this.notifyGroup(
-      { roomId: event.roomId, groupName: event.groupName } as GroupSnapshotRecord,
-      event,
-      `member_left_notification:${event.id}:retry:${Date.now()}`
-    )
-    return this.getState()
-  }
-
  private filePath(): string {
    return path.join(app.getPath('userData'), 'group-exit-monitor.json')
  }

+  /**
+   * 退群事件的 append-only 存储（每行一条 JSON）。
+   *
+   * 与状态文件分开，因为两者的写入模式完全不同：
+   * - **状态**（开关 / 监控范围 / 快照 / 模板）小、且总是整体重写；
+   * - **事件**只增不改，且要求**永久保留**。
+   * 混在一个文件里时，每来一条事件都要把整部历史重新序列化写一遍 —— 越写越慢。
+   */
+  private eventsPath(): string {
+    return path.join(app.getPath('userData'), 'group-exit-monitor-events.jsonl')
+  }
+
+  /** 读全量历史事件。单行损坏只跳过该行，不让整部历史读不出来。 */
+  private readEventsFromDisk(): GroupExitMonitorEvent[] {
+    let raw = ''
+    try {
+      raw = fs.readFileSync(this.eventsPath(), 'utf8')
+    } catch {
+      return []
+    }
+    const events: GroupExitMonitorEvent[] = []
+    for (const line of raw.split('\n')) {
+      const trimmed = line.trim()
+      if (!trimmed) continue
+      try {
+        events.push(JSON.parse(trimmed) as GroupExitMonitorEvent)
+      } catch {
+        // 跳过坏行
+      }
+    }
+    return events
+  }
+
+  /** 只追加新增的那几行，不重写历史。 */
+  private appendEventsToDisk(events: GroupExitMonitorEvent[]): void {
+    if (!events.length) return
+    try {
+      fs.ensureDirSync(path.dirname(this.eventsPath()))
+      const payload = events.map((event) => `${JSON.stringify(event)}\n`).join('')
+      fs.appendFileSync(this.eventsPath(), payload, 'utf8')
+    } catch (error) {
+      console.warn('[GroupMonitor] 追加退群事件失败:', error)
+    }
+  }
+
+  /** 整体重写事件文件（清空、切换账号、老数据迁移时使用）。 */
+  private rewriteEventsToDisk(events: GroupExitMonitorEvent[]): void {
+    try {
+      fs.ensureDirSync(path.dirname(this.eventsPath()))
+      const payload = events.map((event) => `${JSON.stringify(event)}\n`).join('')
+      fs.writeFileSync(this.eventsPath(), payload, 'utf8')
+    } catch (error) {
+      console.warn('[GroupMonitor] 重写退群事件失败:', error)
+    }
+  }
+
  private ensureLoaded(): void {
    if (this.loaded) return
    this.loaded = true
    try {
      const stored = fs.readJsonSync(this.filePath()) as StoredState
      this.enabled = stored.enabled !== false
-      this.events = normalizeEvents(stored.events)
-      this.actionGateway.registerMemberEvents?.(this.events)
+      // 事件从 append-only 文件读；状态文件不再承载它们。
+      const fromDisk = this.readEventsFromDisk()
+      const legacy = normalizeEvents(stored.events)
+      if (legacy.length && !fromDisk.length) {
+        // 老版本把事件塞在状态文件里 —— 一次性迁移过去，避免这批历史丢失。
+        // 落盘按时间**升序**（旧 → 新），与之后 append 的方向一致，避免在
+        // append-only 文件开头留下一段方向相反的旧历史（历史行序错乱的来源）。
+        this.rewriteEventsToDisk(
+          [...legacy].sort((left, right) => left.detectedAt - right.detectedAt)
+        )
+        this.events = sortEventsNewestFirst(legacy)
+      } else {
+        // 磁盘行序不保证时间有序（迁移段与追加段方向相反），读回后必须显式重建
+        // 「新在前」这个内存不变量，否则列表顶部会恒为最旧的一批。
+        this.events = sortEventsNewestFirst(normalizeEvents(fromDisk))
+      }
      this.lastReadAt = Number(stored.lastReadAt) || 0
      this.accountRoot = String(stored.accountRoot || '')
      // 没有显式范围时按空范围处理，保留已有选择。
      this.monitorSelectionConfigured = true
      this.monitoredRoomIds = normalizeRoomIds(stored.monitoredRoomIds || [])
-      this.notificationRoomIds = new Set(
-        Array.from(normalizeRoomIds(stored.notificationRoomIds || [])).filter((roomId) =>
-          this.monitoredRoomIds.has(roomId)
-        )
-      )
-      this.notificationTemplate = normalizeGroupExitNotificationTemplate(
-        stored.notificationTemplate
-      )
      this.snapshots = normalizeSnapshots(stored.snapshots, this.monitoredRoomIds)
    } catch {
-      // 首次启动或文件损坏时从空记录开始。
-      this.events = []
+      // 首次启动或状态文件损坏时从空记录开始 —— 但事件在独立文件里，
+      // 不该被状态文件的问题连累，仍然读回来。
+      this.events = sortEventsNewestFirst(normalizeEvents(this.readEventsFromDisk()))
      this.enabled = true
      this.lastReadAt = 0
      this.monitorSelectionConfigured = true
      this.monitoredRoomIds.clear()
-      this.notificationRoomIds.clear()
-      this.notificationTemplate = normalizeGroupExitNotificationTemplate(undefined)
      this.snapshots.clear()
    }
  }
@@ -725,12 +802,11 @@ class GroupExitMonitorService {
        {
          accountRoot: this.accountRoot,
          enabled: this.enabled,
-          events: this.events,
+          // 事件**不在这里**：它们走 append-only 的 JSONL（见 `eventsPath()`）。
+          // 放进状态文件会让每新增一条事件都把整部历史重写一遍。
          lastReadAt: this.lastReadAt,
          monitorSelectionConfigured: this.monitorSelectionConfigured,
          monitoredRoomIds: Array.from(this.monitoredRoomIds),
-          notificationRoomIds: Array.from(this.notificationRoomIds),
-          notificationTemplate: this.notificationTemplate,
          snapshots: Array.from(this.snapshots.values(), toStoredSnapshot)
        },
        { spaces: 2 }
@@ -927,39 +1003,28 @@ function normalizeEvents(
        ? Number(value.delta)
        : currentCount - previousCount,
      message: String(value.message || `${memberName}退出了${groupName}`),
-      detectedAt,
-      ...(value.notificationStatus
-        ? { notificationStatus: normalizeNotificationStatus(value.notificationStatus) }
-        : {}),
-      ...(value.notification && typeof value.notification === 'object'
-        ? { notification: normalizeNotification(value.notification) }
-        : {})
+      detectedAt
    })
-    if (normalized.length >= MAX_EVENTS) break
+    // 不再按 MAX_EVENTS 截断：事件是永久保留的，截在这里等于每次启动都丢掉历史。
+    // 历史上的 `notificationStatus` / `notification` 字段被**丢弃**：
+    // 通知状态已归 Automation 执行日志，退群监控不再持有它。
  }
  return normalized
 }

-function normalizeNotificationStatus(value: unknown): GroupExitNotificationStatus {
-  const status = String(value || '').trim()
-  return status === 'pending' || status === 'sent' || status === 'blocked' || status === 'failed'
-    ? status
-    : 'not_requested'
-}
-
-function normalizeNotification(value: object): GroupExitNotificationState {
-  const input = value as Partial<GroupExitNotificationState>
-  return {
-    status: normalizeNotificationStatus(input.status),
-    ...(input.actionId ? { actionId: String(input.actionId) } : {}),
-    ...(input.decision === 'allow' || input.decision === 'block'
-      ? { decision: input.decision }
-      : {}),
-    ...(input.errorCode ? { errorCode: String(input.errorCode) } : {}),
-    ...(input.reason ? { reason: String(input.reason) } : {}),
-    ...(input.startedAt ? { startedAt: String(input.startedAt) } : {}),
-    ...(input.finishedAt ? { finishedAt: String(input.finishedAt) } : {})
-  }
+/**
+ * 事件在内存里恒定保持「**新在前**」。
+ *
+ * 这个不变量有三个依赖方：`recordExit` 的 `[event, ...this.events]` 写入方向、
+ * `listEvents()` 的 `.reverse()`（它假定内存是倒序，反转后得到升序）、
+ * 以及 `getState()` 的 `slice(0, MAX_EVENTS)`（要求取到的是**最新**的一批）。
+ *
+ * 必须显式重建它：磁盘是 append-only，行序由「迁移写入的历史 + 之后追加的新事件」
+ * 决定，两段方向相反，整体不保证时间有序。直接信任文件行序会让列表顶部恒为最旧的
+ * 一批，并让 `slice(0, MAX_EVENTS)` 恰好把最新的事件截掉。
+ */
+function sortEventsNewestFirst(events: GroupExitMonitorEvent[]): GroupExitMonitorEvent[] {
+  return [...events].sort((left, right) => right.detectedAt - left.detectedAt)
 }

 function isContactEvent(rawPayload: string): boolean {
@@ -0,0 +1,208 @@
+import { isKnowledgeFresh } from '../../shared/knowledge'
+import {
+  GROUP_STATS_LIMITATION,
+  GROUP_STATS_STALE_LIMITATION,
+  type GroupMemberStatsActiveMember,
+  type GroupMemberStatsQuery,
+  type GroupMemberStatsResult,
+  type GroupMemberStatsSilentMember,
+  type GroupStatsFreshness
+} from '../../shared/group-stats'
+import type { KnowledgeSearchService } from '../knowledge/knowledge-search-service'
+import * as chat from './chat-service'
+
+/**
+ * 触发一次追赶同步的最小间隔。
+ *
+ * 面板是可以反复开关的交互入口，而追赶同步跑的是真实增量 pass（会读 WCDB）。
+ * 没有这个节流，连续点击就等于连续触发索引。
+ */
+const CATCH_UP_MIN_INTERVAL_MS = 30_000
+
+/**
+ * 追赶的有界等待预算。
+ *
+ * 与 Query 侧同一取舍：全量追赶可能以分钟计，交互查询绝不能无限等。
+ * 预算用完之后**如实**把结果标成不完整，而不是假装完整。
+ */
+const FRESHNESS_WAIT_BUDGET_MS = 2_000
+
+/**
+ * 群员统计。
+ *
+ * 数据两路来源，职责严格分开：
+ * - **当前成员名单** 走 WCDB 的轻路径（只取 wxid + 显示名，**不** hydrate 头像）；
+ * - **发言聚合** 走 Knowledge 派生库的 `GROUP BY sender_id`，不读 WCDB 原始消息。
+ *
+ * 「未发言」= 当前成员 ∖ 窗口内有发言的 sender。这个定义**不能**简化成
+ * 「整段窗口都没说话」—— 入群/退群时间在源头就不存在（整个 DB 没有该字段），
+ * 所以结论必须带上 limitation。
+ */
+export class GroupStatsService {
+  constructor(private readonly knowledge: KnowledgeSearchService) {}
+
+  async getMemberStats(query: GroupMemberStatsQuery): Promise<GroupMemberStatsResult> {
+    const { userMd5, startTime, endTime } = query
+    const limitations = [GROUP_STATS_LIMITATION]
+
+    const roomId = chat.resolveGroupRoomId(userMd5)
+    if (!roomId) {
+      // 不是群 / md5 解析不到：如实返回空，不编造统计。
+      return this.build({
+        userMd5,
+        startTime,
+        endTime,
+        freshness: 'unknown',
+        limitations,
+        members: [],
+        activeMembers: [],
+        silentMembers: [],
+        unattributedMessages: 0,
+        excludedSystemMessages: 0,
+        firstMessageTime: null
+      })
+    }
+
+    // 1) 当前群成员 —— 轻路径。只取 wxid，再按需解析显示名；绝不走
+    //    `getGroupSnapshotAsync`（它会 materialize 整群并 hydrate 头像，实测 8 群 ≈ 42s）。
+    const membership = await chat.getGroupMemberIdsAsync(roomId)
+    const memberIds = membership?.memberIds ?? []
+    const members = await chat.getGroupMemberNamesAsync(userMd5, memberIds)
+
+    // 2) 发言聚合（只统计当前群；系统消息在 SQL 层就被排除）
+    let { result, sourceLatestAt } = await this.knowledge.memberStats({
+      conversationId: userMd5,
+      startTime,
+      endTime
+    })
+    let freshness = judgeFreshness(result?.indexLatestAt ?? null, sourceLatestAt)
+
+    // 3) 索引没追平 → 有界追赶一次再判。追不上就如实标 stale，
+    //    **不允许**把落后索引算出来的「未发言」包装成完整结论。
+    if (freshness !== 'fresh') {
+      this.knowledge.requestCatchUp(CATCH_UP_MIN_INTERVAL_MS)
+      const settled = await this.knowledge.waitForIndexingComplete(FRESHNESS_WAIT_BUDGET_MS)
+      if (settled) {
+        const retried = await this.knowledge.memberStats({
+          conversationId: userMd5,
+          startTime,
+          endTime
+        })
+        result = retried.result
+        sourceLatestAt = retried.sourceLatestAt
+        freshness = judgeFreshness(result?.indexLatestAt ?? null, sourceLatestAt)
+      }
+    }
+
+    // 4) 差集：当前成员 × 窗口内发言
+    const statBySender = new Map<string, { messageCount: number; lastMessageTime: number }>()
+    for (const row of result?.senders ?? []) {
+      if (row.senderId) statBySender.set(row.senderId, row)
+    }
+
+    const knownSenderIds = new Set<string>()
+    const activeMembers: GroupMemberStatsActiveMember[] = []
+    const silentMembers: GroupMemberStatsSilentMember[] = []
+
+    for (const member of members) {
+      const wxid = String(member.wxid || '').trim()
+      if (!wxid) continue
+      knownSenderIds.add(wxid)
+      // 与 Knowledge 的 Join key 必须是 wxid，不是昵称 —— 昵称会变。
+      // 显示名沿用项目既有优先级（nickname 已内含 `wechatNickname || groupNickname || username`）。
+      const displayName = member.nickname || member.remark || wxid
+      const groupNickname = member.groupNickname || ''
+      const stat = statBySender.get(wxid)
+      if (stat && stat.messageCount > 0) {
+        activeMembers.push({
+          senderId: wxid,
+          displayName,
+          groupNickname,
+          messageCount: stat.messageCount,
+          lastMessageTime: stat.lastMessageTime
+        })
+      } else {
+        silentMembers.push({ senderId: wxid, displayName, groupNickname })
+      }
+    }
+
+    // 窗口内有发言、但已不在当前成员名单里的人（退群者）。
+    // 默认**不**进活跃榜：那份名单读作「当前群成员」，混入已退群的人会误导。
+    let formerSenderCount = 0
+    for (const senderId of statBySender.keys()) {
+      if (!knownSenderIds.has(senderId)) formerSenderCount += 1
+    }
+
+    activeMembers.sort(
+      (a, b) => b.messageCount - a.messageCount || b.lastMessageTime - a.lastMessageTime
+    )
+    silentMembers.sort((a, b) => a.displayName.localeCompare(b.displayName, 'zh-Hans-CN'))
+
+    if (freshness !== 'fresh') limitations.push(GROUP_STATS_STALE_LIMITATION)
+    if (formerSenderCount > 0) {
+      limitations.push(
+        `另有 ${formerSenderCount} 位窗口内发言者已不在当前群成员名单中，未计入活跃榜。`
+      )
+    }
+
+    return this.build({
+      userMd5,
+      startTime,
+      endTime,
+      freshness,
+      limitations,
+      members,
+      activeMembers,
+      silentMembers,
+      unattributedMessages: result?.unattributedMessages ?? 0,
+      excludedSystemMessages: result?.excludedSystemMessages ?? 0,
+      firstMessageTime: result?.earliestMessageTime ?? null
+    })
+  }
+
+  private build(input: {
+    userMd5: string
+    startTime: number
+    endTime: number
+    freshness: GroupStatsFreshness
+    limitations: string[]
+    members: Array<{ wxid: string }>
+    activeMembers: GroupMemberStatsActiveMember[]
+    silentMembers: GroupMemberStatsSilentMember[]
+    unattributedMessages: number
+    excludedSystemMessages: number
+    firstMessageTime: number | null
+  }): GroupMemberStatsResult {
+    return {
+      conversationId: input.userMd5,
+      startTime: input.startTime,
+      endTime: input.endTime,
+      freshness: input.freshness,
+      // 只有索引确实追平，才允许调用方把「未发言」当作完整结论。
+      complete: input.freshness === 'fresh',
+      memberCount: input.members.length,
+      activeMemberCount: input.activeMembers.length,
+      silentMemberCount: input.silentMembers.length,
+      activeMembers: input.activeMembers,
+      silentMembers: input.silentMembers,
+      unattributedMessages: input.unattributedMessages,
+      excludedSystemMessages: input.excludedSystemMessages,
+      firstMessageTime: input.firstMessageTime,
+      limitations: input.limitations
+    }
+  }
+}
+
+/**
+ * `isKnowledgeFresh` 返回 `boolean | null`（任一侧口径缺失即 null）。
+ * 三态都要能表达：`null` 绝不能当成「新鲜」。
+ */
+function judgeFreshness(
+  indexLatestAt: number | null,
+  sourceLatestAt: number | null
+): GroupStatsFreshness {
+  const fresh = isKnowledgeFresh({ indexLatestAt, sourceLatestAt })
+  if (fresh === true) return 'fresh'
+  if (fresh === false) return 'stale'
+  return 'unknown'
+}
@@ -0,0 +1,116 @@
+import {
+  normalizeLeaveNotificationRoomIds,
+  normalizeLeaveNotificationTemplate,
+  type LeaveNotificationNotifyScope,
+  type LeaveNotificationTarget
+} from '../../shared/automation'
+
+/**
+ * 旧「退群监控 → 通知群聊 / 模板」配置 → 新「自动化 → 退群通知」规则的**纯映射器**。
+ *
+ * 单独成文件是为了能直接单测：这是整轮迁移里**唯一会改变用户既有行为**的地方，
+ * 不允许只靠"跑一遍看看"来验证。
+ *
+ * ## 旧语义（实测自 `group-exit-monitor-service.ts`）
+ *
+ * 旧版是**两层**配置，缺一不可：
+ * 1. **管理群聊**（`monitoredRoomIds`）= 监测哪些群有人退出；
+ * 2. **通知群聊**（`notificationRoomIds`）= 上面这批群里，**哪些要真的发通知**
+ *    —— 即对第 1 层的二次勾选。
+ *
+ * `notificationRoomIds` **不是**"统一通知接收目标列表"，而是
+ * **"哪些被监控群要在自己群里通报成员退出"**（per-group 多值）：
+ * - `applyCurrentMemberships()` 只在 `notificationRoomIds.has(roomId)` 时通知；
+ * - `notifyGroup()` 的收件人**恒等于事件所在群**。
+ *
+ * ## 新语义
+ *
+ * 拆成**两个正交维度**，恰好一一对应旧版那两层：
+ * - `target`      ← 旧版第 2 层里"发回本群"这个事实 ⇒ 恒为 `source_chat`；
+ * - `notifyScope` / `notifyRoomIds` ← 旧版第 2 层是"哪些群"，逐字保留。
+ *
+ * ⇒ **迁移现在是无损的**。早期版本把目标压成单值、丢掉了第 2 层的"哪些群"，
+ * 于是部分勾选只能判成 `needs_review` 并强制停用；恢复 `notifyScope` 之后
+ * 这个特例整体消失，用户不会再有"规则被无辜停用"的体验。
+ */
+
+export interface LegacyLeaveNotificationState {
+  monitoredRoomIds: string[]
+  notificationRoomIds: string[]
+  notificationTemplate?: unknown
+}
+
+export type LeaveNotificationMigrationOutcome =
+  /** 旧配置从不通知任何人 ⇒ 迁移成"关闭"，行为完全一致。 */
+  | 'lossless_off'
+  /** 旧配置在**每个**被监控群里通报 ⇒ `notifyScope: 'all'`，行为完全一致。 */
+  | 'lossless_all_groups'
+  /** 旧配置只在**部分**群里通报 ⇒ `notifyScope: 'selected'` + 原样保留那份勾选。 */
+  | 'lossless_selected_groups'
+
+export interface LeaveNotificationMigrationPlan {
+  outcome: LeaveNotificationMigrationOutcome
+  enabled: boolean
+  target: LeaveNotificationTarget
+  notifyScope: LeaveNotificationNotifyScope
+  notifyRoomIds: string[]
+  targetNeedsReview: boolean
+  template: string
+}
+
+function uniqueRoomIds(values: unknown): string[] {
+  return normalizeLeaveNotificationRoomIds(values)
+}
+
+export function planLeaveNotificationMigration(
+  legacy: LegacyLeaveNotificationState
+): LeaveNotificationMigrationPlan {
+  const monitored = uniqueRoomIds(legacy.monitoredRoomIds)
+  const monitoredSet = new Set(monitored)
+  // 再夹一次：防止脏状态文件把范围外的群带进来。
+  const notifications = uniqueRoomIds(legacy.notificationRoomIds).filter((roomId) =>
+    monitoredSet.has(roomId)
+  )
+  // 模板永远是无损迁移的那一项：它就是用户写的那段文字。
+  const template = normalizeLeaveNotificationTemplate(legacy.notificationTemplate)
+  // 旧版收件人恒为事件所在群 ⇒ 无论勾了几个群，目标都是 source_chat。
+  const target: LeaveNotificationTarget = { type: 'source_chat' }
+
+  if (!notifications.length) {
+    // 旧语义是"从不通知" ⇒ 关掉，避免迁移后突然开始发消息。
+    // 同时把范围写成 `selected` + 空集：配置本身也精确表达"一个都不通知"。
+    return {
+      outcome: 'lossless_off',
+      enabled: false,
+      target,
+      notifyScope: 'selected',
+      notifyRoomIds: [],
+      targetNeedsReview: false,
+      template
+    }
+  }
+
+  if (monitored.length > 0 && notifications.length === monitored.length) {
+    // 每个被监控群都在自己群里通报 ⇒ 等价于"全部已监控群聊"。
+    return {
+      outcome: 'lossless_all_groups',
+      enabled: true,
+      target,
+      notifyScope: 'all',
+      notifyRoomIds: [],
+      targetNeedsReview: false,
+      template
+    }
+  }
+
+  // 部分勾选：**无损**保留那份子集，不再降级成 needs_review。
+  return {
+    outcome: 'lossless_selected_groups',
+    enabled: true,
+    target,
+    notifyScope: 'selected',
+    notifyRoomIds: notifications,
+    targetNeedsReview: false,
+    template
+  }
+}
@@ -0,0 +1,91 @@
+import {
+  LEAVE_NOTIFICATION_TARGET_OPTIONS,
+  leaveNotificationTargetLabel,
+  type LeaveNotificationConfig
+} from '../../shared/automation'
+import type { GroupMemberExitedEvent } from '../../shared/group-exit-event'
+import {
+  automationContactDisplayName,
+  filterSendableFriendContacts as filterSendable,
+  isSendableFriendContact as isSendable,
+  resolveAutomationTarget,
+  type AutomationTargetContact,
+  type AutomationTargetResolution,
+  type ResolvedAutomationTarget
+} from './automation-wechat-target'
+
+/**
+ * 退群通知的**目标解析**（薄适配层）。
+ *
+ * 真正的判定在 `automation-wechat-target.ts` —— 那里才是**唯一实现**，
+ * 定时日报与退群通知共用它。这里只有两件事：
+ * 1. 把 `LeaveNotificationConfig` + 退群事件摊平成中性输入；
+ * 2. 提供退群通知专属文案，保持既有调用方 / 测试不变。
+ *
+ * ⚠️ 这里**不允许**再出现一份 switch / 判定逻辑 —— 两套并存迟早分叉。
+ */
+
+/** 解析所需的最小联系人结构（`FormattedContact` 结构上可赋值给它）。 */
+export type LeaveNotificationCandidateContact = AutomationTargetContact
+
+export type ResolvedLeaveNotificationTarget = ResolvedAutomationTarget
+
+export type LeaveNotificationTargetResolution = AutomationTargetResolution
+
+/** 退群通知的目标文案（唯一一份）。 */
+export const LEAVE_NOTIFICATION_TARGET_MESSAGES = {
+  options: LEAVE_NOTIFICATION_TARGET_OPTIONS,
+  invalidTarget: '退群通知的发送目标无效，请重新选择',
+  sourceMissing: '无法确定发生退群的群聊，本次通知未发送',
+  sourceRecipientName: '当前群聊',
+  sourceDisplayName: '发生退群事件的群聊',
+  selfMissing: '无法确定当前登录的微信账号，本次通知未发送',
+  contactNotChosen: '还没有选择通知联系人，请重新选择',
+  contactUnavailable: '通知联系人已不存在或当前无法发送，请重新选择'
+} as const
+
+/** 联系人显示名：备注 → 微信昵称 → 会话昵称 → 兜底称呼。**永不回落到 wxid**。 */
+export function leaveNotificationContactDisplayName(contact: AutomationTargetContact): string {
+  return automationContactDisplayName(contact)
+}
+
+/** 这个联系人能不能作为「指定好友」的发送目标（与定时日报共用同一条判定）。 */
+export function isSendableFriendContact(
+  contact: AutomationTargetContact,
+  selfWxid?: string
+): boolean {
+  return isSendable(contact, selfWxid)
+}
+
+/** 供 UI 用的「可选好友」过滤（与运行时同一套判定）。 */
+export function filterSendableFriendContacts<T extends AutomationTargetContact>(
+  contacts: T[],
+  selfWxid?: string
+): T[] {
+  return filterSendable(contacts, selfWxid)
+}
+
+export function resolveLeaveNotificationTarget(input: {
+  config: LeaveNotificationConfig | undefined
+  event: GroupMemberExitedEvent
+  contacts: LeaveNotificationCandidateContact[]
+  /** 当前登录账号的 wxid；拿不到时传空串。 */
+  selfWxid?: string
+}): LeaveNotificationTargetResolution {
+  const config = input.config
+  if (!config) return { ok: false, error: '退群通知尚未配置发送目标' }
+  return resolveAutomationTarget(
+    {
+      targetType: config.target?.type as never,
+      sourceConversationId: String(input.event.conversationId || '').trim(),
+      sourceDisplayName: String(input.event.groupName || '').trim(),
+      ...(config.target?.contactId ? { contactId: config.target.contactId } : {}),
+      contacts: input.contacts,
+      ...(input.selfWxid ? { selfWxid: input.selfWxid } : {})
+    },
+    LEAVE_NOTIFICATION_TARGET_MESSAGES
+  )
+}
+
+/** 目标类型 → 界面短标签（保持既有导出面）。 */
+export { leaveNotificationTargetLabel }
@@ -0,0 +1,967 @@
+import {
+  BUILTIN_DAILY_REPORT_RULE_ID,
+  BUILTIN_LEAVE_NOTIFICATION_RULE_ID,
+  calculateNextRunAt,
+  type AutomationExecution,
+  type AutomationRule,
+  type AutomationRuleDraft,
+  type AutomationRuleType
+} from '../../shared/automation'
+import { resolveContact } from './contact-resolution-service'
+import { validateAutomationDraftShape } from '../../shared/agent-api/automation-validation'
+import type {
+  AgentAutomationExecution,
+  AgentAutomationValidationIssue,
+  AgentAutomationValidationResult,
+  ApplicationCapabilities
+} from '../../shared/agent-api/contracts'
+import type {
+  AgentGroupExitMonitorEvent,
+  AgentGroupExitMonitorState
+} from '../../shared/agent-api/group-exit-monitor'
+import type { AgentGroupMemberStats } from '../../shared/agent-api/group-stats'
+import type { GroupExitMonitorEvent } from '../../shared/group-exit-monitor'
+import type { GroupMemberStatsQuery, GroupMemberStatsResult } from '../../shared/group-stats'
+import type { Contact } from '../../shared/types'
+import type { PersonalWechatSendCapability } from '../../shared/personal-wechat'
+import type { AutomationExecutionLogService } from './automation-execution-log-service'
+import type { AutomationRuleStore } from './automation-rule-store'
+
+type AutomationRuleStoreApi = Pick<
+  AutomationRuleStore,
+  'listRules' | 'getRule' | 'createRule' | 'updateRule' | 'deleteRule' | 'setRuleEnabled'
+>
+type AutomationExecutionLogApi = Pick<AutomationExecutionLogService, 'list'>
+type RuleScope = 'any' | 'person' | 'group'
+
+type GroupExitMonitorStateSource = {
+  enabled: boolean
+  running: boolean
+  monitoredRoomIds?: string[]
+  nativeMonitorActive?: boolean
+  monitoredGroupCount?: number
+  monitorSelectionConfigured?: boolean
+  lastCheckedAt?: number
+  lastReadAt?: number
+  unreadCount?: number
+  totalEventCount?: number
+  events?: GroupExitMonitorEvent[]
+}
+
+type GroupExitMonitorConfiguration = {
+  enabled?: boolean
+  monitoredRoomIds?: string[]
+}
+
+type GroupExitMonitorEventQuery = {
+  roomId?: string
+  sinceMs?: number
+  untilMs?: number
+  limit?: number
+}
+
+export class LocalAgentApiError extends Error {
+  constructor(
+    readonly status: number,
+    readonly code: string,
+    message: string,
+    readonly details?: unknown
+  ) {
+    super(message)
+    this.name = 'LocalAgentApiError'
+  }
+}
+
+export interface LocalAgentApiDependencies {
+  automationRuleStore: AutomationRuleStoreApi
+  automationExecutionLogService: AutomationExecutionLogApi
+  listContacts: () => Promise<Contact[]>
+  isDatabaseReady: () => boolean
+  getVersion: () => string
+  getPersonalWechatCapability: () => Promise<PersonalWechatSendCapability>
+  getAgentHubStatus: () => { connector: string }
+  getGroupExitMonitorState: () => GroupExitMonitorStateSource
+  /** Preferred atomic monitor configuration operation. */
+  configureGroupExitMonitor?: (
+    configuration: GroupExitMonitorConfiguration
+  ) => Promise<GroupExitMonitorStateSource> | GroupExitMonitorStateSource
+  setGroupExitMonitorRoomIds?: (roomIds: string[]) => Promise<GroupExitMonitorStateSource>
+  setGroupExitMonitorEnabled?: (enabled: boolean) => Promise<GroupExitMonitorStateSource>
+  listGroupExitMonitorEvents?: (query: GroupExitMonitorEventQuery) => GroupExitMonitorEvent[]
+  getGroupMemberStats?: (query: GroupMemberStatsQuery) => Promise<GroupMemberStatsResult>
+}
+
+const RULE_TYPES: AutomationRuleType[] = ['daily_report', 'scheduled_report', 'leave_notification']
+const EXECUTION_STATUSES: AutomationExecution['status'][] = ['running', 'success', 'failed']
+const UPDATE_FIELDS = new Set([
+  'name',
+  'trigger',
+  'scope',
+  'conditions',
+  'actions',
+  'cooldownSeconds',
+  'replyDelaySeconds',
+  'leaveNotification',
+  'scheduledReport'
+])
+
+function record(value: unknown): Record<string, unknown> | undefined {
+  return value !== null && typeof value === 'object' && !Array.isArray(value)
+    ? (value as Record<string, unknown>)
+    : undefined
+}
+
+function toDraft(rule: AutomationRule): AutomationRuleDraft {
+  return {
+    name: rule.name,
+    enabled: false,
+    ruleType: rule.ruleType,
+    trigger: rule.trigger,
+    scope: rule.scope,
+    conditions: structuredClone(rule.conditions),
+    actions: structuredClone(rule.actions),
+    cooldownSeconds: rule.cooldownSeconds,
+    replyDelaySeconds: rule.replyDelaySeconds,
+    ...(rule.leaveNotification
+      ? { leaveNotification: structuredClone(rule.leaveNotification) }
+      : {}),
+    ...(rule.scheduledReport ? { scheduledReport: structuredClone(rule.scheduledReport) } : {})
+  }
+}
+
+function deepMergeDraft(
+  current: AutomationRuleDraft,
+  patch: Record<string, unknown>
+): Record<string, unknown> {
+  const next: Record<string, unknown> = { ...current, ...patch, enabled: false }
+  if (patch.conditions !== undefined) {
+    next.conditions = { ...current.conditions, ...record(patch.conditions) }
+  }
+  if (patch.leaveNotification !== undefined && current.leaveNotification) {
+    const leavePatch = record(patch.leaveNotification) || {}
+    next.leaveNotification = {
+      ...current.leaveNotification,
+      ...leavePatch,
+      ...(leavePatch.target !== undefined ? { target: leavePatch.target } : {})
+    }
+    if (record(leavePatch.target)?.type !== undefined) {
+      delete (next.leaveNotification as Record<string, unknown>).targetNeedsReview
+    }
+  }
+  if (patch.scheduledReport !== undefined && current.scheduledReport) {
+    const schedulePatch = record(patch.scheduledReport) || {}
+    const currentConfig = current.scheduledReport
+    next.scheduledReport = {
+      ...currentConfig,
+      ...schedulePatch,
+      schedule: { ...currentConfig.schedule, ...record(schedulePatch.schedule) },
+      report: { ...currentConfig.report, ...record(schedulePatch.report) },
+      ...(schedulePatch.target !== undefined ? { target: schedulePatch.target } : {})
+    }
+    if (record(schedulePatch.target)?.type !== undefined) {
+      delete (next.scheduledReport as Record<string, unknown>).targetNeedsReview
+    }
+  }
+  return next
+}
+
+function stableId(contact: Contact): string {
+  return contact.type === 'user' ? contact.wxid || contact.m_nsUsrName : contact.m_nsUsrName
+}
+
+function toIsoTimestamp(value: number | undefined): string | null {
+  if (!Number.isFinite(value) || Number(value) <= 0) return null
+  const date = new Date(Number(value))
+  return Number.isNaN(date.getTime()) ? null : date.toISOString()
+}
+
+export class LocalAgentApiService {
+  private readonly deps: LocalAgentApiDependencies
+
+  constructor(dependencies: LocalAgentApiDependencies) {
+    this.deps = dependencies
+  }
+
+  async getCapabilities(): Promise<ApplicationCapabilities> {
+    const databaseReady = this.deps.isDatabaseReady()
+    const hub = this.deps.getAgentHubStatus()
+    const monitor = this.deps.getGroupExitMonitorState()
+    let personal: Pick<
+      PersonalWechatSendCapability,
+      'supported' | 'ready' | 'status' | 'capabilities'
+    >
+    try {
+      personal = await this.deps.getPersonalWechatCapability()
+    } catch {
+      personal = {
+        supported: true,
+        ready: false,
+        status: 'error',
+        capabilities: { text: false, image: false, voice: false }
+      }
+    }
+
+    return {
+      version: this.deps.getVersion(),
+      apiVersion: 'v1',
+      database: { ready: databaseReady },
+      query: {
+        supported: true,
+        available: databaseReady,
+        ...(!databaseReady ? { reason: 'database_not_ready' } : {})
+      },
+      automations: {
+        supported: true,
+        available: true,
+        ruleTypes: [...RULE_TYPES],
+        operations: [
+          'list',
+          'get',
+          'create',
+          'update',
+          'validate',
+          'enable',
+          'disable',
+          'delete',
+          'executions'
+        ]
+      },
+      groupExitMonitor: {
+        supported: true,
+        available: databaseReady && monitor.running,
+        operations: [
+          'read_state',
+          'configure_scope',
+          'enable',
+          'disable',
+          'list_events'
+        ],
+        ...(!databaseReady
+          ? { reason: 'database_not_ready' }
+          : !monitor.enabled
+            ? { reason: 'disabled' }
+            : !monitor.running
+              ? { reason: 'not_running' }
+              : {})
+      },
+      groupStats: {
+        supported: true,
+        available: databaseReady,
+        operations: ['member_stats'],
+        ...(!databaseReady ? { reason: 'database_not_ready' } : {})
+      },
+      wechat: {
+        personal: {
+          supported: personal.supported,
+          available: personal.ready,
+          status: personal.status,
+          content: { ...personal.capabilities },
+          ...(!personal.ready ? { reason: personal.status } : {})
+        },
+        ilink: {
+          supported: true,
+          available: hub.connector === 'online',
+          status: hub.connector,
+          ...(hub.connector !== 'online' ? { reason: 'connector_not_online' } : {})
+        }
+      }
+    }
+  }
+
+  getGroupExitMonitorState(): AgentGroupExitMonitorState {
+    const state = this.deps.getGroupExitMonitorState()
+    return {
+      enabled: state.enabled === true,
+      running: state.running === true,
+      nativeMonitorActive: state.nativeMonitorActive === true,
+      monitoredConversationIds: [...(state.monitoredRoomIds || [])],
+      monitoredGroupCount: Number.isFinite(state.monitoredGroupCount)
+        ? Number(state.monitoredGroupCount)
+        : state.monitoredRoomIds?.length || 0,
+      monitorSelectionConfigured: state.monitorSelectionConfigured !== false,
+      lastCheckedAt: toIsoTimestamp(state.lastCheckedAt),
+      lastReadAt: toIsoTimestamp(state.lastReadAt),
+      eventCount: Number.isFinite(state.totalEventCount)
+        ? Number(state.totalEventCount)
+        : state.events?.length || 0,
+      unreadCount: Number.isFinite(state.unreadCount) ? Number(state.unreadCount) : 0
+    }
+  }
+
+  async updateGroupExitMonitor(input: unknown): Promise<AgentGroupExitMonitorState> {
+    const patch = record(input)
+    if (!patch) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '请求体必须是 JSON 对象')
+    }
+    const keys = Object.keys(patch)
+    const allowed = new Set(['enabled', 'monitoredConversationIds'])
+    const unknown = keys.find((key) => !allowed.has(key))
+    if (unknown) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `不支持字段：${unknown}`)
+    }
+    if (keys.length === 0) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '至少提供 enabled 或 monitoredConversationIds')
+    }
+
+    const enabled = patch.enabled
+    if (enabled !== undefined && typeof enabled !== 'boolean') {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'enabled 必须是布尔值')
+    }
+
+    let monitoredRoomIds: string[] | undefined
+    if (patch.monitoredConversationIds !== undefined) {
+      if (!Array.isArray(patch.monitoredConversationIds)) {
+        throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'monitoredConversationIds 必须是数组')
+      }
+      monitoredRoomIds = []
+      const seen = new Set<string>()
+      for (const value of patch.monitoredConversationIds) {
+        if (typeof value !== 'string' || !value.trim()) {
+          throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '监控群 ID 不能为空')
+        }
+        const roomId = value.trim()
+        if (seen.has(roomId)) {
+          throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `监控群 ID 重复：${roomId}`)
+        }
+        if (!roomId.endsWith('@chatroom') || roomId.includes('/') || roomId.includes('\\')) {
+          throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `无效的群会话 ID：${roomId}`)
+        }
+        seen.add(roomId)
+        monitoredRoomIds.push(roomId)
+      }
+    }
+
+    if (!this.deps.isDatabaseReady()) {
+      throw new LocalAgentApiError(409, 'DATABASE_NOT_READY', '数据库未就绪，无法修改退群监控配置')
+    }
+    if (monitoredRoomIds !== undefined) {
+      const contacts = await this.deps.listContacts()
+      const groups = new Set(
+        contacts
+          .filter((contact) => contact.type === 'group')
+          .map((contact) => contact.m_nsUsrName)
+      )
+      const missing = monitoredRoomIds.find((roomId) => !groups.has(roomId))
+      if (missing) {
+        throw new LocalAgentApiError(
+          422,
+          'VALIDATION_FAILED',
+          `监控群不存在或不是群联系人：${missing}`,
+          [{ path: 'monitoredConversationIds', code: 'group_not_found', message: missing }]
+        )
+      }
+    }
+
+    const configuration: GroupExitMonitorConfiguration = {
+      ...(enabled !== undefined ? { enabled } : {}),
+      ...(monitoredRoomIds !== undefined ? { monitoredRoomIds } : {})
+    }
+    let nextState: GroupExitMonitorStateSource | undefined
+    if (this.deps.configureGroupExitMonitor) {
+      nextState = await this.deps.configureGroupExitMonitor(configuration)
+    } else {
+      // Test adapters and older embedders may only expose the two primitive operations.
+      // All validation is complete before this fallback starts mutating state.
+      const previous = this.deps.getGroupExitMonitorState()
+      try {
+        if (monitoredRoomIds !== undefined) {
+          if (!this.deps.setGroupExitMonitorRoomIds) {
+            throw new LocalAgentApiError(503, 'CAPABILITY_UNAVAILABLE', '退群监控配置能力尚未就绪')
+          }
+          nextState = await this.deps.setGroupExitMonitorRoomIds(monitoredRoomIds)
+        }
+        if (enabled !== undefined) {
+          if (!this.deps.setGroupExitMonitorEnabled) {
+            throw new LocalAgentApiError(503, 'CAPABILITY_UNAVAILABLE', '退群监控配置能力尚未就绪')
+          }
+          nextState = await this.deps.setGroupExitMonitorEnabled(enabled)
+        }
+      } catch (error) {
+        // Best-effort rollback for legacy adapters. Production uses the atomic operation above.
+        try {
+          if (monitoredRoomIds !== undefined && this.deps.setGroupExitMonitorRoomIds) {
+            await this.deps.setGroupExitMonitorRoomIds(previous.monitoredRoomIds || [])
+          }
+          if (enabled !== undefined && this.deps.setGroupExitMonitorEnabled) {
+            await this.deps.setGroupExitMonitorEnabled(previous.enabled)
+          }
+        } catch {
+          // Preserve the original error; an adapter that cannot roll back is non-atomic by definition.
+        }
+        throw error
+      }
+    }
+    return this.toGroupExitMonitorState(nextState || this.deps.getGroupExitMonitorState())
+  }
+
+  listGroupExitMonitorEvents(query: URLSearchParams): {
+    count: number
+    events: AgentGroupExitMonitorEvent[]
+  } {
+    if (!this.deps.listGroupExitMonitorEvents) {
+      throw new LocalAgentApiError(503, 'CAPABILITY_UNAVAILABLE', '退群事件查询能力尚未就绪')
+    }
+    const rawConversationId = query.get('conversationId')
+    let roomId: string | undefined
+    if (rawConversationId !== null) {
+      roomId = rawConversationId.trim()
+      if (!roomId || !roomId.endsWith('@chatroom') || roomId.includes('/') || roomId.includes('\\')) {
+        throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'conversationId 必须是有效的群会话 ID')
+      }
+    }
+    const since = this.parseDateFilter(query.get('since'), 'since')
+    const until = this.parseDateFilter(query.get('until'), 'until')
+    if (since !== undefined && until !== undefined && since > until) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'since 不能晚于 until')
+    }
+    const rawLimit = query.get('limit')
+    const limit = rawLimit === null ? 50 : Number(rawLimit)
+    if (!Number.isInteger(limit) || limit < 1 || limit > 200) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'limit 必须是 1 到 200 之间的整数')
+    }
+    const events = this.deps
+      .listGroupExitMonitorEvents({ roomId, sinceMs: since, untilMs: until, limit })
+      .slice()
+      .sort((left, right) => left.detectedAt - right.detectedAt)
+      .map((event) => this.toAgentGroupExitEvent(event))
+    return { count: events.length, events }
+  }
+
+  async getGroupMemberStats(
+    conversationId: string,
+    query: URLSearchParams
+  ): Promise<AgentGroupMemberStats> {
+    if (!this.deps.isDatabaseReady()) {
+      throw new LocalAgentApiError(409, 'DATABASE_NOT_READY', '数据库未就绪，无法查询群员统计')
+    }
+    const normalizedConversationId = conversationId.trim()
+    if (!normalizedConversationId || normalizedConversationId.includes('/') || normalizedConversationId.includes('\\')) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'conversationId 格式无效')
+    }
+    const start = this.parseDateFilter(query.get('start'), 'start')
+    const end = this.parseDateFilter(query.get('end'), 'end')
+    if (start === undefined || end === undefined) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'start 和 end 必须同时提供')
+    }
+    if (start > end) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'start 不能晚于 end')
+    }
+
+    const contacts = await this.deps.listContacts()
+    const matches = contacts.filter((contact) => contact.m_nsUsrName === normalizedConversationId)
+    if (matches.length === 0) {
+      throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到指定会话')
+    }
+    if (matches.length > 1) {
+      throw new LocalAgentApiError(409, 'AMBIGUOUS_CONTACT', 'conversationId 匹配到多个会话')
+    }
+    const contact = matches[0]
+    if (contact.type !== 'group') {
+      throw new LocalAgentApiError(422, 'NOT_GROUP_CONVERSATION', 'conversationId 不是群会话')
+    }
+    if (!contact.m_nsUsrName.endsWith('@chatroom')) {
+      throw new LocalAgentApiError(422, 'NOT_GROUP_CONVERSATION', 'conversationId 不是有效的群会话 ID')
+    }
+    if (!this.deps.getGroupMemberStats) {
+      throw new LocalAgentApiError(503, 'CAPABILITY_UNAVAILABLE', '群员统计能力尚未就绪')
+    }
+
+    const result = await this.deps.getGroupMemberStats({
+      userMd5: contact.md5,
+      startTime: start,
+      endTime: end
+    })
+    return this.toAgentGroupMemberStats(contact, result)
+  }
+
+  async listAutomations(filter: {
+    type?: string | null
+    enabled?: string | null
+  }): Promise<unknown[]> {
+    if (filter.type && !RULE_TYPES.includes(filter.type as AutomationRuleType)) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'type 不受支持')
+    }
+    let enabled: boolean | undefined
+    if (filter.enabled !== undefined && filter.enabled !== null) {
+      if (filter.enabled !== 'true' && filter.enabled !== 'false') {
+        throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'enabled 必须是 true 或 false')
+      }
+      enabled = filter.enabled === 'true'
+    }
+    const rules = this.deps.automationRuleStore
+      .listRules()
+      .filter(
+        (rule) =>
+          (!filter.type || rule.ruleType === filter.type) &&
+          (enabled === undefined || rule.enabled === enabled)
+      )
+    return this.toApiRules(rules)
+  }
+
+  async getAutomation(id: string): Promise<unknown> {
+    const rule = this.deps.automationRuleStore.getRule(id)
+    if (!rule) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    return (await this.toApiRules([rule]))[0]
+  }
+
+  async validateAutomation(input: unknown): Promise<AgentAutomationValidationResult> {
+    const structural = validateAutomationDraftShape(input)
+    if (!structural.valid || !structural.normalized) {
+      return { valid: false, issues: structural.issues }
+    }
+
+    const issues = [...structural.issues]
+    const draft = structuredClone(structural.normalized)
+    const references = await this.resolveDraftReferences(draft, issues)
+    if (issues.length) return { valid: false, issues }
+
+    const config = draft.scheduledReport
+    const nextRunAt = config ? calculateNextRunAt(config.schedule.time) : null
+    const effects = this.describeEffects(draft)
+    return {
+      valid: true,
+      issues: [],
+      normalized: draft,
+      effects,
+      nextRunAt,
+      capabilities: {
+        databaseReady: this.deps.isDatabaseReady(),
+        referencedConversationsResolved: references
+      }
+    }
+  }
+
+  async createAutomation(input: unknown): Promise<unknown> {
+    const raw = record(input)
+    if (raw?.ruleType === 'leave_notification') {
+      throw new LocalAgentApiError(409, 'SINGLETON_RULE', '退群通知是系统单例规则，请修改现有规则')
+    }
+    const validation = await this.validateAutomation(input)
+    if (!validation.valid || !validation.normalized) {
+      throw new LocalAgentApiError(
+        422,
+        'VALIDATION_FAILED',
+        '自动化规则校验失败',
+        validation.issues
+      )
+    }
+    const contacts = await this.deps.listContacts()
+    const storedDraft = this.toStoreDraft(validation.normalized as AutomationRuleDraft, contacts)
+    storedDraft.enabled = false
+    const created = this.deps.automationRuleStore.createRule(storedDraft)
+    return (await this.toApiRules([created]))[0]
+  }
+
+  async updateAutomation(id: string, patchInput: unknown): Promise<unknown> {
+    const current = this.deps.automationRuleStore.getRule(id)
+    if (!current) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    const patch = record(patchInput)
+    if (!patch) throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', '请求体必须是 JSON 对象')
+    for (const key of Object.keys(patch)) {
+      if (key === 'ruleType') {
+        throw new LocalAgentApiError(400, 'RULE_TYPE_IMMUTABLE', 'ruleType 不可修改')
+      }
+      if (key === 'id' || key === 'createdAt' || key === 'updatedAt') {
+        throw new LocalAgentApiError(400, 'IMMUTABLE_FIELD', `${key} 不允许由客户端修改`)
+      }
+      if (key === 'enabled') {
+        throw new LocalAgentApiError(400, 'USE_ENABLE_OPERATION', '请使用 enable 或 disable 操作')
+      }
+      if (!UPDATE_FIELDS.has(key)) {
+        throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `不支持字段：${key}`)
+      }
+    }
+    for (const configKey of ['leaveNotification', 'scheduledReport']) {
+      if (Object.hasOwn(record(patch[configKey]) || {}, 'targetNeedsReview')) {
+        throw new LocalAgentApiError(
+          400,
+          'IMMUTABLE_FIELD',
+          `${configKey}.targetNeedsReview 只能通过更新 target 清除`
+        )
+      }
+    }
+
+    const apiCurrent = (await this.toApiRules([current]))[0] as AutomationRule & {
+      requiresReview?: boolean
+    }
+    const candidate = deepMergeDraft(toDraft(apiCurrent), patch)
+    const validation = await this.validateAutomation(candidate)
+    if (!validation.valid || !validation.normalized) {
+      throw new LocalAgentApiError(
+        422,
+        'VALIDATION_FAILED',
+        '自动化规则校验失败',
+        validation.issues
+      )
+    }
+
+    const contacts = await this.deps.listContacts()
+    const storedDraft = this.toStoreDraft(validation.normalized as AutomationRuleDraft, contacts)
+    storedDraft.enabled = current.enabled
+    const updated = this.deps.automationRuleStore.updateRule(id, storedDraft)
+    if (!updated) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    return (await this.toApiRules([updated]))[0]
+  }
+
+  async setAutomationEnabled(id: string, enabled: boolean): Promise<unknown> {
+    const current = this.deps.automationRuleStore.getRule(id)
+    if (!current) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    if (enabled) {
+      if (!this.deps.isDatabaseReady()) {
+        throw new LocalAgentApiError(409, 'VALIDATION_FAILED', '数据库未就绪，无法启用自动化规则', [
+          { path: 'database', code: 'database_not_ready', message: 'TraceMemo 数据库未初始化' }
+        ])
+      }
+      const apiCurrent = (await this.toApiRules([current]))[0] as AutomationRule
+      const draft = toDraft(apiCurrent)
+      const validation = await this.validateAutomation(draft)
+      if (!validation.valid) {
+        throw new LocalAgentApiError(
+          409,
+          'VALIDATION_FAILED',
+          '自动化规则校验失败，未启用',
+          validation.issues
+        )
+      }
+    }
+    const updated = this.deps.automationRuleStore.setRuleEnabled(id, enabled)
+    if (!updated) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    return (await this.toApiRules([updated]))[0]
+  }
+
+  async deleteAutomation(id: string): Promise<{ deletedId: string }> {
+    if (id === BUILTIN_DAILY_REPORT_RULE_ID || id === BUILTIN_LEAVE_NOTIFICATION_RULE_ID) {
+      throw new LocalAgentApiError(409, 'PROTECTED_RULE', '系统内置自动化规则不能删除')
+    }
+    const rule = this.deps.automationRuleStore.getRule(id)
+    if (!rule) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    if (rule.ruleType === 'leave_notification') {
+      throw new LocalAgentApiError(409, 'PROTECTED_RULE', '退群通知是系统单例规则，不能删除')
+    }
+    const deleted = this.deps.automationRuleStore.deleteRule(id)
+    if (!deleted) throw new LocalAgentApiError(404, 'NOT_FOUND', '未找到自动化规则')
+    return { deletedId: id }
+  }
+
+  listExecutions(query: URLSearchParams): {
+    count: number
+    executions: AgentAutomationExecution[]
+  } {
+    const ruleId = query.get('ruleId') || undefined
+    const status = query.get('status') || undefined
+    if (status && !EXECUTION_STATUSES.includes(status as AutomationExecution['status'])) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'status 不受支持')
+    }
+    const since = this.parseDateFilter(query.get('since'), 'since')
+    const until = this.parseDateFilter(query.get('until'), 'until')
+    if (since !== undefined && until !== undefined && since > until) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'since 不能晚于 until')
+    }
+    const rawLimit = query.get('limit')
+    const limit = rawLimit === null ? 50 : Number(rawLimit)
+    if (!Number.isInteger(limit) || limit < 1 || limit > 200) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', 'limit 必须是 1 到 200 之间的整数')
+    }
+    const rulesById = new Map(
+      this.deps.automationRuleStore.listRules().map((rule) => [rule.id, rule])
+    )
+    const executions = this.deps.automationExecutionLogService
+      .list({ limit: 200 })
+      .filter((item) => !ruleId || item.ruleId === ruleId)
+      .filter((item) => !status || item.status === status)
+      .filter((item) => since === undefined || item.triggerTime >= since)
+      .filter((item) => until === undefined || item.triggerTime <= until)
+      .slice(0, limit)
+      .map((item) => this.toApiExecution(item, rulesById.get(item.ruleId)?.ruleType))
+    return { count: executions.length, executions }
+  }
+
+  private toGroupExitMonitorState(state: GroupExitMonitorStateSource): AgentGroupExitMonitorState {
+    const eventCount = Number.isFinite(state.totalEventCount)
+      ? Number(state.totalEventCount)
+      : state.events?.length || 0
+    return {
+      enabled: state.enabled === true,
+      running: state.running === true,
+      nativeMonitorActive: state.nativeMonitorActive === true,
+      monitoredConversationIds: [...(state.monitoredRoomIds || [])],
+      monitoredGroupCount: Number.isFinite(state.monitoredGroupCount)
+        ? Number(state.monitoredGroupCount)
+        : state.monitoredRoomIds?.length || 0,
+      monitorSelectionConfigured: state.monitorSelectionConfigured !== false,
+      lastCheckedAt: toIsoTimestamp(state.lastCheckedAt),
+      lastReadAt: toIsoTimestamp(state.lastReadAt),
+      eventCount,
+      unreadCount: Number.isFinite(state.unreadCount) ? Number(state.unreadCount) : 0
+    }
+  }
+
+  private toAgentGroupExitEvent(event: GroupExitMonitorEvent): AgentGroupExitMonitorEvent {
+    return {
+      eventId: event.id,
+      conversationId: event.roomId,
+      groupName: event.groupName,
+      memberId: event.memberWxid,
+      memberName: event.memberName,
+      wechatName: event.wechatName || '',
+      groupRemark: event.groupRemark || '',
+      contactRemark: event.contactRemark || '',
+      previousCount: event.previousCount,
+      currentCount: event.currentCount,
+      delta: event.delta,
+      message: event.message,
+      detectedAt: new Date(event.detectedAt).toISOString()
+    }
+  }
+
+  private toAgentGroupMemberStats(contact: Contact, result: GroupMemberStatsResult): AgentGroupMemberStats {
+    const conversationName =
+      contact.m_nsNickName || contact.remark || contact.wechatNickname || contact.m_nsUsrName
+    return {
+      conversation: { id: contact.m_nsUsrName, name: conversationName },
+      conversationId: contact.m_nsUsrName,
+      range: {
+        start: new Date(result.startTime).toISOString(),
+        end: new Date(result.endTime).toISOString()
+      },
+      memberCount: result.memberCount,
+      activeMemberCount: result.activeMemberCount,
+      silentMemberCount: result.silentMemberCount,
+      activeMembers: result.activeMembers.map((member) => ({
+        memberId: member.senderId,
+        displayName: member.displayName,
+        groupNickname: member.groupNickname,
+        messageCount: member.messageCount,
+        lastMessageAt: toIsoTimestamp(member.lastMessageTime)
+      })),
+      silentMembers: result.silentMembers.map((member) => ({
+        memberId: member.senderId,
+        displayName: member.displayName,
+        groupNickname: member.groupNickname
+      })),
+      freshness: result.freshness,
+      complete: result.complete,
+      limitations: [...result.limitations],
+      unattributedMessages: result.unattributedMessages,
+      excludedSystemMessages: result.excludedSystemMessages,
+      firstMessageAt:
+        toIsoTimestamp(result.firstMessageTime ?? undefined)
+    }
+  }
+
+  private parseDateFilter(value: string | null, name: string): number | undefined {
+    if (value === null) return undefined
+    if (!/T.*(?:Z|[+-]\d{2}:\d{2})$/.test(value)) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `${name} 必须是带时区的 ISO-8601 时间`)
+    }
+    const timestamp = Date.parse(value)
+    if (!Number.isFinite(timestamp)) {
+      throw new LocalAgentApiError(400, 'INVALID_ARGUMENT', `${name} 不是有效时间`)
+    }
+    return timestamp
+  }
+
+  private toApiExecution(
+    execution: AutomationExecution,
+    ruleType?: AutomationRuleType
+  ): AgentAutomationExecution {
+    return {
+      executionId: execution.executionId,
+      ruleId: execution.ruleId,
+      ruleName: execution.ruleName,
+      ...(ruleType ? { ruleType } : {}),
+      ...(execution.trigger ? { trigger: execution.trigger } : {}),
+      status: execution.status,
+      triggerTime: execution.triggerTime,
+      durationMs: execution.durationMs,
+      sourceDisplayName: execution.sourceDisplayName,
+      startedAt: new Date(execution.triggerTime).toISOString(),
+      finishedAt:
+        execution.status === 'running'
+          ? null
+          : new Date(execution.triggerTime + execution.durationMs).toISOString(),
+      ...(execution.errorSummary ? { error: execution.errorSummary } : {})
+    }
+  }
+
+  private async resolveDraftReferences(
+    draft: AutomationRuleDraft,
+    issues: AgentAutomationValidationIssue[]
+  ): Promise<boolean> {
+    const needsContacts =
+      (draft.ruleType === 'daily_report' && draft.conditions.conversationIds.length > 0) ||
+      draft.ruleType === 'scheduled_report' ||
+      (draft.ruleType === 'leave_notification' &&
+        (draft.leaveNotification?.target.type === 'contact' ||
+          (draft.leaveNotification?.notifyRoomIds.length ?? 0) > 0))
+    if (!needsContacts) return true
+    if (!this.deps.isDatabaseReady()) {
+      issues.push({
+        path: 'database',
+        code: 'database_not_ready',
+        message: 'TraceMemo 数据库未初始化，无法解析联系人或群聊 ID'
+      })
+      return false
+    }
+
+    const contacts = await this.deps.listContacts()
+    let resolved = true
+    const resolve = (query: string, scope: RuleScope, path: string): Contact | undefined => {
+      const direct = contacts.filter((contact) => {
+        const identifiers = [
+          contact.md5,
+          contact.m_nsUsrName,
+          contact.wxid,
+          contact.wechatId,
+          contact.alias
+        ]
+        return identifiers.some(
+          (identifier) => identifier?.toLocaleLowerCase() === query.toLocaleLowerCase()
+        )
+      })
+      const scopedDirect = direct.filter(
+        (contact) =>
+          scope === 'any' ||
+          (scope === 'person' ? contact.type === 'user' : contact.type === 'group')
+      )
+      if (scopedDirect.length === 1) return scopedDirect[0]
+
+      const result = resolveContact(query, contacts, scope)
+      if (result.matched && result.conversationId) {
+        return contacts.find((contact) => contact.md5 === result.conversationId)
+      }
+      resolved = false
+      issues.push({
+        path,
+        code: result.ambiguous ? 'ambiguous_contact' : 'contact_not_found',
+        message: result.ambiguous
+          ? '标识匹配到多个联系人，请使用稳定 ID'
+          : '无法解析到现有联系人或群聊',
+        ...(result.ambiguous
+          ? {
+              details: {
+                candidates: result.candidates.map((candidate) => {
+                  const contact = contacts.find((item) => item.md5 === candidate.conversationId)
+                  return contact
+                    ? { id: stableId(contact), name: candidate.displayName }
+                    : { id: candidate.conversationId, name: candidate.displayName }
+                })
+              }
+            }
+          : {})
+      })
+      return undefined
+    }
+
+    if (draft.ruleType === 'daily_report') {
+      const scope: RuleScope =
+        draft.scope === 'group' ? 'group' : draft.scope === 'direct' ? 'person' : 'any'
+      const ids: string[] = []
+      for (const [index, query] of draft.conditions.conversationIds.entries()) {
+        const contact = resolve(query, scope, `conditions.conversationIds[${index}]`)
+        if (contact) ids.push(stableId(contact))
+      }
+      draft.conditions.conversationIds = ids
+    }
+
+    if (draft.ruleType === 'scheduled_report' && draft.scheduledReport) {
+      const source = resolve(
+        draft.scheduledReport.report.sourceConversationId,
+        'group',
+        'scheduledReport.report.sourceConversationId'
+      )
+      if (source) draft.scheduledReport.report.sourceConversationId = source.m_nsUsrName
+      if (draft.scheduledReport.target.type === 'contact') {
+        const target = resolve(
+          draft.scheduledReport.target.contactId || '',
+          'person',
+          'scheduledReport.target.contactId'
+        )
+        if (target) draft.scheduledReport.target.contactId = stableId(target)
+      }
+    }
+
+    if (draft.ruleType === 'leave_notification' && draft.leaveNotification) {
+      if (draft.leaveNotification.target.type === 'contact') {
+        const target = resolve(
+          draft.leaveNotification.target.contactId || '',
+          'person',
+          'leaveNotification.target.contactId'
+        )
+        if (target) draft.leaveNotification.target.contactId = stableId(target)
+      }
+      const ids: string[] = []
+      for (const [index, query] of draft.leaveNotification.notifyRoomIds.entries()) {
+        const group = resolve(query, 'group', `leaveNotification.notifyRoomIds[${index}]`)
+        if (group) ids.push(group.m_nsUsrName)
+      }
+      draft.leaveNotification.notifyRoomIds = ids
+    }
+    return resolved
+  }
+
+  private toStoreDraft(draft: AutomationRuleDraft, contacts: Contact[]): AutomationRuleDraft {
+    const stored = structuredClone(draft)
+    if (stored.ruleType === 'daily_report') {
+      stored.conditions.conversationIds = stored.conditions.conversationIds.map(
+        (id) => contacts.find((contact) => stableId(contact) === id)?.md5 || id
+      )
+    }
+    return stored
+  }
+
+  private async toApiRules(rules: AutomationRule[]): Promise<unknown[]> {
+    const contacts = await this.deps.listContacts().catch(() => [])
+    return rules.map((rule) => {
+      const copy = structuredClone(rule) as AutomationRule & { requiresReview?: boolean }
+      if (copy.ruleType === 'daily_report') {
+        copy.conditions.conversationIds = copy.conditions.conversationIds.map((id) =>
+          contacts.find((contact) => contact.md5 === id)
+            ? stableId(contacts.find((contact) => contact.md5 === id)!)
+            : id
+        )
+      }
+      if (copy.ruleType === 'scheduled_report' && copy.scheduledReport) {
+        copy.requiresReview = copy.scheduledReport.targetNeedsReview === true
+        delete copy.scheduledReport.legacyTarget
+        delete copy.scheduledReport.legacySourceGroup
+        delete copy.scheduledReport.lastRunAt
+        delete copy.scheduledReport.lastScheduledSlot
+      }
+      if (copy.ruleType === 'leave_notification' && copy.leaveNotification) {
+        copy.requiresReview = copy.leaveNotification.targetNeedsReview === true
+      }
+      return copy
+    })
+  }
+
+  private describeEffects(draft: AutomationRuleDraft): Record<string, unknown> {
+    if (draft.ruleType === 'scheduled_report' && draft.scheduledReport) {
+      const config = draft.scheduledReport
+      return {
+        generatesReport: true,
+        sendsWechatMessage: true,
+        sourceConversationId: config.report.sourceConversationId,
+        target: config.target
+      }
+    }
+    if (draft.ruleType === 'leave_notification' && draft.leaveNotification) {
+      return {
+        sendsWechatMessage: true,
+        monitorScope: 'configured_in_group_exit_monitor',
+        notificationScope: draft.leaveNotification.notifyScope,
+        notificationRoomIds:
+          draft.leaveNotification.notifyScope === 'selected'
+            ? draft.leaveNotification.notifyRoomIds
+            : undefined,
+        target: draft.leaveNotification.target
+      }
+    }
+    return {
+      actions: draft.actions.filter((action) => action.enabled).map((action) => action.type),
+      sendsWechatMessage: draft.actions.some(
+        (action) => action.enabled && action.type === 'sendReportImage'
+      )
+    }
+  }
+}
@@ -46,6 +46,21 @@ function parseBody(bodyText: string, contentType?: string): { json?: unknown; bo
  return { bodyText }
 }

+function materializeEndpointPath(pathname: string, query: Record<string, string>): string {
+  return pathname.replace('{conversationId}', encodeURIComponent(query.conversationId || ''))
+}
+
+function appendQueryParameters(
+  url: URL,
+  endpointPath: string,
+  entries: Array<[string, string]>
+): void {
+  for (const [key, value] of entries) {
+    if (endpointPath.includes(`{${key}}`)) continue
+    if (value.trim()) url.searchParams.set(key, value.trim())
+  }
+}
+
 export function buildLocalApiCurlCommand(payload: unknown): {
  success: boolean
  command?: string
@@ -71,18 +86,17 @@ export function buildLocalApiCurlCommand(payload: unknown): {
  const service = apiServer.getState()
  const targetHost = requestHost(service.host)
  const hostPart = targetHost.includes(':') ? `[${targetHost}]` : targetHost
-  const url = new URL(endpoint.path, `http://${hostPart}:${service.port}`)
-  entries.forEach(([key, value]) => {
-    if (value.trim()) url.searchParams.set(key, value.trim())
-  })
+  const endpointPath = materializeEndpointPath(endpoint.path, query as Record<string, string>)
+  const url = new URL(endpointPath, `http://${hostPart}:${service.port}`)
+  appendQueryParameters(url, endpoint.path, entries)
  const token = endpointId === 'health' ? null : apiTokenStore.getTokenForAuthentication()
  if (endpointId !== 'health' && !token) {
    return { success: false, error: 'API Token 安全存储不可用，请在 API Center 检查 Token 状态' }
  }
  const authHeader = token ? ` -H 'Authorization: Bearer ${token}'` : ''
  const command =
-    endpoint.method === 'POST'
-      ? `curl -X POST '${url.toString()}'${authHeader} -H 'Content-Type: application/json' -d '${body.replaceAll("'", "\\'")}'`
+    endpoint.method === 'POST' || endpoint.method === 'PATCH'
+      ? `curl -X ${endpoint.method} '${url.toString()}'${authHeader} -H 'Content-Type: application/json' -d '${body.replaceAll("'", "\\'")}'`
      : `curl '${url.toString()}'${authHeader}`
  return { success: true, command }
 }
@@ -110,10 +124,9 @@ export async function testLocalApiRequest(payload: unknown): Promise<LocalApiTes
  const targetHost = requestHost(service.host)
  const targetPort = service.port
  const hostPart = targetHost.includes(':') ? `[${targetHost}]` : targetHost
-  const url = new URL(endpoint.path, `http://${hostPart}:${targetPort}`)
-  entries.forEach(([key, value]) => {
-    if (value.trim()) url.searchParams.set(key, value.trim())
-  })
+  const endpointPath = materializeEndpointPath(endpoint.path, query as Record<string, string>)
+  const url = new URL(endpointPath, `http://${hostPart}:${targetPort}`)
+  appendQueryParameters(url, endpoint.path, entries)

  if (!service.running) {
    return {
@@ -150,7 +163,9 @@ export async function testLocalApiRequest(payload: unknown): Promise<LocalApiTes
      })
    }
    const headers: Record<string, string> = {}
-    if (endpoint.method === 'POST') headers['Content-Type'] = 'application/json'
+    if (endpoint.method === 'POST' || endpoint.method === 'PATCH') {
+      headers['Content-Type'] = 'application/json'
+    }
    if (token) headers.Authorization = `Bearer ${token}`
    const request = http.request(
      url,
@@ -208,7 +223,7 @@ export async function testLocalApiRequest(payload: unknown): Promise<LocalApiTes
        error: error.message
      })
    })
-    if (endpoint.method === 'POST') request.write(body)
+    if (endpoint.method === 'POST' || endpoint.method === 'PATCH') request.write(body)
    request.end()
  })
 }
@@ -0,0 +1,101 @@
+import { existsSync, readFileSync, statSync } from 'fs'
+import { join } from 'path'
+
+import type { MacWechatRuntimeManifest } from '../../shared/personal-wechat-mac-runtime'
+
+const REQUIRED_CAPABILITIES = ['text', 'image', 'voice'] as const
+
+function assertFile(path: string, label: string): void {
+  if (!existsSync(path) || !statSync(path).isFile()) {
+    throw new Error(`${label} missing: ${path}`)
+  }
+}
+
+function assertMachOArm64(path: string, label: string): void {
+  const binary = readFileSync(path)
+  if (binary.length < 8 || binary.readUInt32LE(0) !== 0xfeedfacf) {
+    throw new Error(`${label} is not a 64-bit Mach-O binary`)
+  }
+  if (binary.readUInt32LE(4) !== 0x0100000c) {
+    throw new Error(`${label} is not arm64`)
+  }
+}
+
+export function readMacWechatRuntimeManifest(
+  runtimeDir: string,
+  architecture = process.arch
+): MacWechatRuntimeManifest {
+  const manifestPath = join(runtimeDir, 'runtime-manifest.json')
+  const hostPath = join(runtimeDir, 'tm-wechat-host')
+  const dylibPath = join(runtimeDir, 'libtmwechat.dylib')
+  assertFile(manifestPath, 'runtime manifest')
+  assertFile(hostPath, 'runtime host')
+  assertFile(dylibPath, 'runtime dylib')
+
+  let manifest: MacWechatRuntimeManifest
+  try {
+    manifest = JSON.parse(readFileSync(manifestPath, 'utf8')) as MacWechatRuntimeManifest
+  } catch (error) {
+    throw new Error(`runtime manifest is unreadable: ${String(error)}`)
+  }
+
+  if (manifest.runtime !== 'tm-wechat-native') {
+    throw new Error(`unexpected runtime name: ${String(manifest.runtime)}`)
+  }
+  if (manifest.platform !== 'darwin-arm64') {
+    throw new Error(`runtime platform must be darwin-arm64: ${String(manifest.platform)}`)
+  }
+  if (manifest.architecture !== 'arm64' || architecture !== 'arm64') {
+    throw new Error(
+      `runtime architecture mismatch: artifact=${String(manifest.architecture)} process=${architecture}`
+    )
+  }
+  if (manifest.protocolVersion !== 1) {
+    throw new Error(`unsupported runtime protocolVersion: ${String(manifest.protocolVersion)}`)
+  }
+  if (!/^[a-f0-9]{64}$/.test(manifest.tmSendSourceSha256 || '')) {
+    throw new Error('runtime manifest has no valid tmSendSourceSha256')
+  }
+  if (!/^[a-f0-9]{64}$/.test(manifest.addressProfileSha256 || '')) {
+    throw new Error('runtime manifest has no valid addressProfileSha256')
+  }
+  if (
+    !Array.isArray(manifest.supportedWechatVersions) ||
+    manifest.supportedWechatVersions.length === 0
+  ) {
+    throw new Error('runtime manifest has no supported WeChat versions')
+  }
+  const capabilities = new Set(manifest.capabilities || [])
+  for (const capability of REQUIRED_CAPABILITIES) {
+    if (!capabilities.has(capability)) {
+      throw new Error(`runtime manifest capability missing: ${capability}`)
+    }
+  }
+  assertMachOArm64(hostPath, 'runtime host')
+  assertMachOArm64(dylibPath, 'runtime dylib')
+  if (!readFileSync(dylibPath).includes(Buffer.from(manifest.tmSendSourceSha256, 'utf8'))) {
+    throw new Error('runtime dylib embedded agent hash does not match manifest')
+  }
+  if (!readFileSync(dylibPath).includes(Buffer.from(manifest.addressProfileSha256, 'utf8'))) {
+    throw new Error('runtime dylib embedded address profile hash does not match manifest')
+  }
+  return manifest
+}
+
+export function bindingMatchesRuntimeManifest(
+  binding: {
+    hostPid?: number
+    runtimeVersion?: string
+    embeddedAgentSha256?: string
+    embeddedAddressProfileSha256?: string
+  },
+  manifest: MacWechatRuntimeManifest
+): boolean {
+  return (
+    Number.isInteger(binding.hostPid) &&
+    Number(binding.hostPid) > 0 &&
+    binding.runtimeVersion === `tm-wechat-runtime/${manifest.version}` &&
+    binding.embeddedAgentSha256 === manifest.tmSendSourceSha256 &&
+    binding.embeddedAddressProfileSha256 === manifest.addressProfileSha256
+  )
+}
@@ -0,0 +1,724 @@
+import { spawn, type ChildProcess } from 'child_process'
+import { chmodSync, existsSync } from 'fs'
+import { get } from 'http'
+import { join } from 'path'
+
+import { app } from 'electron'
+
+import { appLogger } from '../app-logger'
+import { isPackagedRuntime } from '../runtime-mode'
+import type {
+  MacWechatBindResult,
+  MacWechatBindingStatus,
+  MacWechatRuntimeManifest,
+  MacWechatRuntimeLifecycle,
+  MacWechatRuntimeSnapshot
+} from '../../shared/personal-wechat-mac-runtime'
+import type { PersonalWechatSenderStatus } from '../../shared/personal-wechat'
+import {
+  bindingMatchesRuntimeManifest,
+  readMacWechatRuntimeManifest
+} from './mac-wechat-runtime-artifact'
+
+/*
+ * MacWechatRuntimeManager — lifecycle owner of the built-in macOS native
+ * WeChat runtime (tm-wechat-host + libtmwechat.dylib).
+ *
+ * Process isolation is a hard requirement: Electron NEVER dlopens the dylib.
+ * It spawns tm-wechat-host, which is the ONLY binding owner, and talks to it
+ * over 127.0.0.1 HTTP. A native crash therefore cannot take Electron down.
+ *
+ * Binding lifecycle is "bind once per WeChat process":
+ *   - starting the host NEVER attaches to WeChat (the host is lazy)
+ *   - bind happens only on an explicit user action
+ *   - a host crash while bound leaves an orphaned frida agent inside WeChat
+ *     that frida cannot remove from the outside; the only recovery is
+ *     restarting WeChat. Surfaced as wechatRestartRequired. This was verified
+ *     experimentally (2026-09-21) — do not "fix" it by auto-restarting and
+ *     auto-binding, that path dead-ends in a 20s attach timeout every time.
+ */
+
+const PORT = 4290
+const HOST = '127.0.0.1'
+const START_TIMEOUT_MS = 15_000
+const REQUEST_TIMEOUT_MS = 30_000
+const CRASH_RESTART_DELAY_MS = 1_500
+
+interface MacWechatSenderStatusInput {
+  snapshot: MacWechatRuntimeSnapshot
+  binding: MacWechatBindingStatus | null
+  manifest: MacWechatRuntimeManifest
+  arch?: string
+}
+
+export function shouldClearWechatRestartRequired(
+  orphanWechatPid: number,
+  currentWechatPid: number
+): boolean {
+  return orphanWechatPid > 0 && currentWechatPid > 0 && orphanWechatPid !== currentWechatPid
+}
+
+export function deriveMacWechatSenderStatus({
+  snapshot,
+  binding,
+  manifest,
+  arch = process.arch
+}: MacWechatSenderStatusInput): PersonalWechatSenderStatus {
+  const identityMatches = binding !== null && bindingMatchesRuntimeManifest(binding, manifest)
+  const stale = binding?.bindingState === 'stale'
+
+  let state: PersonalWechatSenderStatus['state'] = 'stopped'
+  let message = '正在初始化微信发送能力'
+  if (snapshot.wechatRestartRequired) {
+    state = 'error'
+    message = '微信发送组件异常退出。由于当前微信进程中可能残留上一发送会话，请重启微信后重新绑定。'
+  } else if (binding && !identityMatches) {
+    state = 'error'
+    message = '运行中的微信发送 host 与当前 Native Runtime 不一致，请退出旧 host 后重启 TraceMemo。'
+  } else if (binding && !binding.wechatRunning) {
+    state = 'wechat_not_running'
+    message = '微信未运行'
+  } else if (binding?.bindingState === 'ready') {
+    state = 'online'
+    message = '个人微信已准备好发送日报'
+  } else if (binding?.bindingState === 'bound' || binding?.bindingState === 'binding') {
+    state = 'hook_not_ready'
+    message = '已绑定，等待捕获微信发送上下文'
+  } else if (stale) {
+    state = 'stopped'
+    message = '检测到微信已重启，请重新绑定'
+  } else if (binding?.bindingState === 'failed') {
+    state = 'error'
+    message = '微信绑定失败'
+  } else {
+    state = 'stopped'
+    message = '尚未绑定当前微信，可点击"绑定微信"'
+  }
+
+  const runtimeCapabilities = new Set(manifest.capabilities)
+  const bindingReady = identityMatches && state === 'online' && binding?.sendContextReady === true
+  const canSendText =
+    bindingReady && runtimeCapabilities.has('text') && binding?.canSendText === true
+  const canSendImage =
+    bindingReady && runtimeCapabilities.has('image') && binding?.canSendImage === true
+  const canSendVoice =
+    bindingReady && runtimeCapabilities.has('voice') && binding?.canSendVoice === true
+  const sessionAttached = identityMatches && (binding?.sessionAttached ?? false)
+  const scriptLoaded = identityMatches && (binding?.scriptLoaded ?? false)
+
+  return {
+    state,
+    platform: 'darwin',
+    arch,
+    // 当前 gadget runtime 不依赖宿主 SIP 状态；该字段仅用于兼容旧 sender contract。
+    sipDisabled: true,
+    wechatRunning: binding?.wechatRunning ?? false,
+    wechatPid: binding?.wechatPid ?? 0,
+    boundWechatPid: binding?.boundWechatPid ?? 0,
+    endpoint: snapshot.endpoint,
+    endpointReady: snapshot.lifecycle === 'online',
+    wechatVersion: binding?.wechatVersion,
+    runtimeReady: snapshot.lifecycle === 'online' && identityMatches,
+    attachReady: sessionAttached,
+    baseAddressReady: sessionAttached,
+    textHookInstalled: scriptLoaded,
+    textHookReady: identityMatches && (binding?.sendContextReady ?? false),
+    imageHookInstalled: identityMatches && (binding?.canSendImage ?? false),
+    imageHookReady: identityMatches && (binding?.canSendImage ?? false),
+    messageListenerReady: scriptLoaded,
+    canSend: canSendText || canSendImage || canSendVoice,
+    canSendText,
+    canSendImage,
+    canSendVoice,
+    message,
+    error: state === 'error' ? message : undefined
+  }
+}
+
+class MacWechatRuntimeManager {
+  private child: ChildProcess | null = null
+  private lifecycle: MacWechatRuntimeLifecycle = 'stopped'
+  private wechatRestartRequired = false
+  /* The WeChat pid that was bound when the host died / bind hit the orphan
+   * wall. The orphan agent lives INSIDE that process; once WeChat is restarted
+   * the pid changes and the orphan is gone, so the flag must auto-clear. */
+  private orphanWechatPid = 0
+  private lastBinding: MacWechatBindingStatus | null = null
+  private manifest: MacWechatRuntimeManifest | null = null
+  private lastError: string | null = null
+  private restartTimer: NodeJS.Timeout | null = null
+
+  getPort(): number {
+    return PORT
+  }
+
+  getSnapshot(): MacWechatRuntimeSnapshot {
+    return {
+      lifecycle: this.lifecycle,
+      wechatRestartRequired: this.wechatRestartRequired,
+      endpoint: `http://${HOST}:${PORT}`,
+      port: PORT,
+      runtimeDir: this.resolveRuntimeDir(),
+      manifest: this.manifest ?? undefined,
+      error: this.lastError ?? undefined,
+      binding: this.lastBinding ?? undefined
+    }
+  }
+
+  private resolveRuntimeDir(): string {
+    const base = isPackagedRuntime() ? process.resourcesPath : app.getAppPath()
+    return join(base, 'resources', 'runtime', 'darwin-arm64')
+  }
+
+  private resolveHostPath(): string {
+    return join(this.resolveRuntimeDir(), 'tm-wechat-host')
+  }
+
+  /** Runtime binaries present and executable? Cheap check for the settings UI. */
+  isRuntimePresent(): boolean {
+    try {
+      this.manifest = readMacWechatRuntimeManifest(this.resolveRuntimeDir())
+      return true
+    } catch {
+      return false
+    }
+  }
+
+  /**
+   * Start the host if it is not already running. Deliberately lazy: the host
+   * comes up WITHOUT touching WeChat (lazy init inside the host), so "runtime
+   * online" and "WeChat bound" stay two independent states.
+   */
+  async ensureStarted(): Promise<MacWechatRuntimeSnapshot> {
+    if (this.lifecycle === 'starting') {
+      return this.getSnapshot()
+    }
+
+    const runtimeDir = this.resolveRuntimeDir()
+    try {
+      this.manifest = readMacWechatRuntimeManifest(runtimeDir)
+      appLogger.write({
+        level: 'info',
+        scope: 'mac-wechat-runtime',
+        message: 'runtime_artifact_validated',
+        details: {
+          runtime: this.manifest.runtime,
+          version: this.manifest.version,
+          protocolVersion: this.manifest.protocolVersion,
+          agentSha256: this.manifest.tmSendSourceSha256,
+          addressProfileSha256: this.manifest.addressProfileSha256
+        }
+      })
+    } catch (error) {
+      this.manifest = null
+      const detail = error instanceof Error ? error.message : String(error)
+      appLogger.write({
+        level: 'warn',
+        scope: 'mac-wechat-runtime',
+        message: 'runtime_artifact_unavailable',
+        details: { detail }
+      })
+      this.lastError = '暂无发送能力'
+      this.lifecycle = 'stopped'
+      return this.getSnapshot()
+    }
+
+    if (this.lifecycle === 'online') {
+      await this.refreshBinding().catch(() => undefined)
+      if (
+        this.lastBinding &&
+        this.manifest &&
+        bindingMatchesRuntimeManifest(this.lastBinding, this.manifest)
+      ) {
+        return this.getSnapshot()
+      }
+      this.lifecycle = 'stopped'
+      this.lastError = '运行中的 host 与 runtime manifest 不一致'
+      return this.getSnapshot()
+    }
+
+    /*
+     * Adopt-first: the host is designed to OUTLIVE TraceMemo ("keep the send
+     * capability process"), so a previous app session may have left it running
+     * with a live, bound session. If something is already answering on our
+     * port, adopt it instead of spawning a second one (which would just die on
+     * the occupied port). Adopting is pure HTTP — the running host keeps its
+     * frida session, so the app is immediately ready with NO re-bind, exactly
+     * the "keep the send capability process" behaviour the user asked to carry over.
+     */
+    if (await this.ping()) {
+      this.lifecycle = 'online'
+      this.child = null
+      this.wechatRestartRequired = false
+      await this.refreshBinding().catch(() => undefined)
+      if (
+        !this.lastBinding ||
+        !this.manifest ||
+        !bindingMatchesRuntimeManifest(this.lastBinding, this.manifest)
+      ) {
+        this.lifecycle = 'stopped'
+        this.lastError =
+          '检测到无法确认身份的微信发送 host。请先退出旧 host，再重新启动 TraceMemo。'
+        return this.getSnapshot()
+      }
+      return this.getSnapshot()
+    }
+    if (this.lifecycle === 'crashed_bound' || this.wechatRestartRequired) {
+      // Restarting the host is safe (it does not touch WeChat), but binding is
+      // not: the orphaned agent from the previous session is still inside the
+      // current WeChat process. Start anyway so the UI can show live status.
+      appLogger.write({
+        level: 'warn',
+        scope: 'mac-wechat-runtime',
+        message: 'restart_host_after_bound_crash',
+        details: { note: 'bind stays blocked until WeChat restarts' }
+      })
+    }
+
+    const hostPath = this.resolveHostPath()
+    if (!existsSync(hostPath)) {
+      this.lastError = `runtime host missing: ${hostPath}`
+      this.lifecycle = 'stopped'
+      return this.getSnapshot()
+    }
+    try {
+      chmodSync(hostPath, 0o755)
+    } catch {
+      /* best effort — dev checkouts usually keep the exec bit */
+    }
+
+    this.lifecycle = 'starting'
+    this.lastError = null
+
+    try {
+      /* Host stderr goes to /dev/null (stdio ignored, see note above), so the
+       * host writes its own log file via TM_WECHAT_LOG_FILE. Without this the
+       * agent's stage beacons (send-enter / cdn-enter / …) are invisible and a
+       * media/CDN stall is undiagnosable. info level so stage beacons show. */
+      const hostLogFile = join(app.getPath('logs'), 'tm-wechat-host.log')
+      this.child = spawn(
+        hostPath,
+        [
+          '--port',
+          String(PORT),
+          '--manifest',
+          join(runtimeDir, 'runtime-manifest.json'),
+          '--log-level',
+          'info'
+        ],
+        {
+          cwd: runtimeDir,
+          stdio: 'ignore',
+          detached: false,
+          env: { ...process.env, TM_WECHAT_LOG_FILE: hostLogFile }
+        }
+      )
+    } catch (error) {
+      this.lifecycle = 'stopped'
+      this.lastError = String(error)
+      return this.getSnapshot()
+    }
+
+    this.child.on('exit', (code) => {
+      const wasBound =
+        this.lastBinding?.bound === true ||
+        this.lastBinding?.sessionAttached === true ||
+        this.lastBinding?.bindingState === 'binding'
+      const orphanPid = this.lastBinding?.boundWechatPid || this.lastBinding?.wechatPid || 0
+      this.child = null
+      this.lastBinding = null
+      if (this.lifecycle === 'stopped') return // deliberate shutdown
+
+      if (wasBound) {
+        this.lifecycle = 'crashed_bound'
+        this.wechatRestartRequired = true
+        this.orphanWechatPid = orphanPid
+        appLogger.write({
+          level: 'error',
+          scope: 'mac-wechat-runtime',
+          message: 'host_died_while_bound',
+          details: { code: code ?? null, orphanWechatPid: orphanPid }
+        })
+      } else {
+        this.lifecycle = 'crashed_unbound'
+        appLogger.write({
+          level: 'warn',
+          scope: 'mac-wechat-runtime',
+          message: 'host_died_while_unbound_auto_restart',
+          details: { code: code ?? null }
+        })
+        if (this.restartTimer) clearTimeout(this.restartTimer)
+        this.restartTimer = setTimeout(() => {
+          void this.ensureStarted()
+        }, CRASH_RESTART_DELAY_MS)
+      }
+    })
+
+    // Wait for /healthz
+    const deadline = Date.now() + START_TIMEOUT_MS
+    while (Date.now() < deadline) {
+      if (await this.ping()) {
+        this.lifecycle = 'online'
+        await this.refreshBinding().catch(() => undefined)
+        if (
+          !this.lastBinding ||
+          !this.manifest ||
+          !bindingMatchesRuntimeManifest(this.lastBinding, this.manifest)
+        ) {
+          this.lifecycle = 'crashed_unbound'
+          this.lastError = '运行中的 host 与 runtime manifest 不一致'
+          return this.getSnapshot()
+        }
+        return this.getSnapshot()
+      }
+      if (this.child === null) break // died during startup
+      await new Promise((resolve) => setTimeout(resolve, 300))
+    }
+
+    if (this.lifecycle === 'starting') {
+      this.lifecycle = 'crashed_unbound'
+      this.lastError = 'host did not answer /healthz in time'
+    }
+    return this.getSnapshot()
+  }
+
+  private request(
+    method: string,
+    path: string,
+    timeoutMs = REQUEST_TIMEOUT_MS
+  ): Promise<{ status: number | null; body: string }> {
+    return new Promise((resolve) => {
+      const request = get(
+        { host: HOST, port: PORT, path, method, timeout: timeoutMs },
+        (response) => {
+          let body = ''
+          response.on('data', (chunk) => {
+            body += chunk
+          })
+          response.on('end', () => {
+            resolve({ status: response.statusCode ?? null, body })
+          })
+        }
+      )
+      request.on('timeout', () => {
+        request.destroy()
+        resolve({ status: null, body: '' })
+      })
+      request.on('error', () => {
+        resolve({ status: null, body: '' })
+      })
+      request.end()
+    })
+  }
+
+  async ping(): Promise<boolean> {
+    const { status } = await this.request('GET', '/healthz', 3_000)
+    if (status === 200) return true
+
+    /*
+     * Adopted hosts have no child handle, so their death is only observable
+     * through failed probes. Classify here: dying while bound means WeChat
+     * keeps an orphaned agent (restart required); dying while unbound is
+     * auto-recoverable.
+     */
+    if (this.lifecycle === 'online') {
+      const wasBound =
+        this.lastBinding?.bound === true ||
+        this.lastBinding?.sessionAttached === true ||
+        this.lastBinding?.bindingState === 'binding'
+      const orphanPid = this.lastBinding?.boundWechatPid || this.lastBinding?.wechatPid || 0
+      this.lastBinding = null
+      if (wasBound) {
+        this.lifecycle = 'crashed_bound'
+        this.wechatRestartRequired = true
+        this.orphanWechatPid = orphanPid
+        appLogger.write({
+          level: 'error',
+          scope: 'mac-wechat-runtime',
+          message: 'adopted_host_died_while_bound',
+          details: { note: 'WeChat restart required', orphanWechatPid: orphanPid }
+        })
+      } else {
+        this.lifecycle = 'crashed_unbound'
+      }
+    }
+    return false
+  }
+
+  /** Pull /bindingStatus and cache it for crash classification. */
+  async refreshBinding(): Promise<MacWechatBindingStatus | null> {
+    const { status, body } = await this.request('GET', '/bindingStatus')
+    if (status !== 200) return null
+    try {
+      this.lastBinding = JSON.parse(body) as MacWechatBindingStatus
+      /*
+       * A 20s bind timeout is NOT proof of the orphan-agent wall: in the
+       * stale-rebind path the host worker waits up to 45s for the old session
+       * teardown before attaching, so the HTTP response times out while the
+       * attach quietly succeeds afterwards. When the binding is later observed
+       * alive and matching, the restart-required verdict was a false positive
+       * and must be withdrawn — otherwise the manager keeps rejecting every
+       * further bind locally while WeChat is actually bound and working.
+       */
+      if (
+        this.wechatRestartRequired &&
+        (this.lastBinding.bound === true || this.lastBinding.bindingState === 'ready') &&
+        this.lastBinding.boundWechatPid === this.lastBinding.wechatPid
+      ) {
+        this.wechatRestartRequired = false
+        appLogger.write({
+          level: 'warn',
+          scope: 'mac-wechat-runtime',
+          message: 'restart_required_withdrawn',
+          details: { note: 'binding observed alive after bind timeout' }
+        })
+      }
+      return this.lastBinding
+    } catch {
+      return null
+    }
+  }
+
+  async getBindingStatus(): Promise<MacWechatBindingStatus | null> {
+    if (this.lifecycle !== 'online') return this.lastBinding
+    await this.refreshBinding().catch(() => null)
+    return this.lastBinding
+  }
+
+  /*
+   * Build the legacy PersonalWechatSenderStatus contract from the live mac
+   * runtime state. Lives here (not in the capability service) so BOTH the
+   * capability mapping and the send path report the same facts without a
+   * circular import.
+   */
+  async buildSenderStatus(): Promise<PersonalWechatSenderStatus> {
+    /*
+     * Lazy-start the host from status reads.
+     *
+     * Every UI surface discovers the runtime through this method, and the
+     * setup guide renders "正在检查 发送运行时…" while runtimeReady is false —
+     * where runtimeReady means lifecycle === 'online'. If nothing here brings a
+     * stopped host up, opening the app (with WeChat already logged in, so no
+     * send is ever attempted) leaves that step spinning forever: the check can
+     * never succeed because the check itself is what should start the host.
+     * Starting it never touches WeChat, so doing it from a read is safe.
+     */
+    if (this.lifecycle === 'stopped') {
+      await this.ensureStarted().catch(() => undefined)
+    }
+    /* Self-heal the status card too: if WeChat was restarted, drop the stale
+     * restartRequired before it is read into the UI as an error. */
+    try {
+      this.manifest = readMacWechatRuntimeManifest(this.resolveRuntimeDir())
+    } catch (error) {
+      const detail = error instanceof Error ? error.message : String(error)
+      appLogger.write({
+        level: 'warn',
+        scope: 'mac-wechat-runtime',
+        message: 'runtime_artifact_unavailable',
+        details: { detail }
+      })
+      const message = '暂无发送能力'
+      this.manifest = null
+      this.lastError = message
+      return {
+        state: 'runtime_missing',
+        platform: 'darwin',
+        arch: process.arch,
+        sipDisabled: true,
+        wechatRunning: false,
+        endpoint: `http://${HOST}:${PORT}`,
+        endpointReady: false,
+        runtimeReady: false,
+        attachReady: false,
+        baseAddressReady: false,
+        textHookInstalled: false,
+        textHookReady: false,
+        imageHookInstalled: false,
+        imageHookReady: false,
+        messageListenerReady: false,
+        canSend: false,
+        canSendText: false,
+        canSendImage: false,
+        canSendVoice: false,
+        message,
+        error: message
+      }
+    }
+    await this.clearRestartRequiredIfWechatRestarted()
+    const snapshot = this.getSnapshot()
+    const binding = await this.getBindingStatus()
+    return deriveMacWechatSenderStatus({ snapshot, binding, manifest: this.manifest })
+  }
+
+  /*
+   * The orphan agent lives inside ONE specific WeChat process. If the WeChat
+   * pid we recorded when restartRequired was raised no longer matches the
+   * WeChat pid the host currently sees, WeChat has been restarted and the
+   * orphan is gone — the flag must clear or every later bind is rejected
+   * forever even though the user DID restart WeChat (the reported bug).
+   */
+  private async clearRestartRequiredIfWechatRestarted(): Promise<void> {
+    if (!this.wechatRestartRequired) return
+    await this.refreshBinding().catch(() => null)
+    const currentPid = this.lastBinding?.wechatPid ?? 0
+    if (shouldClearWechatRestartRequired(this.orphanWechatPid, currentPid)) {
+      this.wechatRestartRequired = false
+      this.orphanWechatPid = 0
+      if (this.lifecycle === 'crashed_bound') this.lifecycle = 'online'
+      appLogger.write({
+        level: 'info',
+        scope: 'mac-wechat-runtime',
+        message: 'restart_required_cleared_wechat_restarted',
+        details: { currentWechatPid: currentPid }
+      })
+    }
+  }
+
+  /**
+   * Explicit user action only. Idempotent on the host side: binding the same
+   * live WeChat pid again returns attached=false and never double-attaches.
+   */
+  async bind(): Promise<MacWechatBindResult> {
+    if (this.lifecycle !== 'online') {
+      const snapshot = await this.ensureStarted()
+      if (snapshot.lifecycle !== 'online') {
+        return {
+          ok: false,
+          attached: false,
+          state: 'runtime_offline',
+          message: snapshot.error ?? 'Runtime 未在线，无法绑定'
+        }
+      }
+    }
+
+    /* Auto-clear if WeChat was restarted since the flag was raised —
+     * otherwise this branch rejects every bind forever (the reported bug). */
+    await this.clearRestartRequiredIfWechatRestarted()
+
+    if (this.wechatRestartRequired) {
+      return {
+        ok: false,
+        attached: false,
+        state: 'wechat_restart_required',
+        message:
+          'Previous WeChat send session was not fully released. ' + 'Restart WeChat and bind again.'
+      }
+    }
+
+    const { status, body } = await this.request('POST', '/bindWeChat')
+    await this.refreshBinding().catch(() => null)
+
+    if (status !== 200) {
+      return {
+        ok: false,
+        attached: false,
+        state: 'host_error',
+        message: `host 返回 HTTP ${status ?? '无响应'}`
+      }
+    }
+    try {
+      const parsed = JSON.parse(body) as MacWechatBindResult
+      if (!parsed.ok && parsed.message?.includes('not fully released')) {
+        this.wechatRestartRequired = true
+        this.orphanWechatPid = this.lastBinding?.boundWechatPid || this.lastBinding?.wechatPid || 0
+        appLogger.write({
+          level: 'error',
+          scope: 'mac-wechat-runtime',
+          message: 'bind_hit_orphan_agent_wall',
+          details: { note: 'WeChat restart required', orphanWechatPid: this.orphanWechatPid }
+        })
+      }
+      return parsed
+    } catch {
+      return { ok: false, attached: false, state: 'bad_response', message: body.slice(0, 200) }
+    }
+  }
+
+  /*
+   * 重新加载发送组件：优雅停掉当前 host 进程，再启动一个新实例。
+   * 用于：更新二进制后换新代码；或 host 卡死时的手动恢复。
+   * 注意：若当前 host 已绑定微信，此操作会让微信里留下孤儿 agent，
+   * 需要重启微信后重新绑定（UI 会提示）。
+   */
+  async reload(): Promise<MacWechatRuntimeSnapshot> {
+    await this.shutdown().catch(() => undefined)
+    await new Promise((resolve) => setTimeout(resolve, 2_000))
+    return this.ensureStarted()
+  }
+
+  private isProcessAlive(pid: number): boolean {
+    if (!Number.isInteger(pid) || pid <= 0) return false
+    try {
+      process.kill(pid, 0)
+      return true
+    } catch {
+      return false
+    }
+  }
+
+  /**
+   * Graceful stop only. A bound host must never be force-killed: doing so can
+   * leave its injected agent inside WeChat and make the next attach hang.
+   */
+  async shutdown(): Promise<void> {
+    if (this.restartTimer) {
+      clearTimeout(this.restartTimer)
+      this.restartTimer = null
+    }
+
+    if (this.lifecycle === 'stopped' && this.child === null) return
+
+    const reachable = await this.ping()
+    if (!reachable && this.child === null) {
+      this.lifecycle = 'stopped'
+      this.lastBinding = null
+      return
+    }
+
+    await this.refreshBinding().catch(() => null)
+    const child = this.child
+    const hostPid = this.lastBinding?.hostPid ?? 0
+    this.lifecycle = 'stopped'
+    this.lastError = null
+
+    const response = await this.request('POST', '/shutdown', 5_000)
+    if (response.status !== 200) {
+      this.lifecycle = 'online'
+      this.lastError = 'Native Runtime 未确认退出请求，已保留 host 以避免微信残留会话。'
+      throw new Error(this.lastError)
+    }
+
+    const deadline = Date.now() + 55_000
+    let transportClosedAt = 0
+    while (Date.now() < deadline) {
+      const childExited = child !== null && child.exitCode !== null
+      const pidExited = hostPid > 0 && !this.isProcessAlive(hostPid)
+      if (childExited || pidExited) {
+        this.child = null
+        this.lastBinding = null
+        return
+      }
+      if (child === null && hostPid === 0) {
+        const transportOpen = await this.ping()
+        if (!transportOpen) {
+          if (transportClosedAt === 0) transportClosedAt = Date.now()
+          /* Compatibility with a pre-hostPid runtime during one upgrade:
+           * its teardown waits at most 45s after closing the listen socket. */
+          if (Date.now() - transportClosedAt >= 46_000) {
+            this.lastBinding = null
+            return
+          }
+        } else {
+          transportClosedAt = 0
+        }
+      }
+      await new Promise((resolve) => setTimeout(resolve, 250))
+    }
+
+    this.lastError = 'Native Runtime 安全退出超时；未强制结束绑定态 host，请先退出微信后再重试。'
+    throw new Error(this.lastError)
+  }
+}
+
+export const macWechatRuntimeManager = new MacWechatRuntimeManager()
@@ -0,0 +1,441 @@
+import { decompress as zstdDecompress } from 'fzstd'
+import type { Wcdb4Client, Wcdb4Message, Wcdb4MonitorEvent } from '../wcdb4-client'
+
+/**
+ * MessageListener —— 实时消息回读底座。
+ *
+ * **职责边界（刻意很窄）**：
+ *
+ * ```
+ * WCDB native change
+ *         ↓
+ *      coalesce
+ *         ↓
+ *   bounded DB readback
+ *         ↓
+ *       dedup
+ *         ↓
+ *   NormalizedIncomingMessage   →  onMessage(callback)
+ * ```
+ *
+ * **不负责**（这些属于下一层，本模块不许碰）：关键词匹配、@我业务判断、日报、
+ * 自动回复、AI 调用、Agent 调度、发送消息。
+ *
+ * 设计约束：一次写入会触发**一连串** native event（十几条），
+ * 所以**绝不能**一个事件触发一次业务动作。
+ */
+
+/** zstd 帧魔数（微信 `source` 列是 zstd 压缩）。 */
+const ZSTD_MAGIC = [0x28, 0xb5, 0x2f, 0xfd]
+
+/**
+ * coalesce 窗口。
+ *
+ * 实测同一批 native event 会**同时**到达（同一毫秒内十几个），所以只需要一个极短的
+ * 合并窗口就能把它们收敛成一次回读。取 120ms：
+ * - 足够吃掉同一批事件（实测事件间隔 < 5ms）；
+ * - 相对「event → 可读」本身就有 ~1.2s 的落库延迟，这点等待**不构成额外延迟**；
+ * - 与项目既有 `wcdb-change` 消费端的 350ms debounce 相比更短，不会叠加成明显卡顿。
+ */
+const DEFAULT_COALESCE_MS = 120
+
+/** 回读窗口：只看最近这么久，绝不扫历史。 */
+const DEFAULT_LOOKBACK_SEC = 30
+/** 未来容忍（时钟漂移 + 秒级取整）。 */
+const DEFAULT_LOOKAHEAD_SEC = 2
+/** 单次回读条数上限。 */
+const DEFAULT_READ_LIMIT = 50
+
+/** dedup 条目存活时间：超过它就可以被淘汰（同一条消息不会在窗口外再被读到）。 */
+const DEFAULT_DEDUP_TTL_MS = 5 * 60 * 1000
+/** dedup 容量上限，防止长时间运行后无限增长。 */
+const DEFAULT_DEDUP_MAX_ENTRIES = 5_000
+
+/**
+ * 规范化后的入站消息。
+ *
+ * 刻意**克制**：只包含 Trigger 后续真正需要的字段，不做「万能 Message DTO」。
+ * 原始行（`raw`）、未解压的 Buffer、头像等一律不往外传。
+ */
+export interface NormalizedIncomingMessage {
+  sessionId: string
+  /** 会话内序号。**单独不保证跨 shard 唯一**，去重必须带上 sessionId。 */
+  localId: string
+  /** 服务器侧全局标识。**可能缺失**，所以只作强标识、不作必需字段。 */
+  serverId?: string
+  /**
+   * Unix **epoch 秒** —— 与 WCDB `create_time` / `getMessages` 参数同口径。
+   *
+   * ⚠️ 本项目另有模块使用 epoch 毫秒。**换算只允许集中在本层**：
+   * 凡是需要毫秒的消费者，自己明确 `/1000` 的**唯一**位置就是这里之后的调用点，
+   * 不许在任意函数里散落 `* 1000` / `/ 1000`。
+   */
+  createTime: number
+  messageType: number
+  /** `mesDes === 0` 表示自己发送（字段语义与直觉相反）。 */
+  isSelf: boolean
+  senderId?: string
+  senderNickname?: string
+  content?: string
+  /** **已解压**的 source XML；原始 zstd Buffer 不往外传。 */
+  source?: string
+  isGroup: boolean
+  /**
+   * 从 `source` 解析出的 @ 目标（**微信 username，不是昵称**）。
+   *
+   * 本轮只产出数据，**不做「是否 @ 我」的业务判断** —— 那需要与
+   * `getMyUsernameCandidates()` 求交集，属于下一层。
+   */
+  mentionTargets: string[]
+}
+
+/** 监听统计（不含任何身份信息，可安全记录）。 */
+export interface MessageListenerStats {
+  /** 收到的 native change event 数。 */
+  nativeEvents: number
+  /** 被 coalesce 合并掉的事件数（未单独触发回读）。 */
+  coalescedEvents: number
+  /** 实际执行的回读次数。 */
+  readbacks: number
+  /** 通过 dedup 并投递出去的消息数。 */
+  delivered: number
+  /** 因 dedup 被丢弃的重复投递数。 */
+  deduped: number
+}
+
+export interface MessageListenerOptions {
+  coalesceMs?: number
+  lookbackSec?: number
+  lookaheadSec?: number
+  readLimit?: number
+  dedupTtlMs?: number
+  dedupMaxEntries?: number
+}
+
+/**
+ * 从 `source`（**已解压的 XML 文本**）提取 @ 目标。
+ *
+ * 纯函数：不依赖任何运行时状态，便于单测。
+ *
+ * 实测两种形式都真实存在，**都必须支持**：
+ * ```xml
+ * <atuserlist><![CDATA[SELF_USERNAME]]></atuserlist>
+ * <atuserlist>SELF_USERNAME</atuserlist>
+ * ```
+ *
+ * ⚠️ 三条硬规则：
+ * 1. **CDATA 外壳必须剥掉**，否则拿到的是 `<![CDATA[xxx]]>` 字面量；
+ * 2. **值就是微信 username，不保证以 `wxid_` 开头** —— 绝不能用前缀做过滤/校验；
+ * 3. **不允许多人分隔符的猜测** —— 多人 @ 的格式尚未在真实样本中验证过，
+ *    这里按「单个值原样返回」处理；一旦拿到真实多人样本需补解析与测试
+ *    （当前形状已经预留为 `string[]`）。
+ */
+export function extractMentionTargets(source: string | undefined): string[] {
+  const text = String(source ?? '')
+  if (!text) return []
+  const match = /<atuserlist>([\s\S]*?)<\/atuserlist>/i.exec(text)
+  if (!match) return []
+  const inner = match[1].trim()
+  const unwrapped = inner.replace(/^<!\[CDATA\[/, '').replace(/\]\]>$/, '').trim()
+  return unwrapped ? [unwrapped] : []
+}
+
+/** 把 WCDB 的 `source` 列（zstd 压缩的 Buffer）还原成 XML 文本。 */
+export function decodeSourcePayload(value: unknown): string | undefined {
+  let buffer: Buffer | null = null
+  if (Buffer.isBuffer(value)) {
+    buffer = value
+  } else {
+    const candidate = value as { type?: string; data?: unknown } | null
+    if (candidate?.type === 'Buffer' && Array.isArray(candidate.data)) {
+      buffer = Buffer.from(candidate.data as number[])
+    }
+  }
+  if (!buffer || buffer.length === 0) return undefined
+  const isZstd = ZSTD_MAGIC.every((byte, index) => buffer![index] === byte)
+  if (!isZstd) return buffer.toString('utf8')
+  try {
+    return Buffer.from(zstdDecompress(buffer)).toString('utf8')
+  } catch {
+    return undefined
+  }
+}
+
+/**
+ * 实时消息监听。
+ *
+ * ⚠️ **已知 BLOCKER（必须如实标注）**：
+ * 本实现**无法从 native event 判断是哪个会话发生了变化** —— 事件 payload 只有
+ * `{db, table, action}`（native 侧 `MonitorEvent` 结构就这两个字段）。
+ * 因此当前策略是**回读「最近活跃会话」**（`sessions[0]`，按 `last_timestamp` 排序），
+ * 这在「刚刚收到消息的会话」这一场景下成立，但：
+ *
+ * - **不是**「全局监听所有微信会话」；
+ * - 若同一窗口内有**多个会话**同时来消息，只会回读到其中最近的那个；
+ * - 正式扩展需要解决「如何从 table event 推断需要回读哪些会话」。
+ *
+ * 详见 `NEXT_STEPS` 注释与 Spike 报告。
+ */
+export class MessageListenerService {
+  private readonly listeners = new Set<(message: NormalizedIncomingMessage) => void>()
+  /**
+   * dedup 缓存：`sessionId:localId` → 首次见到的时间戳。
+   *
+   * 必须有界 —— Spike 里用的是无上限 `Set`，长时间运行会持续吃内存。
+   */
+  private readonly seen = new Map<string, number>()
+  /**
+   * 本批 coalesce 窗口内**出现过精确会话**的事件收集到的 session 集合。
+   *
+   * v2 事件（Native Monitor Event v2）会带上发生变化的 sessionId，这里用 **Set** 收集 ——
+   * 120ms 内收到 `A B A C B` 必须回读 A/B/C 三个，绝不能「最后一个 wins」。
+   */
+  private readonly pendingSessions = new Set<string>()
+  private coalesceTimer: ReturnType<typeof setTimeout> | null = null
+  private readbackInFlight = false
+  private disposed = false
+
+  private readonly coalesceMs: number
+  private readonly lookbackSec: number
+  private readonly lookaheadSec: number
+  private readonly readLimit: number
+  private readonly dedupTtlMs: number
+  private readonly dedupMaxEntries: number
+
+  private counters: MessageListenerStats = {
+    nativeEvents: 0,
+    coalescedEvents: 0,
+    readbacks: 0,
+    delivered: 0,
+    deduped: 0
+  }
+
+  constructor(
+    private readonly client: Wcdb4Client,
+    options: MessageListenerOptions = {}
+  ) {
+    this.coalesceMs = options.coalesceMs ?? DEFAULT_COALESCE_MS
+    this.lookbackSec = options.lookbackSec ?? DEFAULT_LOOKBACK_SEC
+    this.lookaheadSec = options.lookaheadSec ?? DEFAULT_LOOKAHEAD_SEC
+    this.readLimit = options.readLimit ?? DEFAULT_READ_LIMIT
+    this.dedupTtlMs = options.dedupTtlMs ?? DEFAULT_DEDUP_TTL_MS
+    this.dedupMaxEntries = options.dedupMaxEntries ?? DEFAULT_DEDUP_MAX_ENTRIES
+  }
+
+  /**
+   * 订阅新消息。返回取消订阅函数。
+   *
+   * 刻意**只提供回调**：当前规模不需要 EventEmitter / RxJS / 内部队列。
+   */
+  onMessage(listener: (message: NormalizedIncomingMessage) => void): () => void {
+    this.listeners.add(listener)
+    return () => this.listeners.delete(listener)
+  }
+
+  stats(): MessageListenerStats {
+    return { ...this.counters }
+  }
+
+  dispose(): void {
+    this.disposed = true
+    if (this.coalesceTimer) {
+      clearTimeout(this.coalesceTimer)
+      this.coalesceTimer = null
+    }
+    this.listeners.clear()
+    this.seen.clear()
+  }
+
+  /**
+   * 接 native change event。
+   *
+   * 只做计数、收集会话与 coalesce —— **绝不**在这里回读，更不触发任何业务。
+   *
+   * `event.protocol === 2` 时把 `sessionId` 收进本批的集合；legacy 事件（v1，不带会话）
+   * 走到回读时仍然只能退化成「最近活跃会话」。
+   */
+  handleNativeChange(event?: Wcdb4MonitorEvent): void {
+    if (this.disposed) return
+    this.counters.nativeEvents += 1
+    if (event?.protocol === 2 && event.sessionId) {
+      this.pendingSessions.add(event.sessionId)
+    }
+    if (this.coalesceTimer) {
+      // 已经在同一个窗口里：直接合并掉，这就是 17:1 那个比值的收敛点。
+      this.counters.coalescedEvents += 1
+      return
+    }
+    this.coalesceTimer = setTimeout(() => {
+      this.coalesceTimer = null
+      void this.readback()
+    }, this.coalesceMs)
+  }
+
+  /**
+   * 有界回读 → normalize → dedup → 投递。
+   *
+   * 回读范围严格受限：**按会话** + **小时间窗口** + **条数上限**。不遍历会话、不扫全库。
+   *
+   * 目标会话的来源有两种：
+   * - **precise（protocol v2）**：本批事件收集到的 `sessionId` 集合，逐个回读
+   *   （多个会话同时来消息时，A/B/C 都会被读到）；
+   * - **legacy（protocol v1，旧 runtime）**：只能退回「最近活跃会话」`getSessions()[0]`。
+   */
+  private async readback(): Promise<void> {
+    if (this.disposed || this.readbackInFlight) return
+    this.readbackInFlight = true
+    const startedAt = Date.now()
+
+    const preciseSessions = Array.from(this.pendingSessions)
+    this.pendingSessions.clear()
+
+    try {
+      let delivered = 0
+      let sessions = 0
+
+      if (preciseSessions.length > 0) {
+        for (const sessionId of preciseSessions) {
+          if (this.disposed) return
+          delivered += await this.readbackSession(sessionId)
+          sessions += 1
+        }
+      } else {
+        // legacy：旧 runtime 的 payload 不带会话，只能回读「最近活跃会话」。
+        const session = this.client.getSessions()[0]
+        if (!session?.username) return
+        delivered = await this.readbackSession(session.username)
+        sessions = 1
+      }
+
+      if (delivered > 0) {
+        // 日志只记录协议版本、数量与耗时 —— **不含** wxid / 群名 / 昵称 / 正文 / source / sessionId。
+        console.log(
+          `[MessageListener] protocol=${preciseSessions.length > 0 ? 'v2' : 'v1'}` +
+            ` sessions=${sessions} delivered=${delivered} readbackMs=${Date.now() - startedAt}` +
+            ` events=${this.counters.nativeEvents} coalesced=${this.counters.coalescedEvents}`
+        )
+      }
+    } catch (error) {
+      console.warn(
+        `[MessageListener] readback failed: ${error instanceof Error ? error.message : String(error)}`
+      )
+    } finally {
+      this.readbackInFlight = false
+    }
+  }
+
+  /** 回读**单个**会话并投递。返回本次真正投递出去的条数。 */
+  private async readbackSession(sessionId: string): Promise<number> {
+    const nowSec = Math.floor(Date.now() / 1000)
+    this.counters.readbacks += 1
+    const messages = await this.client.getMessagesAsync(
+      sessionId,
+      nowSec - this.lookbackSec,
+      nowSec + this.lookaheadSec,
+      { limit: this.readLimit }
+    )
+    if (this.disposed) return 0
+
+    let delivered = 0
+    for (const message of messages) {
+      const normalized = this.normalize(sessionId, message)
+      if (!normalized) continue
+      // dedup：同一条消息可能在多个读回窗口中反复出现（native 事件本身也可能重复）。
+      if (!this.markSeen(normalized)) {
+        this.counters.deduped += 1
+        continue
+      }
+      this.counters.delivered += 1
+      delivered += 1
+      this.deliver(normalized)
+    }
+    return delivered
+  }
+
+  private deliver(message: NormalizedIncomingMessage): void {
+    for (const listener of this.listeners) {
+      try {
+        listener(message)
+      } catch (error) {
+        console.warn(
+          `[MessageListener] listener threw: ${
+            error instanceof Error ? error.message : String(error)
+          }`
+        )
+      }
+    }
+  }
+
+  /** 把 WCDB 行规范化为克制的 DTO。返回 null 表示这条行缺关键标识，无法投递。 */
+  private normalize(
+    sessionUsername: string,
+    message: Wcdb4Message
+  ): NormalizedIncomingMessage | null {
+    const localId = message.mesLocalID === undefined ? '' : String(message.mesLocalID)
+    if (!localId) return null
+
+    const createTime = Number(message.msgCreateTime) || 0
+    if (!createTime) return null
+
+    const raw = (message.raw ?? {}) as Record<string, unknown>
+    const source = decodeSourcePayload(raw.source)
+    const senderId = message.sender ? String(message.sender) : undefined
+
+    return {
+      sessionId: sessionUsername,
+      localId,
+      serverId: message.serverId ? String(message.serverId) : undefined,
+      createTime,
+      messageType: Number(message.messageType) || 0,
+      // `mesDes === 0` 是自己发送。不用昵称 / sender 文本 / content / username 前缀判断。
+      isSelf: Number(message.mesDes) === 0,
+      senderId,
+      senderNickname: message.senderNickname ? String(message.senderNickname) : undefined,
+      content: message.msgContent ? String(message.msgContent) : undefined,
+      source,
+      isGroup: sessionUsername.endsWith('@chatroom'),
+      mentionTargets: extractMentionTargets(source)
+    }
+  }
+
+  /** 登记并返回「是否是第一次见到」。同时做 TTL + 容量淘汰。 */
+  private markSeen(message: NormalizedIncomingMessage): boolean {
+    const key = `${message.sessionId}:${message.localId}`
+    const now = Date.now()
+    if (this.seen.has(key)) return false
+
+    this.seen.set(key, now)
+    if (this.seen.size > this.dedupMaxEntries) this.evict(now)
+    return true
+  }
+
+  /**
+   * 淘汰策略：先按 TTL 清理过期条目；若仍然超限，再按插入顺序丢掉最旧的一批
+   * （Map 保持插入顺序，所以从头删就是删最旧）。
+   */
+  private evict(now: number): void {
+    for (const [key, seenAt] of this.seen) {
+      if (now - seenAt > this.dedupTtlMs) this.seen.delete(key)
+    }
+    if (this.seen.size <= this.dedupMaxEntries) return
+    const overflow = this.seen.size - this.dedupMaxEntries
+    let removed = 0
+    for (const key of this.seen.keys()) {
+      this.seen.delete(key)
+      removed += 1
+      if (removed >= overflow) break
+    }
+  }
+}
+
+/**
+ * NEXT STEPS（不在本轮范围）
+ *
+ * 1. **会话定位（BLOCKER）**：目前只能回读「最近活跃会话」。要真正监听全部会话，
+ *    需要解决「从 table event 推断变更会话」。可能的路线：
+ *    - 观察 native 侧能否在事件里补上 table 所属的会话标识（需改 dll）；
+ *    - 或改成按 `Session` 表的 `last_timestamp` 变化做「有界的多会话回读」。
+ * 2. **多人 @**：`extractMentionTargets` 目前按单值返回，多人格式**未验证**，不得猜分隔符。
+ * 3. **isMentionedMe**：`mentionTargets` ∩ `getMyUsernameCandidates()`，再接 Trigger。
+ * 4. 本轮**不做**：关键词 / @触发 / 自动回复 / 日报 / AI / Agent / 发送。
+ */
@@ -7,15 +7,28 @@ import {
  personalWechatSendService,
  type PersonalWechatSendService
 } from './personal-wechat-send-service'
+import { macWechatRuntimeManager } from './mac-wechat-runtime-manager'

 /**
 * Converts the detailed sender diagnostics into a small contract that other
- * features can consume without knowing about OneBot, Hook or platform details.
+ * features can consume without knowing about the macOS native runtime, the
+ * Windows hook transport, or platform details.
+ *
+ * On macOS the capability is derived only from MacWechatRuntimeManager and its
+ * validated runtime manifest. Windows continues to use the existing sender.
 */
 export class PersonalWechatCapabilityService {
-  constructor(private readonly sender: Pick<PersonalWechatSendService, 'getStatus'>) {}
+  constructor(
+    private readonly sender: Pick<PersonalWechatSendService, 'getStatus'>,
+    private readonly platform: NodeJS.Platform = process.platform,
+    private readonly getMacStatus: () => Promise<PersonalWechatSenderStatus> = () =>
+      macWechatRuntimeManager.buildSenderStatus()
+  ) {}

  async getPersonalWechatSendCapability(): Promise<PersonalWechatSendCapability> {
+    if (this.platform === 'darwin') {
+      return this.fromMacRuntime()
+    }
    const senderStatus = await this.sender.getStatus()
    return this.fromSenderStatus(senderStatus)
  }
@@ -49,6 +62,16 @@ export class PersonalWechatCapabilityService {
    }
  }

+  /*
+   * macOS capability from the native runtime. Builds a senderStatus snapshot
+   * aligned with the mac binding state and reuses the shared mapping, so the
+   * settings header and the binding card can never disagree.
+   */
+  private async fromMacRuntime(): Promise<PersonalWechatSendCapability> {
+    const senderStatus = await this.getMacStatus()
+    return this.fromSenderStatus(senderStatus)
+  }
+
  private mapState(senderStatus: PersonalWechatSenderStatus): PersonalWechatSendCapabilityState {
    if (senderStatus.platform === 'win32') {
      if (senderStatus.canSend) return 'ready'
@@ -60,6 +83,7 @@ export class PersonalWechatCapabilityService {
      return 'unsupported'
    }
    if (senderStatus.state === 'error') return 'error'
+    if (senderStatus.state === 'runtime_missing') return 'unconfigured'
    const hasCurrentBinding = Boolean(
      senderStatus.endpointReady &&
      senderStatus.attachReady &&
@@ -1,603 +0,0 @@
-import { createHash } from 'crypto'
-import { execFile } from 'child_process'
-import { app, net } from 'electron'
-import { existsSync, readFileSync, writeFileSync } from 'fs'
-import { chmod, copyFile, cp, mkdir, mkdtemp, open, rename, rm, stat } from 'fs/promises'
-import { tmpdir } from 'os'
-import { dirname, join } from 'path'
-import { promisify } from 'util'
-import type {
-  PersonalWechatRuntimeDownloadResult,
-  PersonalWechatRuntimeStatus
-} from '../../shared/personal-wechat-runtime'
-import { findPersonalWechatRuntime } from './personal-wechat-send-service'
-
-const execFileAsync = promisify(execFile)
-const RUNTIME_VERSION = 'v0.0.18'
-const ARCHIVE_NAME = 'onebot_mac_arm64.tar.gz'
-const ARCHIVE_URL = `https://github.com/yincongcyincong/wechat_chatter/releases/download/${RUNTIME_VERSION}/${ARCHIVE_NAME}`
-const ARCHIVE_SIZE = 66_599_785
-const ARCHIVE_SHA256 = 'ee1e11bccef7cec1cf944cd8b2ac3fadaadb9376ba24cd823e3409143e107dab'
-
-function patchPerSendPayload(scriptPath: string): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (!source.includes('var activeTriggerX1Payload = ptr(0);')) return
-
-  const activeSend = `    const payloadData = hexToByteArray(payloadHex);
-    activeTriggerX1Payload = Memory.alloc(payloadData.length);
-    activeTriggerX1Payload.writeByteArray(payloadData);
-    activeTriggerX1Payload.add(0x18).writePointer(info.cgiAddr);
-    activeTriggerX1Payload.add(0xb8).writePointer(activeTriggerX1Payload.add(0xc0));
-    activeTriggerX1Payload.add(0x190).writePointer(activeTriggerX1Payload.add(0x198));`
-  const upstreamSend = `    const payloadData = hexToByteArray(payloadHex);
-    triggerX1Payload.writeByteArray(payloadData);
-    triggerX1Payload.add(0x18).writePointer(info.cgiAddr);
-    triggerX1Payload.add(0xb8).writePointer(triggerX1Payload.add(0xc0));
-    triggerX1Payload.add(0x190).writePointer(triggerX1Payload.add(0x198));`
-  if (!source.includes(activeSend)) {
-    throw new Error('下载的发送组件与当前应用不兼容')
-  }
-  source = source.replace('var triggerX1Payload;\nvar activeTriggerX1Payload = ptr(0);\nvar triggerX0;', 'var triggerX1Payload;\nvar triggerX0;')
-    .replace(activeSend, upstreamSend)
-    .replace('        MMStartTask(triggerX0, activeTriggerX1Payload);', '        MMStartTask(triggerX0, triggerX1Payload);')
-    .replace('        activeTriggerX1Payload = ptr(0);\n        console.error("[!] Error trigger " + msgType + " MMStartTask: " + e);', '        console.error("[!] Error trigger " + msgType + " MMStartTask: " + e);')
-    .replace('\t\t\t\tpendingSendMsgType = "";\n\t\t\t\tactiveTriggerX1Payload = ptr(0);\n\t\t\t\treturn', '\t\t\t\tpendingSendMsgType = "";\n\t\t\t\treturn')
-  writeFileSync(scriptPath, source)
-}
-
-function patchSendContextCapture(scriptPath: string, strict = true): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (source.includes('function isLikelySendContext(')) return
-  const original = `function AttachSendFunc() {
-    Interceptor.attach(sendFuncAddr.add(0x10), {
-        onEnter: function (args) {
-
-            if (triggerX1Payload) {
-                return
-            }
-
-            triggerX0 = this.context.x0;
-            triggerX1Payload = this.context.x1;
-            console.log(\`[+] 捕获到 StartTask 调用，X0：\${triggerX0}, Payload: \${triggerX1Payload}\`);
-        }
-    })
-}`
-  const patched = `function isLikelySendContext(candidateX0, candidateX1) {
-    try {
-        if (!isReadablePointer(candidateX0) || !isReadablePointer(candidateX1)) return false;
-        var manager = readPointerIfReadable(candidateX0.add(0x18));
-        var cgi = readUtf8StringIfReadable(readPointerIfReadable(candidateX1.add(0x18)));
-        console.log("[debug] StartTask candidate x0=" + candidateX0 + " x1=" + candidateX1 + " x0+0x18=" + manager + " cgi=" + cgi);
-        return !manager.equals(ptr(0));
-    } catch (e) {
-        console.error("[debug] StartTask candidate inspect failed: " + e);
-        return false;
-    }
-}
-
-function AttachSendFunc() {
-    Interceptor.attach(sendFuncAddr.add(0x10), {
-        onEnter: function (args) {
-            if (triggerX1Payload) return;
-            var candidateX0 = this.context.x0;
-            var candidateX1 = this.context.x1;
-            if (!isLikelySendContext(candidateX0, candidateX1)) return;
-            triggerX0 = candidateX0;
-            triggerX1Payload = candidateX1;
-            console.log(\`[+] 捕获到有效 StartTask 上下文，X0：\${triggerX0}, Payload: \${triggerX1Payload}\`);
-        }
-    })
-}`
-  if (!source.includes(original)) {
-    if (strict) throw new Error('下载的微信版本与当前应用不兼容')
-    return
-  }
-  source = source.replace(original, patched)
-  writeFileSync(scriptPath, source)
-}
-
-function patchVoiceAudioBuffer(scriptPath: string, strict = true): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (source.includes('voiceAudioDataAddr = Memory.alloc(audioLen + 1);')) return
-
-  const staticAllocation = 'voiceAudioDataAddr = Memory.alloc(5 * 1024 * 1024); // 预分配5MB'
-  if (!source.includes(staticAllocation)) {
-    if (strict) throw new Error('下载的语音组件与当前应用不兼容')
-    return
-  }
-  source = source.replace(
-    staticAllocation,
-    'voiceAudioDataAddr = Memory.alloc(1); // 上传前按语音长度重新分配'
-  )
-  const audioLengthMarker = '    const audioLen = audioBytes.length;\n'
-  if (!source.includes(audioLengthMarker)) {
-    if (strict) throw new Error('下载的语音组件与当前应用不兼容')
-    return
-  }
-  source = source.replace(
-    audioLengthMarker,
-    `${audioLengthMarker}    voiceAudioDataAddr = Memory.alloc(audioLen + 1);\n`
-  )
-  writeFileSync(scriptPath, source)
-}
-
-function patchImageHookReadiness(scriptPath: string): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (
-    source.includes('捕获到图片上传上下文，uploadGlobalX0') &&
-    source.includes('图片上传 Hook Setup Complete')
-  ) {
-    return
-  }
-  const original = `\t\t\tuploadGlobalX0 = this.context.x0;`
-  const patched = `\t\t\tconst capturedUploadX0 = this.context.x0;
-\t\t\tif (uploadGlobalX0.equals(ptr(0)) && !capturedUploadX0.equals(ptr(0))) {
-\t\t\t\tconsole.log("[+] 捕获到图片上传上下文，uploadGlobalX0：" + capturedUploadX0);
-\t\t\t}
-\t\t\tuploadGlobalX0 = capturedUploadX0;`
-  if (!source.includes(original)) throw new Error('下载的媒体组件与当前应用不兼容')
-  source = source.replace(original, patched)
-  source = source.replace(
-    '    })\n}\n\n\n\nfunction patchCdnOnComplete()',
-    '    })\n    console.log("[+] 图片上传 Hook Setup Complete.");\n}\n\n\n\nfunction patchCdnOnComplete()'
-  )
-  writeFileSync(scriptPath, source)
-}
-
-function patchCdnColdStart(scriptPath: string, strict = true): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (source.includes('function resolveCdnManager()')) return
-
-  const initAddresses = `    uploadImageAddr = baseAddr.add({{.uploadImageAddr}});
-    cndOnCompleteAddr = baseAddr.add({{.cndOnCompleteAddr}});`
-  const patchedInitAddresses = `    uploadImageAddr = baseAddr.add({{.uploadImageAddr}});
-    cndOnCompleteAddr = baseAddr.add({{.cndOnCompleteAddr}});
-    // 冷启动 CdnManager 解析（旧版本缺少可选键时保持 hook 捕获行为）
-    {{if .cdnGetServiceAddr}}cdnGetServiceAddr = baseAddr.add({{.cdnGetServiceAddr}});{{end}}
-    {{if .cdnManagerGetterAddr}}cdnManagerGetterAddr = baseAddr.add({{.cdnManagerGetterAddr}});{{end}}`
-  if (!source.includes(initAddresses)) {
-    if (strict) throw new Error('下载的微信版本配置与当前应用不兼容')
-    return
-  }
-  source = source.replace(initAddresses, patchedInitAddresses)
-
-  const downloadChunkEnd = `}
-
-function fillUploadX1AndStart`
-  const resolver = `}
-
-// 上传和下载共用同一个 mars::cdn::CdnManager。冷启动时通过服务定位器
-// 取得 [ctx + 0x40]，避免必须先手动发送图片才能让 Hook 捕获上下文。
-function resolveCdnManager() {
-    if (cdnGetServiceAddr.equals(ptr(0)) || cdnManagerGetterAddr.equals(ptr(0))) {
-        return ptr(0);
-    }
-    try {
-        // libc++ SSO 短字符串：数据在 +0，长度写在 +0x17。
-        var strDefault = Memory.alloc(24);
-        strDefault.writeUtf8String("default");
-        strDefault.add(0x17).writeU8(7);
-
-        var getService = new NativeFunction(cdnGetServiceAddr, 'pointer', ['pointer']);
-        var svc = getService(strDefault);
-        if (!isReadablePointer(svc)) {
-            console.error("[!] GetService(\\"default\\") 返回不可读: " + svc);
-            return ptr(0);
-        }
-        var getCtx = new NativeFunction(cdnManagerGetterAddr, 'pointer', ['pointer']);
-        var ctx = getCtx(svc);
-        if (!isReadablePointer(ctx)) {
-            console.error("[!] CdnManager getter 返回不可读: " + ctx);
-            return ptr(0);
-        }
-        var mgr = readPointerIfReadable(ctx.add(0x40));
-        if (!isReadablePointer(mgr)) {
-            console.error("[!] ctx+0x40 管理器指针不可读: ctx=" + ctx);
-            return ptr(0);
-        }
-        return mgr;
-    } catch (e) {
-        console.error("[!] resolveCdnManager 异常: " + e);
-        return ptr(0);
-    }
-}
-
-function ensureCdnManagerX0() {
-    if (uploadGlobalX0.equals(ptr(0)) && downloadGlobalX0 && !downloadGlobalX0.equals(ptr(0))) {
-        uploadGlobalX0 = downloadGlobalX0;
-        console.log("[+] downloadGlobalX0 回填 uploadGlobalX0: " + uploadGlobalX0);
-    }
-    if ((!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) && !uploadGlobalX0.equals(ptr(0))) {
-        downloadGlobalX0 = uploadGlobalX0;
-        console.log("[+] uploadGlobalX0 回填 downloadGlobalX0: " + downloadGlobalX0);
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {
-        var mgr = resolveCdnManager();
-        if (!mgr.equals(ptr(0))) {
-            uploadGlobalX0 = mgr;
-            if (!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) {
-                downloadGlobalX0 = mgr;
-            }
-            console.log("[+] 冷启动服务定位器解析 CdnManager: " + mgr);
-        }
-    }
-    return !uploadGlobalX0.equals(ptr(0));
-}
-
-function fillUploadX1AndStart`
-  if (!source.includes(downloadChunkEnd)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 媒体上传逻辑')
-    return
-  }
-  source = source.replace(downloadChunkEnd, resolver)
-
-  const declarations = 'var uploadImageAddr;\n'
-  const patchedDeclarations =
-    'var uploadImageAddr;\nvar cdnGetServiceAddr = ptr(0);\nvar cdnManagerGetterAddr = ptr(0);\n'
-  if (!source.includes(declarations)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 媒体地址声明')
-    return
-  }
-  source = source.replace(declarations, patchedDeclarations)
-
-  const uploadGuard = `function fillUploadX1AndStart(idAddr, pathAddr, x1Buffer, receiver, md5, filePath, payloadHex) {
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  const patchedUploadGuard = `function fillUploadX1AndStart(idAddr, pathAddr, x1Buffer, receiver, md5, filePath, payloadHex) {
-    if (uploadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  if (!source.includes(uploadGuard)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 媒体上传入口')
-    return
-  }
-  source = source.replace(uploadGuard, patchedUploadGuard)
-
-  const voiceGuard = `function triggerUploadVoice(receiver, voicePath, payloadHex, audioDataHex, durationMs) {
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  const patchedVoiceGuard = `function triggerUploadVoice(receiver, voicePath, payloadHex, audioDataHex, durationMs) {
-    if (uploadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (uploadGlobalX0.equals(ptr(0))) {`
-  if (!source.includes(voiceGuard)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 语音上传入口')
-    return
-  }
-  source = source.replace(voiceGuard, patchedVoiceGuard)
-
-  const uploadHook = `\t\t\tuploadGlobalX0 = capturedUploadX0;`
-  const patchedUploadHook = `\t\t\tuploadGlobalX0 = capturedUploadX0;
-            if ((!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) && !capturedUploadX0.equals(ptr(0))) {
-                downloadGlobalX0 = capturedUploadX0;
-                console.log("[+] 上传hook回填 downloadGlobalX0: " + downloadGlobalX0);
-            }`
-  if (!source.includes(uploadHook)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 图片 Hook')
-    return
-  }
-  source = source.replace(uploadHook, patchedUploadHook)
-
-  const downloadHook = `            downloadGlobalX0 = this.context.x0;`
-  const patchedDownloadHook = `            downloadGlobalX0 = this.context.x0;
-            if (uploadGlobalX0.equals(ptr(0)) && !downloadGlobalX0.equals(ptr(0))) {
-                uploadGlobalX0 = downloadGlobalX0;
-                console.log("[+] 下载hook回填 uploadGlobalX0: " + uploadGlobalX0);
-            }`
-  if (!source.includes(downloadHook)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 下载 Hook')
-    return
-  }
-  source = source.replace(downloadHook, patchedDownloadHook)
-
-  const downloadGuard = `function triggerDownload(receiver, cdnUrl, aesKey, filePath, fileType) {
-    if (!downloadGlobalX0) {`
-  const patchedDownloadGuard = `function triggerDownload(receiver, cdnUrl, aesKey, filePath, fileType) {
-    if (!downloadGlobalX0 || downloadGlobalX0.equals(ptr(0))) {
-        ensureCdnManagerX0();
-    }
-    if (!downloadGlobalX0) {`
-  if (!source.includes(downloadGuard)) {
-    if (strict) throw new Error('无法定位 wechat_chatter 媒体下载入口')
-    return
-  }
-  source = source.replace(downloadGuard, patchedDownloadGuard)
-
-  writeFileSync(scriptPath, source)
-}
-
-function patchCdnColdStartConfig(configPath: string, strict = true): void {
-  let config: Record<string, unknown>
-  try {
-    config = JSON.parse(readFileSync(configPath, 'utf8')) as Record<string, unknown>
-  } catch {
-    if (strict) throw new Error('4.1.11.53 版本配置不是有效 JSON')
-    return
-  }
-  config.cdnGetServiceAddr = '0x50a15d0'
-  config.cdnManagerGetterAddr = '0x5259290'
-  writeFileSync(configPath, `${JSON.stringify(config, null, 2)}\n`)
-}
-
-function patchWechatCoreModuleBase(scriptPath: string): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (source.includes('WeChat core module base:')) return
-  const initMarker = 'function initAddresses() {'
-  const initIndex = source.indexOf(initMarker)
-  if (initIndex < 0 || !source.startsWith('var targetPath = ')) {
-    throw new Error('下载的微信版本配置与当前应用不兼容')
-  }
-  const patchedHeader = `var targetPath = "/Applications/WeChat.app/Contents/Resources/wechat.dylib";
-var module = Process.enumerateModules().find(function(m) {
-    return m.path === targetPath || m.path.endsWith("/Contents/Resources/wechat.dylib");
-});
-if (!module) {
-    throw new Error("[-] Cannot find WeChat core module: " + targetPath);
-}
-var moduleBase = module.base;
-var baseAddr = moduleBase;
-console.log("[+] WeChat core module base: " + baseAddr + " path=" + module.path);
-setImmediate(initAddresses);
-
-`
-  source = patchedHeader + source.slice(initIndex)
-  writeFileSync(scriptPath, source)
-}
-
-function addModifiedWorkNotice(scriptPath: string): void {
-  let source = readFileSync(scriptPath, 'utf8')
-  if (source.includes('TraceMemo wechat_chatter compatibility modifications')) return
-  source = `/*
- * TraceMemo wechat_chatter compatibility modifications
- * Modified: 2026-08-17
- * Upstream: https://github.com/yincongcyincong/wechat_chatter
- * Runtime version: v0.0.18
- * License: GNU General Public License version 3 (GPL-3.0)
- * Changes: WeChat module discovery, per-send payload isolation, dynamic voice upload buffers,
- * CdnManager cold-start resolution, media hook backfill, and image Hook readiness logging.
- * These modifications are not provided by the upstream author.
- */
-
-${source}`
-  writeFileSync(scriptPath, source)
-}
-
-export class PersonalWechatRuntimeManager {
-  private downloadController: AbortController | null = null
-  private downloadPromise: Promise<PersonalWechatRuntimeDownloadResult> | null = null
-  private downloadedBytes = 0
-  private lastProgressAt = 0
-  private progressListener: ((status: PersonalWechatRuntimeStatus) => void) | null = null
-
-  get directory(): string {
-    return join(app.getPath('userData'), 'connectors', 'wechat-personal', 'darwin-arm64')
-  }
-
-  private get archivePath(): string {
-    return join(
-      app.getPath('userData'),
-      'downloads',
-      `wechat-chatter-${RUNTIME_VERSION}-${ARCHIVE_NAME}`
-    )
-  }
-
-  setProgressListener(listener: ((status: PersonalWechatRuntimeStatus) => void) | null): void {
-    this.progressListener = listener
-  }
-
-  async getStatus(): Promise<PersonalWechatRuntimeStatus> {
-    if (!this.isSupported()) {
-      return this.buildStatus(
-        'unsupported',
-        0,
-        process.platform === 'darwin'
-          ? 'Intel Mac 不支持个人微信发送组件'
-          : process.platform === 'win32'
-            ? 'Windows 暂不支持个人微信发送组件'
-            : '当前系统不支持个人微信发送组件'
-      )
-    }
-    if (this.downloadPromise) return this.buildStatus('downloading', this.downloadedBytes)
-
-    const runtime = findPersonalWechatRuntime()
-    if (runtime) {
-      try {
-        // Apply compatibility fixes to runtimes installed before this version.
-        patchWechatCoreModuleBase(join(runtime.workingDirectory, 'script.js'))
-        patchPerSendPayload(join(runtime.workingDirectory, 'script.js'))
-        patchSendContextCapture(join(runtime.workingDirectory, 'script.js'), false)
-        patchVoiceAudioBuffer(join(runtime.workingDirectory, 'script.js'), false)
-        patchImageHookReadiness(join(runtime.workingDirectory, 'script.js'))
-        patchCdnColdStart(join(runtime.workingDirectory, 'script.js'), false)
-        addModifiedWorkNotice(join(runtime.workingDirectory, 'script.js'))
-        patchCdnColdStartConfig(join(runtime.root, 'wechat_version', '4_1_11_53_mac.json'), false)
-      } catch {
-        // Status discovery should remain available even if an old runtime is read-only.
-      }
-      return this.buildStatus('ready', ARCHIVE_SIZE, undefined, runtime.root)
-    }
-
-    const hasPartialInstall = await this.hasPartialInstall()
-    return this.buildStatus(
-      hasPartialInstall ? 'invalid' : 'missing',
-      0,
-      hasPartialInstall ? '发送组件文件不完整，请重新下载' : undefined,
-      hasPartialInstall ? this.directory : undefined
-    )
-  }
-
-  download(): Promise<PersonalWechatRuntimeDownloadResult> {
-    if (!this.isSupported()) {
-      return this.getStatus().then((status) => ({ success: false, status, error: status.error }))
-    }
-    if (this.downloadPromise) return this.downloadPromise
-    this.downloadedBytes = 0
-    this.downloadController = new AbortController()
-    this.downloadPromise = this.runDownload(this.downloadController.signal).finally(() => {
-      this.downloadPromise = null
-      this.downloadController = null
-    })
-    return this.downloadPromise
-  }
-
-  cancelDownload(): boolean {
-    if (!this.downloadController) return false
-    this.downloadController.abort()
-    return true
-  }
-
-  async remove(): Promise<PersonalWechatRuntimeStatus> {
-    if (this.downloadPromise) return this.buildStatus('downloading', this.downloadedBytes)
-    await Promise.all([
-      rm(this.directory, { recursive: true, force: true }),
-      rm(this.archivePath, { force: true }),
-      rm(`${this.archivePath}.partial`, { force: true })
-    ])
-    return this.getStatus()
-  }
-
-  private async runDownload(signal: AbortSignal): Promise<PersonalWechatRuntimeDownloadResult> {
-    const archive = this.archivePath
-    const downloadsDirectory = dirname(archive)
-    const partial = `${archive}.partial`
-    let extractionDirectory = ''
-    let stagedDirectory = ''
-    try {
-      await mkdir(downloadsDirectory, { recursive: true })
-      await rm(partial, { force: true })
-      const response = await net.fetch(ARCHIVE_URL, { signal })
-      if (!response.ok || !response.body) {
-        throw new Error(`发送组件下载失败：HTTP ${response.status}`)
-      }
-
-      const handle = await open(partial, 'w')
-      const hash = createHash('sha256')
-      try {
-        const reader = response.body.getReader()
-        while (true) {
-          const { done, value } = await reader.read()
-          if (done) break
-          if (signal.aborted) throw new DOMException('Download cancelled', 'AbortError')
-          const chunk = Buffer.from(value)
-          await handle.write(chunk)
-          hash.update(chunk)
-          this.downloadedBytes += chunk.length
-          this.reportProgress(this.buildStatus('downloading', this.downloadedBytes))
-        }
-      } finally {
-        await handle.close()
-      }
-
-      if (this.downloadedBytes !== ARCHIVE_SIZE || hash.digest('hex') !== ARCHIVE_SHA256) {
-        throw new Error('发送组件校验失败，请重新下载')
-      }
-      await rm(archive, { force: true })
-      await rename(partial, archive)
-
-      extractionDirectory = await mkdtemp(join(tmpdir(), 'wechat-chatter-extract-'))
-      await execFileAsync('/usr/bin/tar', ['-xzf', archive, '-C', extractionDirectory])
-      stagedDirectory = `${this.directory}.installing-${process.pid}`
-      await rm(stagedDirectory, { recursive: true, force: true })
-      await mkdir(dirname(stagedDirectory), { recursive: true })
-      await mkdir(join(stagedDirectory, 'onebot'), { recursive: true })
-
-      const sourceOneBot = join(extractionDirectory, 'onebot')
-      const sourceVersions = join(extractionDirectory, 'wechat_version')
-      await Promise.all([
-        cp(sourceVersions, join(stagedDirectory, 'wechat_version'), {
-          recursive: true,
-          force: true
-        }),
-        copyFile(join(sourceOneBot, 'onebot'), join(stagedDirectory, 'onebot', 'onebot')),
-        copyFile(join(sourceOneBot, 'script.js'), join(stagedDirectory, 'onebot', 'script.js'))
-      ])
-
-      const executable = join(stagedDirectory, 'onebot', 'onebot')
-      const script = join(stagedDirectory, 'onebot', 'script.js')
-      const configPath = join(stagedDirectory, 'wechat_version', '4_1_11_53_mac.json')
-      patchCdnColdStartConfig(configPath)
-      patchWechatCoreModuleBase(script)
-      patchPerSendPayload(script)
-      patchSendContextCapture(script)
-      patchVoiceAudioBuffer(script)
-      patchImageHookReadiness(script)
-      patchCdnColdStart(script)
-      addModifiedWorkNotice(script)
-      await chmod(executable, 0o755)
-
-      for (const required of [
-        executable,
-        script,
-        join(stagedDirectory, 'wechat_version', '4_1_11_53_mac.json')
-      ]) {
-        if (!existsSync(required)) throw new Error('发送组件解压后文件不完整')
-      }
-
-      await rm(this.directory, { recursive: true, force: true })
-      await rename(stagedDirectory, this.directory)
-      stagedDirectory = ''
-      const status = this.buildStatus('ready', ARCHIVE_SIZE, undefined, this.directory)
-      this.reportProgress(status, true)
-      return { success: true, status }
-    } catch (error) {
-      const cancelled = signal.aborted
-      const message = cancelled
-        ? '发送组件下载已取消'
-        : error instanceof Error
-          ? error.message
-          : String(error)
-      await rm(partial, { force: true })
-      const status = this.buildStatus(
-        cancelled ? 'missing' : 'error',
-        this.downloadedBytes,
-        message
-      )
-      this.reportProgress(status, true)
-      return { success: false, status, error: message }
-    } finally {
-      if (extractionDirectory) await rm(extractionDirectory, { recursive: true, force: true })
-      if (stagedDirectory) await rm(stagedDirectory, { recursive: true, force: true })
-    }
-  }
-
-  private async hasPartialInstall(): Promise<boolean> {
-    try {
-      await stat(this.directory)
-      return true
-    } catch {
-      return false
-    }
-  }
-
-  private isSupported(): boolean {
-    return process.platform === 'darwin' && process.arch === 'arm64'
-  }
-
-  private buildStatus(
-    state: PersonalWechatRuntimeStatus['state'],
-    downloadedBytes: number,
-    error?: string,
-    directory?: string
-  ): PersonalWechatRuntimeStatus {
-    return {
-      version: RUNTIME_VERSION,
-      state,
-      downloadedBytes,
-      totalBytes: ARCHIVE_SIZE,
-      progress: ARCHIVE_SIZE ? Math.min(1, downloadedBytes / ARCHIVE_SIZE) : 0,
-      platform: process.platform,
-      architecture: process.arch,
-      supported: this.isSupported(),
-      removable: directory === this.directory,
-      ...(directory ? { directory } : {}),
-      ...(error ? { error } : {})
-    }
-  }
-
-  private reportProgress(status: PersonalWechatRuntimeStatus, force = false): void {
-    const now = Date.now()
-    if (!force && now - this.lastProgressAt < 100) return
-    this.lastProgressAt = now
-    this.progressListener?.(status)
-  }
-}
@@ -7,11 +7,8 @@ import type {
  PersonalWechatVoiceRuntimeComponent
 } from '../../shared/personal-wechat-voice-runtime'
 import { PERSONAL_WECHAT_PILK_VERSION } from '../../shared/personal-wechat-voice-runtime'
-import {
-  buildPersonalWechatRuntimeEnvironment,
-  findPersonalWechatRuntime
-} from './personal-wechat-send-service'
-import type { RuntimeLayout } from './personal-wechat-send-service'
+import { buildPersonalWechatRuntimeEnvironment } from './personal-wechat-send-service'
+import { macWechatRuntimeManager } from './mac-wechat-runtime-manager'
 import { appLogger } from '../app-logger'

 const execFileAsync = promisify(execFile)
@@ -37,8 +34,8 @@ type CommandRunner = (
 interface PersonalWechatVoiceEnvironmentServiceOptions {
  platform?: NodeJS.Platform
  architecture?: string
-  findRuntime?: () => RuntimeLayout | null
-  buildEnvironment?: (runtimeRoot?: string) => NodeJS.ProcessEnv
+  isRuntimePresent?: () => boolean
+  buildEnvironment?: () => NodeJS.ProcessEnv
  runCommand?: CommandRunner
  now?: () => Date
 }
@@ -100,18 +97,26 @@ function unsupportedEnvironment(): PersonalWechatVoiceEncodingEnvironment {
  }
 }

+function defaultIsRuntimePresent(): boolean {
+  try {
+    return macWechatRuntimeManager.isRuntimePresent()
+  } catch {
+    return false
+  }
+}
+
 export class PersonalWechatVoiceEnvironmentService {
  private readonly platform: NodeJS.Platform
  private readonly architecture: string
-  private readonly findRuntime: () => RuntimeLayout | null
-  private readonly buildEnvironment: (runtimeRoot?: string) => NodeJS.ProcessEnv
+  private readonly isRuntimePresent: () => boolean
+  private readonly buildEnvironment: () => NodeJS.ProcessEnv
  private readonly runCommand: CommandRunner
  private readonly now: () => Date

  constructor(options: PersonalWechatVoiceEnvironmentServiceOptions = {}) {
    this.platform = options.platform || process.platform
    this.architecture = options.architecture || process.arch
-    this.findRuntime = options.findRuntime || findPersonalWechatRuntime
+    this.isRuntimePresent = options.isRuntimePresent || defaultIsRuntimePresent
    this.buildEnvironment = options.buildEnvironment || buildPersonalWechatRuntimeEnvironment
    this.runCommand = options.runCommand || runCommand
    this.now = options.now || (() => new Date())
@@ -123,8 +128,8 @@ export class PersonalWechatVoiceEnvironmentService {
    }

    logEnvironmentLine('Checking voice encoding environment')
-    const runtime = this.findRuntime()
-    const environment = this.buildEnvironment(runtime?.root)
+    const runtimeReady = this.isRuntimePresent()
+    const environment = this.buildEnvironment()
    const python = blankComponent()
    const pilk = blankComponent()
    const ffmpeg = blankComponent()
@@ -197,26 +202,25 @@ export class PersonalWechatVoiceEnvironmentService {
      logEnvironmentWarning('ffmpeg: unavailable', { error: ffmpeg.error })
    }

-    const ready = Boolean(runtime && python.ready && pilk.ready && ffmpeg.ready)
+    const ready = Boolean(runtimeReady && python.ready && pilk.ready && ffmpeg.ready)
    const result: PersonalWechatVoiceEncodingEnvironment = {
      state: ready ? 'ready' : 'incomplete',
      ready,
      checkedAt: this.now().toISOString(),
-      runtimeReady: Boolean(runtime),
-      ...(runtime ? { runtimeRoot: runtime.root } : {}),
+      runtimeReady,
      python,
      pilk,
      ffmpeg,
      encoder: pilk.ready
        ? 'pilk'
-        : runtime && python.ready && ffmpeg.ready
-          ? 'go-silk'
+        : runtimeReady && python.ready && ffmpeg.ready
+          ? 'silk'
          : 'unavailable',
      message: ready
        ? '语音编码环境正常，可以使用 pilk 编码'
-        : runtime
-          ? '语音编码环境不完整，OneBot 可能回退到 go-silk'
-          : '微信发送组件尚未安装，请先准备 OneBot 运行时'
+        : runtimeReady
+          ? '语音编码环境不完整，语音将回退到内置 SILK 编码'
+          : '当前版本暂未提供微信消息发送功能'
    }
    logEnvironmentLine(
      ready ? 'Voice encoding environment is ready' : 'Voice encoding environment is NOT ready',
@@ -245,7 +249,7 @@ export class PersonalWechatVoiceEnvironmentService {
        pythonExecutable,
        ['-m', 'pip', 'install', '--user', `pilk==${PERSONAL_WECHAT_PILK_VERSION}`],
        {
-          env: this.buildEnvironment(before.runtimeRoot || undefined),
+          env: this.buildEnvironment(),
          timeout: INSTALL_TIMEOUT_MS
        }
      )
@@ -27,8 +27,7 @@ export interface ScheduledReportApiDependencies {
    | 'deleteTask'
    | 'setTaskEnabled'
    | 'runScheduledReportNow'
-  > &
-    Partial<Pick<ScheduledReportService, 'retryScheduledReportSend'>>
+  >
  getCapability: () => Promise<PersonalWechatSendCapability>
  listContacts: () => FormattedContact[]
  isDatabaseReady: () => boolean
@@ -250,16 +249,20 @@ export class ScheduledReportApiService {
    return result.data
  }

-  async retrySend(executionId: string): Promise<ScheduledReportExecution> {
-    const retry = this.deps.service.retryScheduledReportSend
-    if (!retry) {
-      throw new ScheduledReportApiError(501, 'not_supported', '当前运行时不支持重新发送日报')
-    }
-    const result = await retry.call(this.deps.service, executionId)
-    if (!result.data) {
-      throw new ScheduledReportApiError(404, 'not_found', result.error || '未找到定时日报执行记录')
-    }
-    return result.data
+  /**
+   * 「重新发送」不再支持。
+   *
+   * 日报图片归日报历史所有，发送目标由规则配置决定，「把某次执行记录里那张 PNG
+   * 就地再发一遍」这个语义已经不存在 —— 与其在此伪造一个只对部分规则有效的重发
+   * 语义，不如**明确返回 501**。
+   * 用户的替代路径：在日报历史里手动转发，或对该规则点「立即执行」重跑一次。
+   */
+  async retrySend(_executionId: string): Promise<ScheduledReportExecution> {
+    throw new ScheduledReportApiError(
+      501,
+      'not_supported',
+      '当前运行时不支持重新发送日报；请在 TraceMemo 的自动化页面使用「立即执行」，或从日报历史手动转发。'
+    )
  }

  async executions(taskId: string): Promise<ScheduledReportApiExecution[]> {
@@ -0,0 +1,207 @@
+import { createHash } from 'node:crypto'
+import {
+  normalizeScheduledReportConfig,
+  type ScheduledReportAutomationConfig
+} from '../../shared/automation'
+import type {
+  ScheduledReportMemberNameMode,
+  ScheduledReportMessageType,
+  ScheduledReportRange
+} from '../../shared/scheduled-report'
+import type { SelectableReportTemplateId } from '../../shared/report-templates'
+
+/**
+ * 旧「定时日报任务」→ 新「AutomationRule(ruleType='scheduled_report')」的**纯映射器**。
+ *
+ * 单独成文件是为了能直接单测：这是整轮迁移里**唯一会改变用户既有行为**的地方，
+ * 不允许只靠"跑一遍看看"来验证。
+ *
+ * ## 旧 target 的真实语义（实测）
+ *
+ * `ScheduledReportService.resolveTarget()`：
+ * - `target` 以 `@chatroom` 结尾 ⇒ **直接发到那个群**（可能是另一个群）；
+ * - 否则 ⇒ **静默回落到来源群**。
+ *
+ * ⇒ 历史行为**事实上支持「生成 A 群日报 → 发 B 群」**，也存在一处静默 fallback。
+ * 新产品明确不支持"指定另一个群聊"，所以：
+ * **能证明同群 → 无损映射成 `source_chat`；不能证明 → 停用 + 要求用户重选**，
+ * 绝不自动改写、绝不偷偷丢掉目标。
+ */
+
+/** 旧任务里迁移真正用得到的字段（其余字段不参与，也不需要读）。 */
+export interface LegacyScheduledReportTask {
+  id?: unknown
+  name?: unknown
+  group?: unknown
+  scheduleTime?: unknown
+  reportRange?: unknown
+  messageTypes?: unknown
+  templateId?: unknown
+  memberNameMode?: unknown
+  timeoutSeconds?: unknown
+  target?: unknown
+  enabled?: unknown
+  createdAt?: unknown
+  updatedAt?: unknown
+  lastRunAt?: unknown
+  lastScheduledSlot?: unknown
+}
+
+export type ScheduledReportMigrationOutcome =
+  /** 目标与来源群是同一群 ⇒ 等价于 `source_chat`，行为完全一致。 */
+  | 'lossless_source_chat'
+  /** 目标是另一个群 / 无法确认是否同群 ⇒ 单值四选一表达不了，需用户重选。 */
+  | 'needs_review'
+
+export interface ScheduledReportMigrationPlan {
+  outcome: ScheduledReportMigrationOutcome
+  /** 迁移后的规则 id（确定性，可重复执行且不产生新 id）。 */
+  ruleId: string
+  name: string
+  enabled: boolean
+  config: ScheduledReportAutomationConfig
+  createdAt: number
+  updatedAt: number
+}
+
+/** 把任意输入收敛成字符串。 */
+function asText(value: unknown): string {
+  return typeof value === 'string' ? value.trim() : String(value ?? '').trim()
+}
+
+/**
+ * 确定性 ruleId。
+ *
+ * 优先**原样复用旧 taskId**（旧 id 形如 `scheduled_report_<uuid>`，与自动化现有
+ * id 空间 —— `randomUUID()` / `builtin-*` —— 不可能冲突）。
+ * 旧 id 缺失时（文件被手改坏），退化成一个**由稳定字段派生**的 id，
+ * 保证"重复迁移不会生成新规则"。
+ */
+export function scheduledReportMigrationRuleId(task: LegacyScheduledReportTask): string {
+  const id = asText(task.id)
+  if (id) return id
+  const digest = createHash('sha1')
+    .update(
+      [asText(task.name), asText(task.group), asText(task.scheduleTime), asText(task.target)].join(
+        '\u0001'
+      )
+    )
+    .digest('hex')
+  return `scheduled-report:${digest.slice(0, 32)}`
+}
+
+/** 旧 `createdAt` / `updatedAt` 是 ISO 字符串；非法时回落到迁移时刻。 */
+function parseIso(value: unknown, fallback: number): number {
+  const text = asText(value)
+  if (!text) return fallback
+  const parsed = Date.parse(text)
+  return Number.isFinite(parsed) ? parsed : fallback
+}
+
+/**
+ * 单条旧任务 → 迁移计划。
+ *
+ * `resolveConversationId` 由调用方注入（main 侧是 `chat-service.resolveMd5`，
+ * 单测里是假的联系人表）。**它拿不到就说明主进程还没准备好**，
+ * 调用方必须**推迟整批迁移**，而不是用"解析不到"去误判成 needs_review。
+ */
+export function planScheduledReportMigration(
+  task: LegacyScheduledReportTask,
+  resolveConversationId: (raw: string) => string | undefined,
+  now: number
+): ScheduledReportMigrationPlan {
+  const sourceRaw = asText(task.group)
+  const targetRaw = asText(task.target)
+  const sourceResolved = resolveConversationId(sourceRaw)
+  const targetResolved = resolveConversationId(targetRaw)
+
+  /*
+   * 三种"能证明是同一个群"的情况：
+   * 1. target 为空 —— 旧 `normalizeInput` 本就把空 target 填成 group；
+   * 2. 两者文本完全相等 —— 无论能否解析，用户表达的就是同一个东西；
+   * 3. 两者都能解析且解析结果相同 —— 旧 UI 的常态（group=会话 md5，target=roomId）。
+   */
+  const sameGroup =
+    !targetRaw || targetRaw === sourceRaw || Boolean(sourceResolved && targetResolved && sourceResolved === targetResolved)
+
+  const config = normalizeScheduledReportConfig({
+    schedule: { time: asText(task.scheduleTime) },
+    report: {
+      // 解析得到就用稳定会话 id；解析不到**保留原值** —— 运行期会如实失败，
+      // 绝不在这里"猜一个群"。
+      sourceConversationId: sourceResolved ?? sourceRaw,
+      range: asText(task.reportRange) as ScheduledReportRange,
+      messageTypes: task.messageTypes as ScheduledReportMessageType[],
+      templateId: asText(task.templateId) as SelectableReportTemplateId,
+      memberNameMode: asText(task.memberNameMode) as ScheduledReportMemberNameMode,
+      timeoutSeconds: Number(task.timeoutSeconds)
+    },
+    target: { type: 'source_chat' },
+    ...(sameGroup ? {} : { targetNeedsReview: true }),
+    // 旧目标原文只作展示；`sameGroup` 时不保留，避免留下看起来像"还生效"的残值。
+    ...(sameGroup || !targetRaw ? {} : { legacyTarget: targetRaw }),
+    // 来源群被改写过（md5/群名 → roomId）时留下原文，便于用户核对。
+    ...(sourceRaw && sourceResolved && sourceRaw !== sourceResolved
+      ? { legacySourceGroup: sourceRaw }
+      : {}),
+    ...(asText(task.lastRunAt) ? { lastRunAt: asText(task.lastRunAt) } : {}),
+    ...(asText(task.lastScheduledSlot) ? { lastScheduledSlot: asText(task.lastScheduledSlot) } : {})
+  })
+
+  const name = asText(task.name) || '定时日报'
+  return {
+    outcome: sameGroup ? 'lossless_source_chat' : 'needs_review',
+    ruleId: scheduledReportMigrationRuleId(task),
+    name,
+    // 不自动改写、也不自动启用：无法确认目标的规则一律停用。
+    enabled: sameGroup ? task.enabled !== false : false,
+    config,
+    createdAt: parseIso(task.createdAt, now),
+    updatedAt: parseIso(task.updatedAt, now)
+  }
+}
+
+export interface ScheduledReportMigrationSummary {
+  total: number
+  migrated: number
+  lossless: number
+  needsReview: number
+  duplicatesSkipped: number
+}
+
+/**
+ * 整批任务的迁移结果：**按 id 去重**，保证幂等（同一份旧文件跑两次结果一致）。
+ */
+export function planScheduledReportMigrationBatch(
+  tasks: LegacyScheduledReportTask[],
+  resolveConversationId: (raw: string) => string | undefined,
+  now: number
+): { plans: ScheduledReportMigrationPlan[]; summary: ScheduledReportMigrationSummary } {
+  const plans: ScheduledReportMigrationPlan[] = []
+  const seen = new Set<string>()
+  let duplicatesSkipped = 0
+  let lossless = 0
+  let needsReview = 0
+  for (const task of tasks) {
+    if (!task || typeof task !== 'object') continue
+    const plan = planScheduledReportMigration(task, resolveConversationId, now)
+    if (seen.has(plan.ruleId)) {
+      duplicatesSkipped += 1
+      continue
+    }
+    seen.add(plan.ruleId)
+    plans.push(plan)
+    if (plan.outcome === 'lossless_source_chat') lossless += 1
+    else needsReview += 1
+  }
+  return {
+    plans,
+    summary: {
+      total: plans.length + duplicatesSkipped,
+      migrated: plans.length,
+      lossless,
+      needsReview,
+      duplicatesSkipped
+    }
+  }
+}
@@ -0,0 +1,81 @@
+import {
+  SCHEDULED_REPORT_TARGET_OPTIONS,
+  scheduledReportTargetLabel,
+  type ScheduledReportAutomationConfig
+} from '../../shared/automation'
+import {
+  resolveAutomationTarget,
+  resolveGroupConversationId,
+  resolveGroupDisplayName,
+  type AutomationTargetContact,
+  type AutomationTargetResolution
+} from './automation-wechat-target'
+
+/**
+ * 定时日报的**目标解析**（薄适配层）。
+ *
+ * 判定逻辑在 `automation-wechat-target.ts`，与退群通知**共用同一份**。
+ * 差异只有一处：来源会话不是"事件所在群"，而是**这条规则配置的日报来源群**。
+ */
+
+/** 定时日报的目标文案（唯一一份）。 */
+export const SCHEDULED_REPORT_TARGET_MESSAGES = {
+  options: SCHEDULED_REPORT_TARGET_OPTIONS,
+  invalidTarget: '定时日报的发送目标无效，请重新选择',
+  sourceMissing: '无法确定日报来源群，本次日报未发送',
+  sourceRecipientName: '日报来源群',
+  sourceDisplayName: '日报来源群',
+  selfMissing: '无法确定当前登录的微信账号，本次日报未发送',
+  contactNotChosen: '还没有选择发送联系人，请重新选择',
+  contactUnavailable: '发送联系人已不存在或当前无法发送，请重新选择'
+} as const
+
+export function resolveScheduledReportTarget(input: {
+  config: ScheduledReportAutomationConfig | undefined
+  contacts: AutomationTargetContact[]
+  /** 当前登录账号的 wxid；拿不到时传空串。 */
+  selfWxid?: string
+}): AutomationTargetResolution {
+  const config = input.config
+  if (!config) return { ok: false, error: '定时日报尚未配置发送目标' }
+  const sourceConversationId = String(config.report?.sourceConversationId || '').trim()
+  return resolveAutomationTarget(
+    {
+      targetType: config.target?.type as never,
+      sourceConversationId,
+      // 来源群的可读名从通讯录解析；解析不到时由中性层给兜底称呼。
+      sourceDisplayName: resolveGroupDisplayName(sourceConversationId, input.contacts),
+      ...(config.target?.contactId ? { contactId: config.target.contactId } : {}),
+      contacts: input.contacts,
+      ...(input.selfWxid ? { selfWxid: input.selfWxid } : {})
+    },
+    SCHEDULED_REPORT_TARGET_MESSAGES
+  )
+}
+
+/**
+ * 日报来源群的可读显示名（执行日志的 `sourceDisplayName` 用它）。
+ *
+ * **绝不回落到裸 id**：解析不到就说「日报来源群」。
+ */
+export function scheduledReportSourceDisplayName(
+  config: ScheduledReportAutomationConfig | undefined,
+  contacts: AutomationTargetContact[]
+): string {
+  const raw = String(config?.report?.sourceConversationId || '').trim()
+  if (!raw) return '日报来源群'
+  return resolveGroupDisplayName(raw, contacts) || '日报来源群'
+}
+
+/** 日报来源群解析成稳定会话 id（执行前预检用；解析不到返回 undefined）。 */
+export function scheduledReportSourceConversationId(
+  config: ScheduledReportAutomationConfig | undefined,
+  contacts: AutomationTargetContact[]
+): string | undefined {
+  return resolveGroupConversationId(
+    String(config?.report?.sourceConversationId || ''),
+    contacts
+  )
+}
+
+export { scheduledReportTargetLabel }
@@ -39,9 +39,8 @@ export interface AppSettings {
  showStartupProgress: boolean
  ttsSelectedVoiceId: string
  ttsModel: TextToSpeechModel
-  /** Keep a running personal-WeChat OneBot process across app restarts. */
-  keepPersonalWechatProcess?: boolean
  windowsWechatPort: string
+  reportImagePostfixText: string
  /**
   * Query Agent 是否为桌面「问问微信」与 Agent Hub 查询类问题的主路径。
   * 默认开启；关闭后回退到 Legacy AI Search Pipeline（仅作 runtime regression 时的回退开关，
@@ -130,8 +129,8 @@ const DEFAULT_SETTINGS: AppSettings = {
  showStartupProgress: true,
  ttsSelectedVoiceId: '',
  ttsModel: 's2.1-pro-free',
-  keepPersonalWechatProcess: false,
  windowsWechatPort: '',
+  reportImagePostfixText: '今日日报',
  queryAgentEnabled: true
 }

@@ -150,7 +149,10 @@ export function loadSettings(): AppSettings {
  if (cache) return cache
  try {
    if (fs.existsSync(SETTINGS_FILE)) {
-      const raw = fs.readJsonSync(SETTINGS_FILE) as Partial<AppSettings>
+      const stored = fs.readJsonSync(SETTINGS_FILE) as Partial<AppSettings> & {
+        keepPersonalWechatProcess?: unknown
+      }
+      const { keepPersonalWechatProcess: retiredKeepProcess, ...raw } = stored
      cache = { ...DEFAULT_SETTINGS, ...raw }
      if (raw.autoLogin === undefined) {
        const hasSavedDatabaseKey = fs.existsSync(
@@ -176,7 +178,7 @@ export function loadSettings(): AppSettings {
      }
      // 防撤回已下线（设置入口已隐藏）：历史版本可能把它持久化为 true。
      // 这里强制收敛为 false 并回写磁盘，确保旧的撤回监听与撤回日志不会继续运行。
-      if (cache.recallProtectionEnabled) {
+      if (cache.recallProtectionEnabled || retiredKeepProcess !== undefined) {
        cache.recallProtectionEnabled = false
        saveSettings(cache)
      }
@@ -192,7 +194,11 @@ export function loadSettings(): AppSettings {
 export function saveSettings(next: AppSettings): AppSettings {
  // 防撤回已下线：所有写入路径（含 settings:set 补丁）统一收敛为 false，
  // 避免遗留入口或旧版本把它重新打开。
-  cache = { ...next, recallProtectionEnabled: false }
+  const sanitized = { ...next } as AppSettings & {
+    keepPersonalWechatProcess?: unknown
+  }
+  delete sanitized.keepPersonalWechatProcess
+  cache = { ...sanitized, recallProtectionEnabled: false }
  try {
    ensureDir()
    fs.writeJsonSync(SETTINGS_FILE, cache, { spaces: 2 })
@@ -5,14 +5,14 @@ import path from 'path'
 import type {
  PersonalWechatSendCapability,
  PersonalWechatSendRequest,
-  PersonalWechatSendResult
+  PersonalWechatSendResult,
+  PersonalWechatSenderStatus
 } from '../../shared/personal-wechat'
 import type {
  PolicyDecision,
  WechatActionAuditRecord,
  WechatActionContent,
  WechatActionErrorCode,
-  WechatActionMemberEventReference,
  WechatActionRequest,
  WechatActionResult
 } from '../../shared/wechat-action'
@@ -22,17 +22,40 @@ import { wechatSendGateway } from './wechat-send-gateway'
 const MAX_AUDIT_RECORDS = 500
 const MAX_CONTENT_PREVIEW_LENGTH = 240
 export const AUTOMATION_SEND_INTERVAL_MS = 3_000
-const AUTOMATION_PURPOSE_ALLOWLIST = new Set(['scheduled_report', 'member_left_notification'])
+const AUTOMATION_PURPOSE_ALLOWLIST = new Set([
+  'scheduled_report',
+  // Automation（@我生成日报）。必须与 `src/shared/automation.ts` 的
+  // `AUTOMATION_SEND_PURPOSE` 保持一致，否则会被下面 evaluateWechatActionPolicy
+  // 以 ACTION_NOT_ALLOWED 拦下 —— 那是有意的闸门，不是 bug。
+  'automation_reply',
+  'automation_report',
+  // 退群通知（由 Automation 发出）。目标可以是群 / 自己 / 文件传输助手 / 指定好友，
+  // 所以**不绑定**任何 "只能发回原群" 的作用域锁。
+  'automation_leave_notification',
+  // 定时日报（由 Automation 的 `scheduled_report` 规则发出）。目标是四选一，
+  // 同样**不绑定**任何作用域锁。
+  'automation_scheduled_report',
+  // 定时日报的「后置词」（图片 sent 之后补发的那条文本）。与图片是两个独立
+  // purpose，各自有幂等位 —— 少这一条会被 evaluateWechatActionPolicy 拦下。
+  'automation_scheduled_report_postfix',
+  'manual_report_image',
+  'manual_report_postfix'
+])
+
+export interface ReportImageSequenceRequest {
+  recipient: WechatActionRequest['recipient']
+  imagePath: string
+  postfixText: string
+}
+
+export interface ReportImageSequenceResult {
+  image: WechatActionResult
+  postfix?: WechatActionResult
+}

 export interface WechatActionGatewayDependencies {
  getCapability?: () => Promise<PersonalWechatSendCapability>
  send?: (request: PersonalWechatSendRequest) => Promise<PersonalWechatSendResult>
-  getMemberEvent?: (
-    sourceId: string
-  ) =>
-    | WechatActionMemberEventReference
-    | Promise<WechatActionMemberEventReference | undefined>
-    | undefined
  getUserDataPath?: () => string
  now?: () => Date
  wait?: (milliseconds: number) => Promise<void>
@@ -43,10 +66,6 @@ interface LoadedAuditState {
  records: WechatActionAuditRecord[]
 }

-export interface WechatActionPolicyContext {
-  memberEvent?: WechatActionMemberEventReference
-}
-
 const defaultDependencies = (): Required<
  Pick<
    WechatActionGatewayDependencies,
@@ -76,7 +95,6 @@ export class WechatActionGateway {
      WechatActionGatewayDependencies,
      'getCapability' | 'send' | 'getUserDataPath' | 'now' | 'wait'
    >
-  private readonly memberEvents = new Map<string, WechatActionMemberEventReference>()
  private readonly inFlight = new Map<string, Promise<WechatActionResult>>()
  private auditState: LoadedAuditState | null = null
  private automationSendTail: Promise<void> = Promise.resolve()
@@ -86,20 +104,33 @@ export class WechatActionGateway {
    this.deps = { ...defaultDependencies(), ...deps }
  }

-  /** 记录退群事件，发送通知前可确认事件所属群聊。 */
-  registerMemberEvent(event: WechatActionMemberEventReference): void {
-    const id = String(event?.id || '').trim()
-    const roomId = String(event?.roomId || '').trim()
-    if (!id || !roomId) return
-    this.memberEvents.set(id, { id, roomId })
-  }
+  /**
+   * 用户确认后的日报发送序列。两步都进入 automation 发送队列，从而复用既有 3 秒间隔；
+   * 后置词 action 只在图片明确 sent 后创建，图片失败/blocked 时严格短路。
+   */
+  async executeReportImageSequence(
+    request: ReportImageSequenceRequest
+  ): Promise<ReportImageSequenceResult> {
+    const image = await this.execute({
+      origin: 'user_manual',
+      purpose: 'manual_report_image',
+      triggerType: 'automation',
+      recipient: request.recipient,
+      content: { type: 'image', path: String(request.imagePath || '') }
+    })
+    if (image.status !== 'sent') return { image }

-  registerMemberEvents(events: WechatActionMemberEventReference[]): void {
-    for (const event of events) this.registerMemberEvent(event)
-  }
+    const postfixText = String(request.postfixText || '').trim()
+    if (!postfixText) return { image }

-  clearMemberEvents(): void {
-    this.memberEvents.clear()
+    const postfix = await this.execute({
+      origin: 'user_manual',
+      purpose: 'manual_report_postfix',
+      triggerType: 'automation',
+      recipient: request.recipient,
+      content: { type: 'text', text: postfixText }
+    })
+    return { image, postfix }
  }

  listAuditRecords(): WechatActionAuditRecord[] {
@@ -145,8 +176,7 @@ export class WechatActionGateway {
      )
    }

-    const eventContext = await this.resolveEventContext(request)
-    const policy = evaluateWechatActionPolicy(request, eventContext)
+    const policy = evaluateWechatActionPolicy(request)
    if (policy.decision !== 'allow') {
      return this.finishBlocked(
        actionId,
@@ -240,28 +270,6 @@ export class WechatActionGateway {
    return queued
  }

-  private async resolveEventContext(
-    request: WechatActionRequest
-  ): Promise<WechatActionPolicyContext> {
-    if (request.purpose !== 'member_left_notification') return {}
-    const sourceId = String(request.sourceId || '').trim()
-    if (!sourceId) return {}
-    let memberEvent: WechatActionMemberEventReference | undefined = this.memberEvents.get(sourceId)
-    if (!memberEvent && this.deps.getMemberEvent) {
-      let resolved: WechatActionMemberEventReference | undefined
-      try {
-        resolved = await this.deps.getMemberEvent(sourceId)
-      } catch {
-        resolved = undefined
-      }
-      if (resolved && String(resolved.id || '').trim() === sourceId) {
-        memberEvent = { id: String(resolved.id), roomId: String(resolved.roomId) }
-        this.registerMemberEvent(memberEvent)
-      }
-    }
-    return memberEvent ? { memberEvent } : {}
-  }
-
  private async findExisting(idempotencyKey: string): Promise<WechatActionResult | undefined> {
    const state = this.loadAuditState()
    const existing = state.records.find((record) => record.idempotencyKey === idempotencyKey)
@@ -280,9 +288,6 @@ export class WechatActionGateway {
  private idempotencyKey(request: WechatActionRequest): string | undefined {
    const explicit = String(request.idempotencyKey || '').trim()
    if (explicit) return explicit
-    if (request.purpose === 'member_left_notification' && request.sourceId) {
-      return `member_left_notification:${request.sourceId}`
-    }
    if (request.origin === 'scheduled_report' && request.executionId) {
      return `scheduled_report:${request.executionId}`
    }
@@ -429,10 +434,7 @@ export class WechatActionGateway {
  }
 }

-export function evaluateWechatActionPolicy(
-  request: WechatActionRequest,
-  context: WechatActionPolicyContext = {}
-): PolicyDecision {
+export function evaluateWechatActionPolicy(request: WechatActionRequest): PolicyDecision {
  if (!request || typeof request !== 'object') {
    return {
      decision: 'block',
@@ -463,29 +465,6 @@ export function evaluateWechatActionPolicy(
      reason: `自动化动作不允许执行 purpose=${request.purpose}`
    }
  }
-  if (request.purpose === 'member_left_notification') {
-    if (
-      !request.sourceId ||
-      !context.memberEvent ||
-      context.memberEvent.id !== request.sourceId ||
-      !context.memberEvent.roomId
-    ) {
-      return {
-        decision: 'block',
-        source: 'deterministic',
-        reasonCode: 'INVALID_REQUEST',
-        reason: '退群通知必须关联已记录的退群事件'
-      }
-    }
-    if (request.recipient.type !== 'group' || request.recipient.id !== context.memberEvent.roomId) {
-      return {
-        decision: 'block',
-        source: 'deterministic',
-        reasonCode: 'RECIPIENT_SCOPE_VIOLATION',
-        reason: '退群通知只能发送回原事件所在群聊'
-      }
-    }
-  }
  return { decision: 'allow', source: 'deterministic' }
 }

@@ -600,6 +579,32 @@ function toPersonalWechatSendRequest(request: WechatActionRequest): PersonalWech
  }
 }

+/**
+ * 把网关结果还原成既有调用方期望的 `PersonalWechatSendResult`。
+ *
+ * 为什么需要：手动发送的 IPC（`wechat-personal:send`）返回契约是
+ * `PersonalWechatSendResult`，界面上靠 `response.success` / `response.error` 判读。
+ * 改成走网关之后必须把这个契约**原样**还回去，否则「统一发送路径」会把 UI 一起改坏。
+ *
+ * 网关成功时会把底层 `sendResult` 原样挂在 `action.sendResult` 上，优先用它
+ * （它带着真实的 `status`）；拿不到时按 `action.status` 合成一个。
+ */
+export function toPersonalWechatSendResult(
+  action: WechatActionResult,
+  fallbackStatus: PersonalWechatSenderStatus
+): PersonalWechatSendResult {
+  const raw = action.sendResult
+  if (raw && typeof raw === 'object' && 'success' in (raw as Record<string, unknown>)) {
+    return raw as PersonalWechatSendResult
+  }
+  if (action.status === 'sent') return { success: true, status: fallbackStatus }
+  return {
+    success: false,
+    status: fallbackStatus,
+    error: action.reason || action.errorCode || '微信发送失败'
+  }
+}
+
 function contentAudit(
  content: WechatActionContent
 ): Pick<WechatActionAuditRecord, 'contentPreview' | 'contentHash'> {
@@ -0,0 +1,51 @@
+import { readFileSync } from 'fs'
+
+export interface TrayNativeImage {
+  addRepresentation(options: { scaleFactor: number; dataURL: string }): void
+  isEmpty(): boolean
+  resize(options: { width: number; height: number; quality: 'best' }): TrayNativeImage
+  setTemplateImage(option: boolean): void
+}
+
+export interface TrayNativeImageFactory<T extends TrayNativeImage = TrayNativeImage> {
+  createEmpty(): T
+  createFromPath(path: string): T
+}
+
+export interface TrayTemplateIconPaths {
+  oneX: string
+  twoX: string
+}
+
+type ReadImageFile = (path: string) => Buffer
+
+function toPngDataUrl(buffer: Buffer): string {
+  return `data:image/png;base64,${buffer.toString('base64')}`
+}
+
+export function createTrayImage<T extends TrayNativeImage>(
+  platform: NodeJS.Platform,
+  appIconPath: string,
+  templateIconPaths: TrayTemplateIconPaths,
+  nativeImage: TrayNativeImageFactory<T>,
+  readImageFile: ReadImageFile = readFileSync
+): T {
+  if (platform === 'darwin') {
+    const image = nativeImage.createEmpty()
+    image.addRepresentation({
+      scaleFactor: 1,
+      dataURL: toPngDataUrl(readImageFile(templateIconPaths.oneX))
+    })
+    image.addRepresentation({
+      scaleFactor: 2,
+      dataURL: toPngDataUrl(readImageFile(templateIconPaths.twoX))
+    })
+    image.setTemplateImage(true)
+    return image
+  }
+
+  const image = nativeImage.createFromPath(appIconPath)
+  return image.isEmpty()
+    ? nativeImage.createEmpty()
+    : (image.resize({ width: 24, height: 24, quality: 'best' }) as T)
+}
@@ -19,9 +19,9 @@ export interface VoiceSilkMetadata {
 }

 /**
- * Validate the exact PCM contract consumed by the bundled OneBot encoder.
- * The Go Silk encoder emits a header-only payload for sub-frame input, which
- * is accepted by the old path but produces an unplayable WeChat voice.
+ * Validate the exact PCM contract consumed by the bundled SILK encoder.
+ * The encoder emits a header-only payload for sub-frame input, which would
+ * produce an unplayable WeChat voice.
 */
 export function validateVoicePcm(
  pcm: Uint8Array,
@@ -384,6 +384,79 @@ export function resolveWindowsNativeAccountRoot(
  }
 }

+/**
+ * native monitor 事件（pipe / socket 上的一行 JSON）。
+ *
+ * ## 两个协议版本
+ *
+ * **v1（历史）** —— 变更信号来自**文件系统通知**（Windows `ReadDirectoryChangesW` /
+ * macOS `kqueue`），native 只知道「哪个文件被写了」：
+ *
+ * ```json
+ * {"db":"message_0.db","table":"message","action":"update"}
+ * ```
+ *
+ * 它**不携带会话**，所以上层只能回读「最近活跃会话」。`protocol` 会标成 1。
+ *
+ * **v2（Native Monitor Event v2）** —— native 侧在收到文件变化后做一次
+ * SessionTable watermark diff，把**每一个真正变化的会话**各发一条：
+ *
+ * ```json
+ * {"version":2,"kind":"message_change","session_id":"xxx@chatroom",
+ *  "db":"message_0.db","table":"message","action":"update","observed_at_ms":1}
+ * ```
+ *
+ * `protocol = 2` 且带 `sessionId` 时，上层可以**直接按会话回读**，不再依赖
+ * `getSessions()[0]`。
+ *
+ * **降级规则**：只要 `version`/`kind`/`session_id` 任一不符合 v2 契约，
+ * 就按 v1 处理（`protocol = 1`）—— 宁可不精确，也不能拿一个不可信的会话 id 去读。
+ */
+export interface Wcdb4MonitorEvent {
+  /** pipe 上的原始一行（原样转发给需要它的消费者，例如退群监控）。 */
+  raw: string
+  /** 语义化的动作类型。当事件语义出现时用它，比如 `'user_change'`。 */
+  type: string
+  action: string
+  /** `2` = 精确会话事件；`1` = legacy（不含会话）。 */
+  protocol: 1 | 2
+  /** 粗表分类：`database` / `message` / `Session` / `contact`。 */
+  table: string
+  /** `protocol === 2` 时才有：发生变化的会话（`xxx@chatroom` 或 wxid）。 */
+  sessionId?: string
+  /** native 侧观测时刻（epoch ms），仅用于诊断。 */
+  observedAtMs?: number
+}
+
+/** 把 pipe 上的一行 payload 解析成事件。纯函数，便于单测。 */
+export function parseMonitorEvent(rawPayload: string): Wcdb4MonitorEvent {
+  const raw = rawPayload.trim()
+  let parsed: Record<string, unknown> = {}
+  try {
+    const value = JSON.parse(raw) as unknown
+    if (value && typeof value === 'object' && !Array.isArray(value)) {
+      parsed = value as Record<string, unknown>
+    }
+  } catch {
+    // 不是 JSON：当作 legacy 事件处理，保留原始 payload。
+  }
+
+  const action = typeof parsed.action === 'string' && parsed.action ? parsed.action : 'update'
+  const sessionId = typeof parsed.session_id === 'string' ? parsed.session_id.trim() : ''
+  // 三个条件缺一不可：版本、语义、以及**非空**的会话 id。
+  const precise = parsed.version === 2 && parsed.kind === 'message_change' && Boolean(sessionId)
+
+  return {
+    raw,
+    type: action,
+    action,
+    protocol: precise ? 2 : 1,
+    table: typeof parsed.table === 'string' ? parsed.table : '',
+    ...(precise ? { sessionId } : {}),
+    ...(typeof parsed.observed_at_ms === 'number' ? { observedAtMs: parsed.observed_at_ms } : {})
+  }
+}
+
 export class Wcdb4Client {
  static readonly defaultRoot = Wcdb4Client.findExistingDefaultRoot()

@@ -495,7 +568,7 @@ export class Wcdb4Client {
  private wcdbStopMonitorPipe: (() => void) | null = null
  private wcdbGetMonitorPipeName: ((outName: WcdbVoidOut) => number) | null = null
  private monitorPipeClient: Socket | null = null
-  private monitorCallback: ((type: string, json: string) => void) | null = null
+  private monitorCallback: ((event: Wcdb4MonitorEvent) => void) | null = null
  private monitorConnectTimer: ReturnType<typeof setTimeout> | null = null
  private monitorReconnectTimer: ReturnType<typeof setTimeout> | null = null
  private monitorPipePath = ''
@@ -773,7 +846,7 @@ export class Wcdb4Client {
    })
  }

-  async startMonitor(callback: (type: string, json: string) => void): Promise<boolean> {
+  async startMonitor(callback: (event: Wcdb4MonitorEvent) => void): Promise<boolean> {
    if (this.closing || !this.wcdbStartMonitorPipe || !this.wcdbGetMonitorPipeName || !this.koffi) {
      return false
    }
@@ -911,13 +984,7 @@ export class Wcdb4Client {
  private emitMonitorPayload(rawPayload: string): void {
    const payload = rawPayload.trim()
    if (!payload || !this.monitorCallback) return
-
-    try {
-      const parsed = JSON.parse(payload) as { action?: string }
-      this.monitorCallback(parsed.action || 'update', payload)
-    } catch {
-      this.monitorCallback('update', payload)
-    }
+    this.monitorCallback(parseMonitorEvent(payload))
  }

  private scheduleMonitorReconnect(): void {
@@ -82,26 +82,28 @@ import type {
  PersonalWechatVoiceDiagnostic
 } from '../shared/personal-wechat'
 import type { PersonalWechatSendCapability } from '../shared/personal-wechat'
+import type { GroupMemberStatsQuery, GroupMemberStatsResult } from '../shared/group-stats'
 import type {
-  ScheduledReportCreateInput,
  ScheduledReportExecution,
+  ScheduledReportNotification,
+  ScheduledReportNotificationCapability,
  ScheduledReportNotificationSettings,
  ScheduledReportNotificationSettingsResult,
-  ScheduledReportResult,
-  ScheduledReportTask,
-  ScheduledReportUpdateInput
+  ScheduledReportResult
 } from '../shared/scheduled-report'
-import type {
-  PersonalWechatRuntimeDownloadResult,
-  PersonalWechatRuntimeProgressEvent,
-  PersonalWechatRuntimeStatus
-} from '../shared/personal-wechat-runtime'
 import type {
  PersonalWechatVoiceEncodingEnvironment,
  PersonalWechatVoiceEncodingEnvironmentResult
 } from '../shared/personal-wechat-voice-runtime'
 import type { AppLogEntry } from '../shared/app-log'
-import type { GroupExitMonitorState } from '../shared/group-exit-monitor'
+import type { GroupExitMonitorEvent, GroupExitMonitorState } from '../shared/group-exit-monitor'
+import type {
+  AutomationExecution,
+  AutomationRule,
+  AutomationRuleDraft,
+  AutomationStatusSummary,
+  ScheduledRuleRunOutcome
+} from '../shared/automation'
 import type { ActionLogEntry } from '../shared/action-log'
 import type {
  AppUpdateCheckResult,
@@ -293,18 +295,37 @@ declare global {
          avatar: string
        }[]
      } | null>
+      getGroupMemberStats: (request: GroupMemberStatsQuery) => Promise<GroupMemberStatsResult>
      getGroupExitMonitorState: () => Promise<GroupExitMonitorState>
+      listGroupExitMonitorEvents: (query?: {
+        roomId?: string
+        sinceMs?: number
+        untilMs?: number
+        limit?: number
+      }) => Promise<GroupExitMonitorEvent[]>
      setGroupExitMonitorEnabled: (enabled: boolean) => Promise<GroupExitMonitorState>
-      setGroupExitMonitorGroups: (
-        roomIds: string[],
-        notificationRoomIds?: string[]
-      ) => Promise<GroupExitMonitorState>
-      setGroupExitMonitorNotificationTemplate: (template: string) => Promise<GroupExitMonitorState>
+      /** 保存**监控范围**。通知配置已迁到「自动化 → 退群通知」。 */
+      setGroupExitMonitorGroups: (roomIds: string[]) => Promise<GroupExitMonitorState>
      checkGroupExitMonitorNow: () => Promise<GroupExitMonitorState>
      clearGroupExitMonitorEvents: () => Promise<GroupExitMonitorState>
-      resendGroupExitMonitorEvent: (eventId: string) => Promise<GroupExitMonitorState>
      markGroupExitMonitorRead: (readAt?: number) => Promise<GroupExitMonitorState>
      listWechatActionLogs: () => Promise<ActionLogEntry[]>
+      getAutomationStatus: () => Promise<AutomationStatusSummary>
+      listAutomationRules: () => Promise<AutomationRule[]>
+      createAutomationRule: (draft: AutomationRuleDraft) => Promise<AutomationRule>
+      updateAutomationRule: (
+        id: string,
+        draft: AutomationRuleDraft
+      ) => Promise<AutomationRule | null>
+      deleteAutomationRule: (id: string) => Promise<boolean>
+      setAutomationRuleEnabled: (id: string, enabled: boolean) => Promise<AutomationRule | null>
+      listAutomationExecutions: (query?: { limit?: number }) => Promise<AutomationExecution[]>
+      clearAutomationExecutions: () => Promise<boolean>
+      listAutomationGroups: () => Promise<Array<{ id: string; name: string }>>
+      /** 保存「退群通知」规则（singleton upsert）。 */
+      saveLeaveNotificationRule: (draft: AutomationRuleDraft) => Promise<AutomationRule>
+      /** 「指定好友」的可选项；已在 main 侧过滤掉群聊 / 公众号 / 文件传输助手 / 自己。 */
+      listSendableContacts: () => Promise<Array<{ id: string; name: string }>>
      onGroupExitMonitorState: (callback: (state: GroupExitMonitorState) => void) => () => void
      search: (keyword: string) => Promise<string | null>
      searchKnowledge: (request: KnowledgeSearchIpcRequest) => Promise<KnowledgeSearchIpcResult>
@@ -521,6 +542,7 @@ declare global {
          ttsSelectedVoiceId: string
          ttsModel: import('../shared/text-to-speech').TextToSpeechModel
          windowsWechatPort: string
+          reportImagePostfixText: string
          imageXorKey: string
          imageAesKey: string
        }
@@ -561,6 +583,7 @@ declare global {
          ttsSelectedVoiceId: string
          ttsModel: import('../shared/text-to-speech').TextToSpeechModel
          windowsWechatPort: string
+          reportImagePostfixText: string
          imageXorKey: string
          imageAesKey: string
        }
@@ -584,6 +607,7 @@ declare global {
          ttsSelectedVoiceId: string
          ttsModel: import('../shared/text-to-speech').TextToSpeechModel
          windowsWechatPort: string
+          reportImagePostfixText: string
          imageXorKey: string
          imageAesKey: string
        }>
@@ -697,21 +721,11 @@ declare global {
      onImageTextIndexStatus: (callback: (status: ImageTextIndexStatus) => void) => () => void
      getPersonalWechatSenderStatus: () => Promise<PersonalWechatSenderStatus>
      getPersonalWechatSendCapability: () => Promise<PersonalWechatSendCapability>
-      getPersonalWechatKeepOneBotProcess: () => Promise<boolean>
-      setPersonalWechatKeepOneBotProcess: (keep: boolean) => Promise<boolean>
      checkPersonalWechatSenderStatus: (port?: string) => Promise<PersonalWechatSenderStatus>
      checkPersonalWechatVoiceEncodingEnvironment: () => Promise<PersonalWechatVoiceEncodingEnvironment>
      installPersonalWechatPilk: () => Promise<PersonalWechatVoiceEncodingEnvironmentResult>
      openPersonalWechatVoicePythonDownload: () => Promise<{ success: boolean; error?: string }>
      openPersonalWechatVoiceFfmpegDownload: () => Promise<{ success: boolean; error?: string }>
-      getPersonalWechatRuntimeStatus: () => Promise<PersonalWechatRuntimeStatus>
-      downloadPersonalWechatRuntime: () => Promise<PersonalWechatRuntimeDownloadResult>
-      cancelPersonalWechatRuntimeDownload: () => Promise<{ success: boolean }>
-      removePersonalWechatRuntime: () => Promise<PersonalWechatRuntimeStatus>
-      openPersonalWechatRuntimeDirectory: () => Promise<{ success: boolean; error?: string }>
-      onPersonalWechatRuntimeProgress: (
-        callback: (status: PersonalWechatRuntimeProgressEvent) => void
-      ) => () => void
      rebindPersonalWechatSender: () => Promise<PersonalWechatSenderStatus>
      sendGeneratedTtsVoice: (
        request: PersonalWechatGeneratedTtsVoiceRequest
@@ -719,35 +733,23 @@ declare global {
      sendPersonalWechatMessage: (
        request: PersonalWechatSendRequest
      ) => Promise<PersonalWechatSendResult>
-      listScheduledReports: () => Promise<ScheduledReportTask[]>
-      listScheduledReportExecutions: (taskId?: string) => Promise<ScheduledReportExecution[]>
+      /** 定时日报：「立即执行」——与 scheduler 走同一条执行链路（manual trigger）。 */
+      runScheduledReportRule: (
+        ruleId: string
+      ) => Promise<{ success: boolean; error?: string; data?: ScheduledRuleRunOutcome }>
+      /** 定时日报：旧执行记录**只读存档**（旧记录无法无损转换，原样展示）。 */
+      listScheduledReportLegacyExecutions: (
+        ruleId?: string
+      ) => Promise<ScheduledReportExecution[]>
+      /** 定时日报：微信异常通知开关与能力检测。 */
      getScheduledReportNotificationSettings: () => Promise<ScheduledReportNotificationSettings>
+      getScheduledReportNotificationCapability: () => Promise<ScheduledReportNotificationCapability>
      setScheduledReportNotificationEnabled: (
        enabled: boolean
      ) => Promise<ScheduledReportNotificationSettingsResult>
-      createScheduledReport: (
-        request: ScheduledReportCreateInput
-      ) => Promise<ScheduledReportResult<ScheduledReportTask>>
-      updateScheduledReport: (
-        taskId: string,
-        request: ScheduledReportUpdateInput
-      ) => Promise<ScheduledReportResult<ScheduledReportTask>>
-      deleteScheduledReport: (
-        taskId: string
-      ) => Promise<ScheduledReportResult<{ deletedId: string }>>
-      setScheduledReportEnabled: (
-        taskId: string,
-        enabled: boolean
-      ) => Promise<ScheduledReportResult<ScheduledReportTask>>
-      runScheduledReportNow: (
-        taskId: string
-      ) => Promise<ScheduledReportResult<ScheduledReportExecution>>
-      retryScheduledReportSend: (
-        executionId: string
-      ) => Promise<ScheduledReportResult<ScheduledReportExecution>>
      testScheduledReportErrorNotification: (
-        taskId: string
-      ) => Promise<ScheduledReportResult<ScheduledReportExecution>>
+        ruleId: string
+      ) => Promise<ScheduledReportResult<ScheduledReportNotification>>
      getPersonalWechatVoiceDiagnostic: () => Promise<PersonalWechatVoiceDiagnostic | null>
      getAgentHubStatus: () => Promise<AgentHubStatus>
      getAgentHubLogs: () => Promise<AgentHubLogEntry[]>
@@ -52,26 +52,28 @@ import type {
 } from '../shared/personal-wechat'
 import type { PersonalWechatSendCapability } from '../shared/personal-wechat'
 import type {
-  ScheduledReportCreateInput,
  ScheduledReportExecution,
+  ScheduledReportNotification,
+  ScheduledReportNotificationCapability,
  ScheduledReportNotificationSettings,
  ScheduledReportNotificationSettingsResult,
-  ScheduledReportResult,
-  ScheduledReportTask,
-  ScheduledReportUpdateInput
+  ScheduledReportResult
 } from '../shared/scheduled-report'
-import type {
-  PersonalWechatRuntimeDownloadResult,
-  PersonalWechatRuntimeProgressEvent,
-  PersonalWechatRuntimeStatus
-} from '../shared/personal-wechat-runtime'
 import type {
  PersonalWechatVoiceEncodingEnvironment,
  PersonalWechatVoiceEncodingEnvironmentResult
 } from '../shared/personal-wechat-voice-runtime'
 import type { AppLogEntry } from '../shared/app-log'
 import type { AppUpdateState } from '../shared/app-update'
-import type { GroupExitMonitorState } from '../shared/group-exit-monitor'
+import type { GroupExitMonitorEvent, GroupExitMonitorState } from '../shared/group-exit-monitor'
+import type {
+  AutomationExecution,
+  AutomationRule,
+  AutomationRuleDraft,
+  AutomationStatusSummary,
+  ScheduledRuleRunOutcome
+} from '../shared/automation'
+import type { GroupMemberStatsQuery, GroupMemberStatsResult } from '../shared/group-stats'
 import type { ActionLogEntry } from '../shared/action-log'
 import type { CacheClearScope, CacheSummary } from '../shared/cache'
 import type { ExportRequest, ExportJobProgress } from '../shared/export'
@@ -167,29 +169,59 @@ const api = {
  ): Promise<MessagesAroundResult> =>
    ipcRenderer.invoke('db:getMessagesAround', userMd5, messageId, anchorSeconds, radiusSeconds),
  getGroupSnapshot: (userMd5: string) => ipcRenderer.invoke('db:getGroupSnapshot', userMd5),
+  getGroupMemberStats: (request: GroupMemberStatsQuery): Promise<GroupMemberStatsResult> =>
+    ipcRenderer.invoke('group-stats:getMemberStats', request),
  getGroupExitMonitorState: (): Promise<GroupExitMonitorState> =>
    ipcRenderer.invoke('group-exit-monitor:getState'),
+  listGroupExitMonitorEvents: (query?: {
+    roomId?: string
+    sinceMs?: number
+    untilMs?: number
+    limit?: number
+  }): Promise<GroupExitMonitorEvent[]> =>
+    ipcRenderer.invoke('group-exit-monitor:listEvents', query),
  setGroupExitMonitorEnabled: (enabled: boolean): Promise<GroupExitMonitorState> =>
    ipcRenderer.invoke('group-exit-monitor:setEnabled', enabled),
-  setGroupExitMonitorGroups: (
-    roomIds: string[],
-    notificationRoomIds?: string[]
-  ): Promise<GroupExitMonitorState> =>
-    notificationRoomIds === undefined
-      ? ipcRenderer.invoke('group-exit-monitor:setGroups', roomIds)
-      : ipcRenderer.invoke('group-exit-monitor:setGroups', roomIds, notificationRoomIds),
-  setGroupExitMonitorNotificationTemplate: (template: string): Promise<GroupExitMonitorState> =>
-    ipcRenderer.invoke('group-exit-monitor:setTemplate', template),
+  /** 保存**监控范围**。通知配置已迁到自动化规则，这里不再有第二个参数。 */
+  setGroupExitMonitorGroups: (roomIds: string[]): Promise<GroupExitMonitorState> =>
+    ipcRenderer.invoke('group-exit-monitor:setGroups', roomIds),
  checkGroupExitMonitorNow: (): Promise<GroupExitMonitorState> =>
    ipcRenderer.invoke('group-exit-monitor:checkNow'),
  clearGroupExitMonitorEvents: (): Promise<GroupExitMonitorState> =>
    ipcRenderer.invoke('group-exit-monitor:clearEvents'),
-  resendGroupExitMonitorEvent: (eventId: string): Promise<GroupExitMonitorState> =>
-    ipcRenderer.invoke('group-exit-monitor:resendEvent', eventId),
  markGroupExitMonitorRead: (readAt?: number): Promise<GroupExitMonitorState> =>
    ipcRenderer.invoke('group-exit-monitor:markRead', readAt),
  listWechatActionLogs: (): Promise<ActionLogEntry[]> =>
    ipcRenderer.invoke('wechat-action-log:list'),
+  // ---- Automation v1（@我生成日报）。命名与既有扁平风格一致。 ----
+  getAutomationStatus: (): Promise<AutomationStatusSummary> =>
+    ipcRenderer.invoke('automation:getStatus'),
+  listAutomationRules: (): Promise<AutomationRule[]> => ipcRenderer.invoke('automation:listRules'),
+  createAutomationRule: (draft: AutomationRuleDraft): Promise<AutomationRule> =>
+    ipcRenderer.invoke('automation:createRule', draft),
+  updateAutomationRule: (id: string, draft: AutomationRuleDraft): Promise<AutomationRule | null> =>
+    ipcRenderer.invoke('automation:updateRule', { id, draft }),
+  deleteAutomationRule: (id: string): Promise<boolean> =>
+    ipcRenderer.invoke('automation:deleteRule', id),
+  setAutomationRuleEnabled: (id: string, enabled: boolean): Promise<AutomationRule | null> =>
+    ipcRenderer.invoke('automation:setRuleEnabled', { id, enabled }),
+  listAutomationExecutions: (query?: { limit?: number }): Promise<AutomationExecution[]> =>
+    ipcRenderer.invoke('automation:listExecutions', query),
+  clearAutomationExecutions: (): Promise<boolean> =>
+    ipcRenderer.invoke('automation:clearExecutions'),
+  /** 「在哪些聊天生效」的可选项。`id` 为 `xxx@chatroom`，与规则内 conversationIds 同口径。 */
+  listAutomationGroups: (): Promise<Array<{ id: string; name: string }>> =>
+    ipcRenderer.invoke('automation:listGroups'),
+  /**
+   * 保存「退群通知」规则（singleton upsert，id 由 main 侧固定，渲染层不拼 id）。
+   */
+  saveLeaveNotificationRule: (draft: AutomationRuleDraft): Promise<AutomationRule> =>
+    ipcRenderer.invoke('automation:saveLeaveNotificationRule', draft),
+  /**
+   * 「指定好友」的可选项。已在 main 侧过滤掉群聊 / 公众号 / 文件传输助手 / 自己。
+   */
+  listSendableContacts: (): Promise<Array<{ id: string; name: string }>> =>
+    ipcRenderer.invoke('automation:listSendableContacts'),
  onGroupExitMonitorState: (callback: (state: GroupExitMonitorState) => void) => {
    const listener = (_event: Electron.IpcRendererEvent, state: GroupExitMonitorState): void =>
      callback(state)
@@ -519,10 +551,6 @@ const api = {
    ipcRenderer.invoke('wechat-personal:getStatus'),
  getPersonalWechatSendCapability: (): Promise<PersonalWechatSendCapability> =>
    ipcRenderer.invoke('wechat-personal:getSendCapability'),
-  getPersonalWechatKeepOneBotProcess: (): Promise<boolean> =>
-    ipcRenderer.invoke('wechat-personal:getKeepProcess'),
-  setPersonalWechatKeepOneBotProcess: (keep: boolean): Promise<boolean> =>
-    ipcRenderer.invoke('wechat-personal:setKeepProcess', keep),
  checkPersonalWechatSenderStatus: (port?: string): Promise<PersonalWechatSenderStatus> =>
    ipcRenderer.invoke('wechat-personal:checkStatus', port),
  checkPersonalWechatVoiceEncodingEnvironment:
@@ -534,26 +562,6 @@ const api = {
    ipcRenderer.invoke('wechat-personal:openVoicePythonDownload'),
  openPersonalWechatVoiceFfmpegDownload: (): Promise<{ success: boolean; error?: string }> =>
    ipcRenderer.invoke('wechat-personal:openVoiceFfmpegDownload'),
-  getPersonalWechatRuntimeStatus: (): Promise<PersonalWechatRuntimeStatus> =>
-    ipcRenderer.invoke('wechat-personal:getRuntimeStatus'),
-  downloadPersonalWechatRuntime: (): Promise<PersonalWechatRuntimeDownloadResult> =>
-    ipcRenderer.invoke('wechat-personal:downloadRuntime'),
-  cancelPersonalWechatRuntimeDownload: (): Promise<{ success: boolean }> =>
-    ipcRenderer.invoke('wechat-personal:cancelRuntimeDownload'),
-  removePersonalWechatRuntime: (): Promise<PersonalWechatRuntimeStatus> =>
-    ipcRenderer.invoke('wechat-personal:removeRuntime'),
-  openPersonalWechatRuntimeDirectory: (): Promise<{ success: boolean; error?: string }> =>
-    ipcRenderer.invoke('wechat-personal:openRuntimeDirectory'),
-  onPersonalWechatRuntimeProgress: (
-    callback: (status: PersonalWechatRuntimeProgressEvent) => void
-  ) => {
-    const listener = (
-      _event: Electron.IpcRendererEvent,
-      status: PersonalWechatRuntimeProgressEvent
-    ): void => callback(status)
-    ipcRenderer.on('wechat-personal:runtimeProgress', listener)
-    return () => ipcRenderer.removeListener('wechat-personal:runtimeProgress', listener)
-  },
  rebindPersonalWechatSender: (): Promise<PersonalWechatSenderStatus> =>
    ipcRenderer.invoke('wechat-personal:rebind'),
  sendGeneratedTtsVoice: (
@@ -564,44 +572,30 @@ const api = {
  sendPersonalWechatMessage: (
    request: PersonalWechatSendRequest
  ): Promise<PersonalWechatSendResult> => ipcRenderer.invoke('wechat-personal:send', request),
-  listScheduledReports: (): Promise<ScheduledReportTask[]> =>
-    ipcRenderer.invoke('scheduled-report:list'),
-  listScheduledReportExecutions: (taskId?: string): Promise<ScheduledReportExecution[]> =>
-    ipcRenderer.invoke('scheduled-report:listExecutions', taskId),
+  /**
+   * 定时日报（Automation 的 `scheduled_report` 规则类型）。
+   *
+   * 规则 CRUD 复用上面的 `automation:*` 通道；这里只暴露三块专属能力：
+   * 立即执行、微信异常通知、旧执行记录只读存档。
+   */
+  runScheduledReportRule: (
+    ruleId: string
+  ): Promise<{ success: boolean; error?: string; data?: ScheduledRuleRunOutcome }> =>
+    ipcRenderer.invoke('automation:runScheduledReportRule', ruleId),
+  listScheduledReportLegacyExecutions: (ruleId?: string): Promise<ScheduledReportExecution[]> =>
+    ipcRenderer.invoke('automation:listScheduledReportLegacyExecutions', ruleId),
  getScheduledReportNotificationSettings: (): Promise<ScheduledReportNotificationSettings> =>
-    ipcRenderer.invoke('scheduled-report:getNotificationSettings'),
+    ipcRenderer.invoke('automation:getScheduledReportNotificationSettings'),
+  getScheduledReportNotificationCapability: (): Promise<ScheduledReportNotificationCapability> =>
+    ipcRenderer.invoke('automation:getScheduledReportNotificationCapability'),
  setScheduledReportNotificationEnabled: (
    enabled: boolean
  ): Promise<ScheduledReportNotificationSettingsResult> =>
-    ipcRenderer.invoke('scheduled-report:setNotificationEnabled', enabled),
-  createScheduledReport: (
-    request: ScheduledReportCreateInput
-  ): Promise<ScheduledReportResult<ScheduledReportTask>> =>
-    ipcRenderer.invoke('scheduled-report:create', request),
-  updateScheduledReport: (
-    taskId: string,
-    request: ScheduledReportUpdateInput
-  ): Promise<ScheduledReportResult<ScheduledReportTask>> =>
-    ipcRenderer.invoke('scheduled-report:update', taskId, request),
-  deleteScheduledReport: (taskId: string): Promise<ScheduledReportResult<{ deletedId: string }>> =>
-    ipcRenderer.invoke('scheduled-report:delete', taskId),
-  setScheduledReportEnabled: (
-    taskId: string,
-    enabled: boolean
-  ): Promise<ScheduledReportResult<ScheduledReportTask>> =>
-    ipcRenderer.invoke('scheduled-report:setEnabled', taskId, enabled),
-  runScheduledReportNow: (
-    taskId: string
-  ): Promise<ScheduledReportResult<ScheduledReportExecution>> =>
-    ipcRenderer.invoke('scheduled-report:runNow', taskId),
-  retryScheduledReportSend: (
-    executionId: string
-  ): Promise<ScheduledReportResult<ScheduledReportExecution>> =>
-    ipcRenderer.invoke('scheduled-report:retrySend', executionId),
+    ipcRenderer.invoke('automation:setScheduledReportNotificationEnabled', enabled),
  testScheduledReportErrorNotification: (
-    taskId: string
-  ): Promise<ScheduledReportResult<ScheduledReportExecution>> =>
-    ipcRenderer.invoke('scheduled-report:testErrorNotification', taskId),
+    ruleId: string
+  ): Promise<ScheduledReportResult<ScheduledReportNotification>> =>
+    ipcRenderer.invoke('automation:testScheduledReportErrorNotification', ruleId),
  getPersonalWechatVoiceDiagnostic: (): Promise<PersonalWechatVoiceDiagnostic | null> =>
    ipcRenderer.invoke('wechat-personal:getVoiceDiagnostic'),
  getAgentHubStatus: () => ipcRenderer.invoke('agent-hub:getStatus'),
@@ -18,7 +18,6 @@ import { ReportInfoPanel } from './components/reports/ReportInfoPanel'
 import { ReportSourceSidebar } from './components/reports/ReportSourceSidebar'
 import { ReportTaskStatusPanel } from './components/reports/ReportTaskStatusPanel'
 import { ReportViewer } from './components/reports/ReportViewer'
-import { ScheduledReportsWorkspace } from './components/reports/ScheduledReportsWorkspace'
 import { ReportTemplateMarketWorkspace } from './components/reports/ReportTemplateMarketWorkspace'
 import { contactDisplayName } from './components/reports/types'
 import type { GeneratedReportRecord, ReportWorkspaceView } from './components/reports/types'
@@ -42,10 +41,11 @@ import { isRelevantMessageMonitorEvent, parseWcdbMonitorEvent } from './utils/me
 import { enrichQuotedMessages } from './utils/quoted-messages'
 import type { ReportTemplateSelectionId } from '../../shared/report-templates'
 import { switchGeneratedReportTemplate } from './utils/report-template-switch'
-import { runtimePlatform, supportsPersonalWechatSend } from './utils/runtime-environment'
-import { useToast } from './components/ui'
+import { runtimePlatform } from './utils/runtime-environment'
+import { Button, useToast } from './components/ui'
 import { AppUpdatePrompt } from './features/app-update/AppUpdatePrompt'
-import { GroupExitMonitorWorkspace } from './features/group-exit-monitor/GroupExitMonitorWorkspace'
+import { GroupExitMonitorWorkspace, type GroupExitMonitorOpenViewRequest } from './features/group-exit-monitor/GroupExitMonitorWorkspace'
+import { AutomationWorkspace, type AutomationOpenRuleRequest } from './features/automation/AutomationWorkspace'
 import { selectContactAvatarRefreshUsernames } from './utils/contact-avatar'
 import {
  buildContactSearchIndex,
@@ -257,6 +257,17 @@ function App(): React.ReactElement {
  const [databaseEnvironment, setDatabaseEnvironment] = useState<DatabaseKeyEnvironment>()
  const connectionOperationRef = React.useRef(0)
  const [activePage, setActivePage] = useState<AppPage>('archive')
+  /**
+   * 「退群监控 ⇄ 自动化」之间的跨页深链。
+   *
+   * 项目没有 react-router，一级菜单就是 `activePage` 这一份 state，
+   * 所以深链同样是 state —— 用一个单调递增的 requestId 表达"又点了一次"，
+   * 子页面据此重新落位（同一个对象引用不会重复触发 effect）。
+   */
+  const [automationOpenRuleRequest, setAutomationOpenRuleRequest] =
+    React.useState<AutomationOpenRuleRequest | null>(null)
+  const [exitMonitorOpenViewRequest, setExitMonitorOpenViewRequest] =
+    React.useState<GroupExitMonitorOpenViewRequest | null>(null)
  const [archiveJumpTime, setArchiveJumpTime] = useState<number | null>(null)
  /**
   * 精确跳转目标（规范化后的消息 id）。
@@ -268,7 +279,7 @@ function App(): React.ReactElement {
  const [settingsCategory, setSettingsCategory] = useState<SettingsCategoryId>('account-database')
  const [reportSourceContact, setReportSourceContact] = useState<Contact | null>(null)
  const [reportWorkspaceView, setReportWorkspaceView] = useState<ReportWorkspaceView>('result')
-  const [reportSection, setReportSection] = useState<'today' | 'scheduled' | 'market'>('today')
+  const [reportSection, setReportSection] = useState<'today' | 'market'>('today')
  const [generatedReports, setGeneratedReports] = useState<GeneratedReportRecord[]>([])
  const [selectedReportId, setSelectedReportId] = useState<string | null>(null)
  const [latestGeneratedReportId, setLatestGeneratedReportId] = useState<string | null>(null)
@@ -1639,10 +1650,49 @@ function App(): React.ReactElement {
    setActivePage('settings')
  }

+  /**
+   * 跳到「设置 · 本地索引」。
+   *
+   * 用在群统计这类场景：结果不完整的原因是索引没追平，用户需要一个**可操作的去处**，
+   * 而不是只看到一句「结果可能不完整」却不知道去哪解决。
+   */
+  const openLocalIndexSettings = (): void => {
+    setSettingsCategory('local-index')
+    setActivePage('settings')
+  }
+
  const openAgentHub = (): void => {
    setActivePage('agent-hub')
  }

+  /**
+   * 「退群监控 → 退群通知自动化」。
+   *
+   * 直接落到「自动化 → 规则 → 退群通知」，不是只跳到自动化首页。
+   * 这是纯导航：不改退群监控的任何配置，也不触发任何发送。
+   */
+  const openLeaveNotificationAutomation = (): void => {
+    setAutomationOpenRuleRequest({ ruleType: 'leave_notification', requestId: Date.now() })
+    setActivePage('automation')
+  }
+
+  /**
+   * 「定时日报」已正式迁入自动化，日报页只保留「今日日报 | 社区模板市场」两个 Tab。
+   *
+   * 这里仍然留一个导航入口：用户是从日报页产生"要定时发日报"这个念头的，
+   * 不给路会显得功能被删了。它**不是**第三个 Tab，只是一句去处的说明。
+   */
+  const openScheduledReportAutomation = (): void => {
+    setAutomationOpenRuleRequest({ ruleType: 'scheduled_report', requestId: Date.now() })
+    setActivePage('automation')
+  }
+
+  /** 「自动化 → 退群通知 → 管理监控群聊 / 查看群聊」：回到退群监控的管理群聊页。 */
+  const openExitMonitorGroups = (): void => {
+    setExitMonitorOpenViewRequest({ view: 'manage', requestId: Date.now() })
+    setActivePage('exit-monitor')
+  }
+
  const dismissFirstUseWelcome = (): void => {
    try {
      localStorage.setItem(FIRST_USE_WELCOME_SEEN_KEY, '1')
@@ -1901,12 +1951,12 @@ function App(): React.ReactElement {
        contentFilter={contentFilter}
        onContentFilterChange={setContentFilter}
        onRefresh={() => selectedContact && handleSelectContact(selectedContact, true)}
-        onRefreshData={loadContacts}
        onReloadAvatars={handleReloadCurrentAvatars}
        onLoadOlderMessages={handleLoadOlderMessages}
        onCreateGroupReport={handleOpenReportWorkspace}
        onOpenTextToSpeechSettings={openTextToSpeechSettings}
        onOpenPersonalWechatSettings={openWechatSendSettings}
+        onOpenLocalIndexSettings={openLocalIndexSettings}
        isAiLoading={reportGeneration.isGenerating}
        jumpToTime={archiveJumpTime}
        jumpToMessageId={archiveJumpMessageId}
@@ -1926,22 +1976,6 @@ function App(): React.ReactElement {
        >
          今日日报
        </button>
-        <button
-          type="button"
-          role="tab"
-          aria-selected={reportSection === 'scheduled'}
-          aria-disabled={!supportsPersonalWechatSend}
-          className={`${reportSection === 'scheduled' ? 'active' : ''} ${!supportsPersonalWechatSend ? 'unsupported' : ''}`}
-          onClick={() => {
-            if (!supportsPersonalWechatSend) {
-              toast({ description: '定时日报目前仅支持 macOS 和 Windows。', duration: 3200 })
-              return
-            }
-            setReportSection('scheduled')
-          }}
-        >
-          定时日报{!supportsPersonalWechatSend && <small>仅 macOS / Windows</small>}
-        </button>
        <button
          type="button"
          role="tab"
@@ -1952,23 +1986,26 @@ function App(): React.ReactElement {
          社区模板市场
        </button>
      </div>
+      {/*
+        「定时日报已并入自动化」的指引条。
+
+        这里以前是裸 `<p>` + 裸 `<button>`，并且挂了 `.report-workspace-hint` /
+        `.report-workspace-hint-link` 两个**样式表里根本不存在**的类 ——
+        于是就成了一条没样式的文字 + 一个长得不像系统里任何按钮的按钮。
+        现在补上真实样式，按钮统一走 UI 组件库。
+      */}
+      <div className="report-workspace-hint">
+        <span>需要「定时日报」？它已经并入「自动化」，可以按时间自动生成并发送日报。</span>
+        <Button variant="link" size="sm" onClick={openScheduledReportAutomation}>
+          去自动化配置 →
+        </Button>
+      </div>
      <div className="report-workspace-body">
        {reportSection === 'market' ? (
          <ReportTemplateMarketWorkspace
            value={reportGeneration.templateId}
            onChange={reportGeneration.setTemplateId}
          />
-        ) : reportSection === 'scheduled' ? (
-          <ScheduledReportsWorkspace
-            contacts={contacts}
-            platformSupported={supportsPersonalWechatSend}
-            onOpenWechatSettings={openWechatSendSettings}
-            onOpenAgentHub={openAgentHub}
-            onOpenModelSettings={openModelSettings}
-            onNotice={(message, variant) =>
-              toast({ description: message, variant, duration: 3200 })
-            }
-          />
        ) : reportWorkspaceView === 'result' ? (
          <div className="report-center-page">
            <ReportHistorySidebar
@@ -2084,6 +2121,20 @@ function App(): React.ReactElement {
            dbReady={isDatabaseConnected}
            contacts={contacts}
            onOpenSendSettings={openWechatSendSettings}
+            openViewRequest={exitMonitorOpenViewRequest}
+            onOpenViewRequestHandled={() => setExitMonitorOpenViewRequest(null)}
+            onOpenLeaveNotificationAutomation={openLeaveNotificationAutomation}
+          />
+        )
+      case 'automation':
+        return (
+          <AutomationWorkspace
+            dbReady={isDatabaseConnected}
+            onOpenSendSettings={openWechatSendSettings}
+            onOpenExitMonitorGroups={openExitMonitorGroups}
+            openRuleRequest={automationOpenRuleRequest}
+            onOpenModelSettings={openModelSettings}
+            onOpenAgentHub={openAgentHub}
          />
        )
      case 'agent-hub':
@@ -7,6 +7,8 @@ import { DataTrustBar } from './chat/DataTrustBar'
 import { EmptyConversationState } from './chat/EmptyConversationState'
 import { MessageList } from './chat/MessageList'
 import { PersonalWechatSendDialog } from './chat/PersonalWechatSendDialog'
+import { mergeGroupExitEvents } from '../../../shared/group-exit-event-message'
+import { useGroupExitEvents } from '../hooks/useGroupExitEvents'

 interface ChatWindowProps {
  contact: Contact | null
@@ -16,12 +18,13 @@ interface ChatWindowProps {
  contentFilter?: string
  onContentFilterChange?: (keyword: string) => void
  onRefresh?: () => void
-  onRefreshData?: () => void
  onReloadAvatars?: () => Promise<void>
  onLoadOlderMessages?: () => Promise<void>
  onCreateGroupReport?: () => void
  onOpenTextToSpeechSettings?: () => void
  onOpenPersonalWechatSettings?: () => void
+  /** 跳到「设置 · 本地索引」（群统计发现索引没追平时用）。 */
+  onOpenLocalIndexSettings?: () => void
  isAiLoading?: boolean
  jumpToTime?: number | null
  /** 精确跳转目标（消息 id）。与 jumpToTime 取或：任一存在就说明"这是一次跳转"。 */
@@ -36,12 +39,12 @@ const ChatWindow: React.FC<ChatWindowProps> = ({
  contentFilter,
  onContentFilterChange,
  onRefresh,
-  onRefreshData,
  onReloadAvatars,
  onLoadOlderMessages,
  onCreateGroupReport,
  onOpenTextToSpeechSettings,
  onOpenPersonalWechatSettings,
+  onOpenLocalIndexSettings,
  isAiLoading = false,
  jumpToTime,
  jumpToMessageId
@@ -122,8 +125,20 @@ const ChatWindow: React.FC<ChatWindowProps> = ({
    }
  }

+  /**
+   * 退群推断事件并入档案消息流。
+   *
+   * **先合并再过滤**：反过来的话搜索词就命中不了事件，而「谁退群了」
+   * 恰恰是最需要能被搜到的内容之一。
+   */
+  const groupExitEvents = useGroupExitEvents(contact)
+  const allMessages = React.useMemo(
+    () => (groupExitEvents.length ? mergeGroupExitEvents(messages, groupExitEvents) : messages),
+    [messages, groupExitEvents]
+  )
+
  const filteredMessages = React.useMemo(() => {
-    return messages.filter((msg) => {
+    return allMessages.filter((msg) => {
      const filterTypes = (import.meta.env.VITE_FILTER_MSG_TYPES || '')
        .split(',')
        .map((type) => type.trim())
@@ -132,7 +147,7 @@ const ChatWindow: React.FC<ChatWindowProps> = ({
      const contentMatch = !contentFilter || msg.content.includes(contentFilter)
      return typeMatch && contentMatch
    })
-  }, [messages, contentFilter])
+  }, [allMessages, contentFilter])

  const handleOpenPersonalWechatSend = useCallback(async (): Promise<void> => {
    try {
@@ -165,9 +180,9 @@ const ChatWindow: React.FC<ChatWindowProps> = ({
        isAiLoading={isAiLoading}
        onContentFilterChange={onContentFilterChange || (() => undefined)}
        onRefresh={onRefresh}
-        onRefreshData={onRefreshData}
        onTestSend={() => void handleOpenPersonalWechatSend()}
        onOpenAiSettings={onCreateGroupReport || (() => undefined)}
+        onOpenLocalIndexSettings={onOpenLocalIndexSettings}
      />
      <DataTrustBar messageCount={messages.length} />
      <MessageList
@@ -26,11 +26,7 @@ export function AccountSummary({
  const displayName =
    showAccount && selfInfo ? selfInfo.nickname || selfInfo.wxid || '当前账号' : '未连接'
  const subtitle = showAccount && selfInfo ? selfInfo.wxid : '打开设置'
-  const statusText = dbReady
-    ? '数据库已连接'
-    : dbConnecting
-      ? '正在连接数据库'
-      : '数据库未连接'
+  const statusText = dbReady ? '数据库已连接' : dbConnecting ? '正在连接数据库' : '数据库未连接'
  const statusClass = dbReady ? 'ready' : dbConnecting ? 'connecting' : ''
  const initial = (displayName || '?').charAt(0)
  const title = `${displayName}\n${subtitle}`
@@ -66,8 +62,8 @@ export function AccountSummary({
      </span>
      <button type="button" className="account-summary-settings" onClick={onClick} title="设置">
        <svg viewBox="0 0 24 24" aria-hidden="true" focusable="false">
-          <path d="M12 8.5a3.5 3.5 0 1 0 0 7 3.5 3.5 0 0 0 0-7Z" />
-          <path d="M19.4 15a8.2 8.2 0 0 0 .1-1l2-1.5-2-3.5-2.4 1a7.5 7.5 0 0 0-1.7-1l-.3-2.5h-4l-.4 2.5a7.5 7.5 0 0 0-1.7 1l-2.3-1-2 3.5 2 1.5a8.2 8.2 0 0 0 0 2l-2 1.5 2 3.5 2.3-1a7.5 7.5 0 0 0 1.7 1l.4 2.5h4l.3-2.5a7.5 7.5 0 0 0 1.7-1l2.4 1 2-3.5-2.1-1.5Z" />
+          <path d="M12.22 2h-.44a2 2 0 0 0-2 1.72l-.15 1.58a8 8 0 0 0-1.77 1.03l-1.48-.6a2 2 0 0 0-2.5.88l-.22.39a2 2 0 0 0 .51 2.68l1.25.98a8 8 0 0 0 0 2.08l-1.25.98a2 2 0 0 0-.51 2.68l.22.39a2 2 0 0 0 2.5.88l1.48-.6c.54.43 1.14.78 1.77 1.03l.15 1.58a2 2 0 0 0 2 1.72h.44a2 2 0 0 0 2-1.72l.15-1.58c.63-.25 1.23-.6 1.77-1.03l1.48.6a2 2 0 0 0 2.5-.88l.22-.39a2 2 0 0 0-.51-2.68l-1.25-.98a8 8 0 0 0 0-2.08l1.25-.98a2 2 0 0 0 .51-2.68l-.22-.39a2 2 0 0 0-2.5-.88l-1.48.6a8 8 0 0 0-1.77-1.03l-.15-1.58a2 2 0 0 0-2-1.72Z" />
+          <circle cx="12" cy="12" r="3" />
        </svg>
      </button>
    </div>
@@ -1,19 +1,10 @@
 import React, { useState } from 'react'
 import { Contact } from '../../../../shared/types'
-import {
-  Button,
-  DropdownMenu,
-  DropdownMenuContent,
-  DropdownMenuItem,
-  DropdownMenuTrigger,
-  IconButton,
-  Tooltip,
-  TooltipContent,
-  TooltipTrigger
-} from '../ui'
+import { Button, IconButton, Tooltip, TooltipContent, TooltipTrigger } from '../ui'
 import { ConversationContentSearch } from './ConversationContentSearch'
-import { AiIcon, MoreIcon, RefreshIcon, SearchIcon, SendIcon } from './icons'
+import { AiIcon, RefreshIcon, SearchIcon, SendIcon, StatsIcon } from './icons'
 import { supportsPersonalWechatSend } from '../../utils/runtime-environment'
+import { GroupMemberStatsDialog } from '../group-stats/GroupMemberStatsDialog'

 interface ChatHeaderProps {
  contact: Contact
@@ -24,9 +15,10 @@ interface ChatHeaderProps {
  isAiLoading: boolean
  onContentFilterChange: (value: string) => void
  onRefresh?: () => void
-  onRefreshData?: () => void
  onTestSend: () => void
  onOpenAiSettings: () => void
+  /** 跳到「设置 · 本地索引」（群统计发现索引没追平时用）。 */
+  onOpenLocalIndexSettings?: () => void
 }

 export function ChatHeader({
@@ -38,11 +30,12 @@ export function ChatHeader({
  isAiLoading,
  onContentFilterChange,
  onRefresh,
-  onRefreshData,
  onTestSend,
-  onOpenAiSettings
+  onOpenAiSettings,
+  onOpenLocalIndexSettings
 }: ChatHeaderProps): React.ReactElement {
  const [searchOpen, setSearchOpen] = useState(Boolean(contentFilter))
+  const [statsOpen, setStatsOpen] = useState(false)
  const displayName = contact.m_nsNickName || contact.m_nsUsrName || '未命名会话'
  const typeLabel = isGroupChat ? '群聊' : '联系人'
  const visibleCount = contentFilter ? filteredCount : loadedCount
@@ -91,16 +84,20 @@ export function ChatHeader({
        <IconButton label="刷新聊天记录" variant="ghost" className="h-8 w-8" onClick={onRefresh}>
          <RefreshIcon />
        </IconButton>
-        <DropdownMenu>
-          <DropdownMenuTrigger asChild>
-            <IconButton label="更多" tooltip="" variant="ghost" className="h-8 w-8">
-              <MoreIcon />
-            </IconButton>
-          </DropdownMenuTrigger>
-          <DropdownMenuContent align="end">
-            <DropdownMenuItem onSelect={() => onRefreshData?.()}>刷新数据</DropdownMenuItem>
-          </DropdownMenuContent>
-        </DropdownMenu>
+        {/* 群发言统计只对群聊有意义；单聊没有「成员名单」这个概念。 */}
+        {isGroupChat ? (
+          <Button
+            variant="outline"
+            size="sm"
+            className="chat-header-text-action"
+            aria-label="群发言统计"
+            title="群发言统计"
+            onClick={() => setStatsOpen(true)}
+          >
+            <StatsIcon />
+            <span>群发言统计</span>
+          </Button>
+        ) : null}
        {supportsPersonalWechatSend ? (
          <Button
            variant="outline"
@@ -148,6 +145,14 @@ export function ChatHeader({
          <span>{isAiLoading ? '生成中' : '生成 AI 日报'}</span>
        </Button>
      </div>
+      {isGroupChat ? (
+        <GroupMemberStatsDialog
+          open={statsOpen}
+          onOpenChange={setStatsOpen}
+          contact={contact}
+          onOpenLocalIndexSettings={onOpenLocalIndexSettings}
+        />
+      ) : null}
    </div>
  )
 }
@@ -0,0 +1,27 @@
+import React from 'react'
+import type { Message } from '../../../../shared/types'
+
+interface GroupExitEventBubbleProps {
+  message: Message
+}
+
+/**
+ * 退群推断事件在档案里的展示。
+ *
+ * ⚠️ 与真实微信系统消息**必须视觉可区分**，原因不是审美：
+ * 这条内容**不是微信说的话**，而是 TraceMemo 通过「群成员快照前后对比」推断出来的。
+ * 两者的可信度完全不同 —— 微信系统消息是事实原文，这条是推断产物，
+ * 而且它记录的是**检测时刻**（可能比真实退群时间晚几分钟到几天）。
+ *
+ * 所以它带一个「本地推断」标记，且样式与 `wechat-system-message` 明确区分。
+ */
+export function GroupExitEventBubble({ message }: GroupExitEventBubbleProps): React.ReactElement {
+  return (
+    <div className="group-exit-event-row">
+      <div className="group-exit-event-bubble">
+        <span className="group-exit-event-tag">本地推断</span>
+        <span className="group-exit-event-text">{message.content}</span>
+      </div>
+    </div>
+  )
+}
@@ -1,5 +1,7 @@
 import React from 'react'
 import { Contact } from '../../../../shared/types'
+import { isGroupExitEventMessage } from '../../../../shared/group-exit-event-message'
+import { GroupExitEventBubble } from './GroupExitEventBubble'
 import { MessageBubble } from './MessageBubble'
 import { MessageGroupModel } from './messageGrouping'

@@ -26,11 +28,17 @@ export function MessageGroup({
    return (
      <>
        {group.timeLabel && <div className="chat-time-separator">{group.timeLabel}</div>}
-        {group.messages.map((message) => (
-          <div key={message.id} className="wechat-system-message-row">
-            <div className="wechat-system-message">{message.content}</div>
-          </div>
-        ))}
+        {group.messages.map((message) =>
+          // 退群推断事件与真实微信系统消息都落在 isSystem 分支，
+          // 但必须分开渲染：前者是本地推断，后者是微信原话。
+          isGroupExitEventMessage(message) ? (
+            <GroupExitEventBubble key={message.id} message={message} />
+          ) : (
+            <div key={message.id} className="wechat-system-message-row">
+              <div className="wechat-system-message">{message.content}</div>
+            </div>
+          )
+        )}
      </>
    )
  }
@@ -5,16 +5,14 @@ import type {
  PersonalWechatSenderStatus,
  PersonalWechatVoiceDiagnostic
 } from '../../../../shared/personal-wechat'
-import type {
-  PersonalWechatRuntimeProgressEvent,
-  PersonalWechatRuntimeStatus
-} from '../../../../shared/personal-wechat-runtime'
 import { Button, Dialog, DialogContent, DialogDescription, DialogHeader, DialogTitle } from '../ui'
 import { isMac, isWindows } from '../../utils/runtime-environment'
 import { PersonalWechatChatComposer, type ChatMessage } from './PersonalWechatChatComposer'
 import { PersonalWechatSetupGuide } from './PersonalWechatSetupGuide'
 import { PersonalWechatVoiceDiagnosticDialog } from './PersonalWechatVoiceDiagnosticDialog'
 import { PersonalWechatWindowsSendDialog } from './PersonalWechatWindowsSendDialog'
+import { ReportImagePostfixInput } from './ReportImagePostfixInput'
+import { useReportImagePostfixSetting } from './useReportImagePostfixSetting'

 type SelectedLocalFile = { path: string; name: string }

@@ -36,7 +34,7 @@ function fallbackStatus(error: unknown): PersonalWechatSenderStatus {
    sipDisabled: false,
    wechatRunning: false,
    runtimeReady: false,
-    endpoint: '127.0.0.1:58080',
+    endpoint: '',
    endpointReady: false,
    attachReady: false,
    baseAddressReady: false,
@@ -67,18 +65,14 @@ function PersonalWechatMacSendDialog({
  initialImage = null
 }: PersonalWechatSendDialogProps): React.ReactElement {
  const [senderStatus, setSenderStatus] = useState<PersonalWechatSenderStatus | null>(null)
-  const [runtimeStatus, setRuntimeStatus] = useState<PersonalWechatRuntimeStatus | null>(null)
-  const [runtimeProgress, setRuntimeProgress] = useState<PersonalWechatRuntimeProgressEvent | null>(
-    null
-  )
  const [detecting, setDetecting] = useState(true)
  const [binding, setBinding] = useState(false)
-  const [runtimeBusy, setRuntimeBusy] = useState(false)
  const [sendBusy, setSendBusy] = useState(false)
-  // 状态可能来自之前的 OneBot 进程或日志；发送入口只信任当前语音能力状态。
+  // 发送入口只信任当前 native runtime 上报的能力状态。
  const [sessionBound, setSessionBound] = useState(false)
  const [messages, setMessages] = useState<ChatMessage[]>([])
  const [sendError, setSendError] = useState<string | null>(null)
+  const [sendSuccess, setSendSuccess] = useState<string | null>(null)
  const [voiceDiagnostic, setVoiceDiagnostic] = useState<PersonalWechatVoiceDiagnostic | null>(null)
  const [voiceDiagnosticOpen, setVoiceDiagnosticOpen] = useState(false)
  const requestIdRef = useRef(0)
@@ -86,23 +80,19 @@ function PersonalWechatMacSendDialog({
  const closingRef = useRef(false)
  const displayName = contact.m_nsNickName || contact.m_nsUsrName || '未命名会话'
  const targetId = contact.m_nsUsrName
-  const isBusy = binding || runtimeBusy || sendBusy
+  const isBusy = binding || sendBusy
  const setupReady = Boolean(initialImage ? senderStatus?.canSendImage : senderStatus?.canSendVoice)
+  const { postfixText, setPostfixText, persistPostfixText } = useReportImagePostfixSetting(
+    Boolean(initialImage)
+  )

  const refreshStatus = useCallback(async (): Promise<void> => {
    const requestId = ++requestIdRef.current
    setDetecting(true)
    setSendError(null)
    try {
-      const [nextRuntime, nextSender] = await Promise.all([
-        isMac
-          ? window.api.getPersonalWechatRuntimeStatus?.() || Promise.resolve(null)
-          : Promise.resolve(null),
-        window.api.getPersonalWechatSenderStatus()
-      ])
+      const nextSender = await window.api.getPersonalWechatSenderStatus()
      if (requestId !== requestIdRef.current) return
-      setRuntimeStatus(nextRuntime)
-      setRuntimeProgress(nextRuntime?.state === 'downloading' ? nextRuntime : null)
      setSenderStatus(nextSender)
      setSessionBound(
        nextSender.state === 'online' ||
@@ -122,13 +112,6 @@ function PersonalWechatMacSendDialog({

  useEffect(() => {
    void refreshStatus()
-    if (!isMac) return undefined
-    const unsubscribe = window.api.onPersonalWechatRuntimeProgress?.((status) => {
-      setRuntimeProgress(status)
-      setRuntimeStatus(status)
-      if (status.state === 'ready') void refreshStatus()
-    })
-    return unsubscribe
  }, [refreshStatus])

  useEffect(() => {
@@ -146,22 +129,6 @@ function PersonalWechatMacSendDialog({
    return () => window.clearInterval(timer)
  }, [refreshStatus, senderStatus])

-  const handleDownloadRuntime = async (): Promise<void> => {
-    if (runtimeBusy) return
-    setRuntimeBusy(true)
-    setSendError(null)
-    try {
-      const result = await window.api.downloadPersonalWechatRuntime()
-      setRuntimeStatus(result.status)
-      if (!result.success && result.error) setSendError(result.error)
-      if (result.success) await refreshStatus()
-    } catch (error) {
-      setSendError(error instanceof Error ? error.message : String(error))
-    } finally {
-      setRuntimeBusy(false)
-    }
-  }
-
  const handleBind = async (): Promise<void> => {
    if (binding) return
    setBinding(true)
@@ -232,18 +199,27 @@ function PersonalWechatMacSendDialog({
    if (!initialImage || !senderStatus?.canSendImage || sendBusy) return
    setSendBusy(true)
    setSendError(null)
+    setSendSuccess(null)
    try {
+      await persistPostfixText()
      const response = await window.api.sendPersonalWechatMessage({
        type: 'image',
        to: targetId,
        isGroup: isGroupChat,
-        filePath: initialImage.path
+        filePath: initialImage.path,
+        postfixText
      } satisfies PersonalWechatSendRequest)
      setSenderStatus(response.status)
      if (!response.success) {
        setSendError(response.error || '日报图片发送失败')
        return
      }
+      if (response.postfixError) {
+        setSendError(`日报图片已发送，但${response.postfixError}`)
+        setSendSuccess('日报图片发送成功')
+      } else {
+        setSendSuccess(postfixText.trim() ? '日报图片和后置词发送成功' : '日报图片发送成功')
+      }
      setMessages((current) => [
        ...current,
        {
@@ -287,7 +263,7 @@ function PersonalWechatMacSendDialog({
              {displayName.slice(0, 1)}
            </div>
            <div className="personal-wechat-chat-heading">
-              <DialogTitle>文字转语音</DialogTitle>
+              <DialogTitle>{initialImage ? '发送日报图片' : '文字转语音'}</DialogTitle>
              <DialogDescription>
                发送给 {displayName} · {setupReady ? '微信已连接' : '配置微信发送能力'}
              </DialogDescription>
@@ -319,14 +295,10 @@ function PersonalWechatMacSendDialog({

            {!setupReady && senderStatus && (
              <PersonalWechatSetupGuide
-                runtimeStatus={runtimeStatus}
                senderStatus={senderStatus}
-                runtimeProgress={runtimeProgress}
-                runtimeBusy={runtimeBusy}
                binding={binding}
                detecting={detecting}
                sessionBound={sessionBound}
-                onDownloadRuntime={() => void handleDownloadRuntime()}
                onBind={() => void handleBind()}
                onStartSending={() => undefined}
                onOpenTextToSpeechSettings={handleOpenSettings}
@@ -336,6 +308,16 @@ function PersonalWechatMacSendDialog({
            {setupReady && initialImage && (
              <section className="personal-wechat-composer" aria-label="日报图片发送">
                <p>已准备日报图片：{initialImage.name}</p>
+                <ReportImagePostfixInput
+                  value={postfixText}
+                  onChange={setPostfixText}
+                  onBlur={() =>
+                    void persistPostfixText().catch((error) =>
+                      setSendError(error instanceof Error ? error.message : '发送后置词保存失败')
+                    )
+                  }
+                  disabled={sendBusy}
+                />
                <Button size="sm" onClick={() => void handleSendReportImage()} disabled={sendBusy}>
                  {sendBusy ? '发送中…' : '发送日报图片'}
                </Button>
@@ -358,8 +340,13 @@ function PersonalWechatMacSendDialog({
                {sendError}
              </div>
            )}
+            {sendSuccess && (
+              <div className="personal-wechat-global-success" role="status">
+                {sendSuccess}
+              </div>
+            )}
          </div>
-          {setupReady && isMac && (
+          {setupReady && isMac && !initialImage && (
            <div className="personal-wechat-chat-footer flex items-center justify-end">
              <Button variant="link" size="sm" onClick={() => void handleOpenVoiceDiagnostic()}>
                语音发送诊断
@@ -1,22 +1,13 @@
-import { useState } from 'react'
 import type { PersonalWechatSenderStatus } from '../../../../shared/personal-wechat'
-import type {
-  PersonalWechatRuntimeProgressEvent,
-  PersonalWechatRuntimeStatus
-} from '../../../../shared/personal-wechat-runtime'
-import { Button, Dialog, DialogContent, DialogDescription, DialogHeader, DialogTitle } from '../ui'
-import { isMac } from '../../utils/runtime-environment'
-import { PersonalWechatSupportedVersionsContent } from './PersonalWechatSupportedVersionsContent'
+import { Button } from '../ui'
+
+const GROUP_README_URL = 'https://github.com/Wxw-Gu/TraceMemo#-交流与反馈'

 interface PersonalWechatSetupGuideProps {
-  runtimeStatus: PersonalWechatRuntimeStatus | null
  senderStatus: PersonalWechatSenderStatus | null
-  runtimeProgress: PersonalWechatRuntimeProgressEvent | null
-  runtimeBusy: boolean
  binding: boolean
  detecting: boolean
  sessionBound: boolean
-  onDownloadRuntime: () => void
  onBind: () => void
  onStartSending: () => void
  onOpenTextToSpeechSettings?: () => void
@@ -27,21 +18,18 @@ function capabilityLabel(ready: boolean, initializing: boolean): string {
  return initializing ? '初始化中' : '未就绪'
 }

-function diagnosticValue(value: unknown): string {
-  if (value === undefined || value === null || value === '') return '未检测到'
-  return String(value)
-}
-
 function bindingHint(status: PersonalWechatSenderStatus): string {
-  if (status.state === 'wechat_not_running') return '请先启动并登录微信。'
+  if (status.state === 'wechat_not_running') {
+    return '请保持微信未登录窗口状态，点击“绑定微信”后，再点击微信窗口登录。'
+  }
  if (status.state === 'unsupported_platform') return '当前系统暂不支持个人微信发送。'
  if (status.state === 'unsupported_version')
-    return '当前微信版本暂不支持，请查看微信发送设置中的支持版本。'
+    return '当前微信版本暂不支持，请联系群主确认可用版本。'
  if (status.state === 'runtime_missing') {
-    return isMac ? '请先完成 OneBot 运行时准备。' : '请先完成发送运行时准备。'
+    return '请先完成发送运行时准备。'
  }
  if (status.state === 'error') return '连接微信时遇到问题，请稍后重试。'
-  return '请启动并登录当前微信，TraceMemo 会自动绑定正在使用的账号。'
+  return '请保持微信未登录窗口状态，点击“绑定微信”后，再点击微信窗口登录。'
 }

 function isWechatBound(status: PersonalWechatSenderStatus): boolean {
@@ -54,28 +42,21 @@ function isWechatBound(status: PersonalWechatSenderStatus): boolean {
 }

 export function PersonalWechatSetupGuide({
-  runtimeStatus,
  senderStatus,
-  runtimeProgress,
-  runtimeBusy,
  binding,
  detecting,
  sessionBound,
-  onDownloadRuntime,
  onBind,
  onStartSending,
  onOpenTextToSpeechSettings
 }: PersonalWechatSetupGuideProps): React.ReactElement {
-  const [showSupportedVersions, setShowSupportedVersions] = useState(false)
-  const runtimeLabel = isMac ? 'OneBot 运行时' : '发送运行时'
-  const runtimeReady = runtimeStatus?.state === 'ready' || senderStatus?.runtimeReady === true
-  const runtimeDownloading = runtimeBusy || runtimeStatus?.state === 'downloading'
+  const runtimeUnavailable = senderStatus?.state === 'runtime_missing'
+  const runtimeLabel = '发送运行时'
+  const runtimeReady = senderStatus?.runtimeReady === true
  const connected = sessionBound && (senderStatus ? isWechatBound(senderStatus) : false)
  const canSendVoice = Boolean(senderStatus?.canSendVoice)
  const initializing = connected && !canSendVoice && senderStatus?.state !== 'error'
  const allReady = connected && Boolean(senderStatus?.canSend)
-  const progress = runtimeProgress || runtimeStatus
-  const progressPercent = Math.max(0, Math.min(100, Math.round((progress?.progress || 0) * 100)))

  return (
    <section className="personal-wechat-setup" aria-label="微信消息功能配置">
@@ -89,38 +70,21 @@ export function PersonalWechatSetupGuide({
      </div>

      <ol className="personal-wechat-steps">
-        <li
-          className={runtimeReady ? 'is-complete' : runtimeDownloading ? 'is-active' : 'is-current'}
-        >
+        <li className={runtimeReady ? 'is-complete' : 'is-current'}>
          <span className="personal-wechat-step-number">{runtimeReady ? '✓' : '1'}</span>
          <div className="personal-wechat-step-content">
            <strong>准备 {runtimeLabel}</strong>
-            <p>个人微信发送需要 {runtimeLabel}，首次使用时下载一次即可。</p>
-            {runtimeDownloading ? (
-              <div className="personal-wechat-download-progress" aria-live="polite">
-                <div>
-                  <span>正在准备 {runtimeLabel}</span>
-                  <span>{progressPercent}%</span>
-                </div>
-                <div className="personal-wechat-progress-track">
-                  <span style={{ width: `${progressPercent}%` }} />
-                </div>
-              </div>
-            ) : runtimeReady ? (
-              <span className="personal-wechat-step-status">✓ {runtimeLabel}已准备</span>
+            <p>
+              {runtimeUnavailable
+                ? `个人微信发送需要${runtimeLabel}，授权后即可使用。`
+                : `${runtimeLabel}随 TraceMemo 一起提供，无需额外安装。`}
+            </p>
+            {runtimeReady ? (
+              <span className="personal-wechat-step-status">✓ {runtimeLabel}已就绪</span>
+            ) : runtimeUnavailable ? (
+              <span className="personal-wechat-step-error">当前版本暂未提供微信消息发送功能</span>
            ) : (
-              <Button
-                size="sm"
-                onClick={onDownloadRuntime}
-                disabled={runtimeBusy || runtimeStatus?.state === 'unsupported'}
-              >
-                下载运行时
-              </Button>
-            )}
-            {runtimeStatus?.error && !runtimeDownloading && !runtimeReady && (
-              <p className="personal-wechat-step-error">
-                运行时准备失败，请重试或查看微信发送设置。
-              </p>
+              <span className="personal-wechat-step-status">正在检查 {runtimeLabel}…</span>
            )}
          </div>
        </li>
@@ -139,7 +103,7 @@ export function PersonalWechatSetupGuide({
          <span className="personal-wechat-step-number">{connected ? '✓' : '2'}</span>
          <div className="personal-wechat-step-content">
            <strong>绑定个人微信</strong>
-            <p>请启动并登录当前微信，TraceMemo 会自动绑定正在使用的账号。</p>
+            <p>请保持微信未登录窗口状态，点击“绑定微信”后，再点击微信窗口登录。</p>
            {connected ? (
              <span className="personal-wechat-step-status">✓ 微信已绑定</span>
            ) : (
@@ -147,15 +111,28 @@ export function PersonalWechatSetupGuide({
                size="sm"
                variant="outline"
                onClick={onBind}
-                disabled={!runtimeReady || binding || runtimeDownloading}
+                disabled={runtimeUnavailable || !runtimeReady || binding}
              >
                {binding ? '正在绑定…' : '绑定微信'}
              </Button>
            )}
-            {!connected && senderStatus && (
+            {!connected && senderStatus && !runtimeUnavailable && (
              <p className="personal-wechat-step-hint">{bindingHint(senderStatus)}</p>
            )}
-            {!connected && (
+            {!connected && runtimeUnavailable && (
+              <div className="personal-wechat-authorization-warning" role="alert">
+                <span aria-hidden>!</span>
+                <p>
+                  发送能力属授权制，需要联系群主。请先加入交流群，然后在群内添加群主申请授权。
+                  进群请点击{' '}
+                  <a href={GROUP_README_URL} target="_blank" rel="noreferrer">
+                    这里
+                  </a>{' '}
+                  跳转。
+                </p>
+              </div>
+            )}
+            {!connected && !runtimeUnavailable && (
              <p className="personal-wechat-step-warning" role="note">
                绑定微信可能导致当前微信异常闪退，这是正常现象。若微信退出，请重新启动微信后，再回到这里重新检测/绑定。
                <br />
@@ -163,14 +140,6 @@ export function PersonalWechatSetupGuide({
                通用”，取消勾选“有更新时自动升级微信”，否则版本变化后可能无法绑定。
              </p>
            )}
-            <Button
-              className="w-fit"
-              variant="link"
-              size="sm"
-              onClick={() => setShowSupportedVersions(true)}
-            >
-              查看支持的微信版本
-            </Button>
          </div>
        </li>

@@ -225,7 +194,7 @@ export function PersonalWechatSetupGuide({
        </li>
      </ol>

-      {!runtimeReady && runtimeStatus?.state === 'unsupported' && onOpenTextToSpeechSettings && (
+      {!runtimeReady && runtimeUnavailable && onOpenTextToSpeechSettings && (
        <Button
          variant="link"
          size="sm"
@@ -235,54 +204,6 @@ export function PersonalWechatSetupGuide({
          查看语音设置
        </Button>
      )}
-
-      <Dialog open={showSupportedVersions} onOpenChange={setShowSupportedVersions}>
-        <DialogContent className="max-h-[calc(100vh-3rem)] max-w-[620px] overflow-y-auto">
-          <DialogHeader className="pr-8">
-            <DialogTitle className="text-lg">支持的微信版本</DialogTitle>
-            <DialogDescription>请安装下列完整版本之一。</DialogDescription>
-          </DialogHeader>
-          <PersonalWechatSupportedVersionsContent />
-        </DialogContent>
-      </Dialog>
-
-      <details className="personal-wechat-diagnostics">
-        <summary>高级诊断</summary>
-        <p>仅用于排查连接问题，普通使用无需关注这些信息。</p>
-        <dl>
-          {[
-            ['微信进程', senderStatus?.wechatPid ? `PID ${senderStatus.wechatPid}` : '未检测到'],
-            ...(isMac
-              ? [['OneBot', senderStatus?.oneBotPid ? `PID ${senderStatus.oneBotPid}` : '未启动']]
-              : []),
-            ['绑定进程', diagnosticValue(senderStatus?.boundWechatPid)],
-            [
-              '接口监听',
-              `${diagnosticValue(senderStatus?.endpoint)} · ${senderStatus?.endpointReady ? '监听中' : '未监听'}`
-            ],
-            [
-              '基址扫描',
-              senderStatus?.baseAddress || (senderStatus?.baseAddressReady ? '已完成' : '未完成')
-            ],
-            ['文字 Hook', senderStatus?.textHookReady ? '已就绪' : '未就绪'],
-            ['图片 Hook', senderStatus?.imageHookReady ? '已就绪' : '未就绪'],
-            ['语音能力', senderStatus?.canSendVoice ? '可发送' : '未就绪'],
-            ['消息监听', senderStatus?.messageListenerReady ? '正常' : '未就绪'],
-            ['微信版本', diagnosticValue(senderStatus?.wechatVersion)],
-            [
-              '运行时',
-              runtimeStatus?.version
-                ? `${runtimeStatus.version} · ${runtimeStatus.state}`
-                : diagnosticValue(senderStatus?.runtimeReady)
-            ]
-          ].map(([label, value]) => (
-            <div key={label}>
-              <dt>{label}</dt>
-              <dd>{value}</dd>
-            </div>
-          ))}
-        </dl>
-      </details>
    </section>
  )
 }
@@ -1,46 +0,0 @@
-import type React from 'react'
-
-export const WECHAT_VERSION_DOWNLOAD_URL = 'https://github.com/zsbai/wechat-versions/releases'
-
-export const BUNDLED_WECHAT_VERSIONS = [
-  '4.1.6.12',
-  '4.1.6.46',
-  '4.1.6.47',
-  '4.1.7.31',
-  '4.1.7.55',
-  '4.1.7.57',
-  '4.1.8.28',
-  '4.1.8.29',
-  '4.1.8.104',
-  '4.1.8.107',
-  '4.1.9.52',
-  '4.1.9.55',
-  '4.1.9.58',
-  '4.1.10.53',
-  '4.1.11.53'
-] as const
-
-export function PersonalWechatSupportedVersionsContent(): React.ReactElement {
-  return (
-    <>
-      <a
-        className="w-fit rounded-md border border-primary/30 bg-primary/10 px-3 py-2 text-xs font-semibold text-primary hover:border-primary"
-        href={WECHAT_VERSION_DOWNLOAD_URL}
-        target="_blank"
-        rel="noreferrer"
-      >
-        下载微信历史版本 ↗
-      </a>
-      <div className="grid grid-cols-2 gap-2 sm:grid-cols-3">
-        {BUNDLED_WECHAT_VERSIONS.map((version) => (
-          <span
-            className="rounded-md border border-border bg-background px-2 py-2 text-center text-xs text-muted-foreground"
-            key={version}
-          >
-            {version}
-          </span>
-        ))}
-      </div>
-    </>
-  )
-}
@@ -6,6 +6,8 @@ import type {
 import { Button, Dialog, DialogContent, DialogDescription, DialogHeader, DialogTitle } from '../ui'
 import { PersonalWechatChatComposer, type ChatMessage } from './PersonalWechatChatComposer'
 import type { PersonalWechatSendDialogProps } from './PersonalWechatSendDialog'
+import { ReportImagePostfixInput } from './ReportImagePostfixInput'
+import { useReportImagePostfixSetting } from './useReportImagePostfixSetting'

 function fallbackStatus(error: unknown): PersonalWechatSenderStatus {
  return {
@@ -45,7 +47,7 @@ function statusLabel(status: PersonalWechatSenderStatus | null): string {
 }

 function statusText(value: string): string {
-  return value.replace(/OneBot|Hook/gi, '微信发送能力').replace(/个人微信发送组件/g, '微信发送能力')
+  return value.replace(/Hook/gi, '微信发送能力').replace(/个人微信发送组件/g, '微信发送能力')
 }

 function statusDescription(status: PersonalWechatSenderStatus | null): string {
@@ -64,12 +66,17 @@ export function PersonalWechatWindowsSendDialog({
  const [status, setStatus] = useState<PersonalWechatSenderStatus | null>(null)
  const [detecting, setDetecting] = useState(true)
  const [sendBusy, setSendBusy] = useState(false)
+  const [sendError, setSendError] = useState<string | null>(null)
+  const [sendSuccess, setSendSuccess] = useState<string | null>(null)
  const [messages, setMessages] = useState<ChatMessage[]>([])
  const requestIdRef = useRef(0)
  const restoreFocusRef = useRef<HTMLElement | null>(null)
  const closingRef = useRef(false)
  const displayName = contact.m_nsNickName || contact.m_nsUsrName || '未命名会话'
  const targetId = contact.m_nsUsrName
+  const { postfixText, setPostfixText, persistPostfixText } = useReportImagePostfixSetting(
+    Boolean(initialImage)
+  )

  const refreshStatus = useCallback(async (): Promise<void> => {
    const requestId = ++requestIdRef.current
@@ -103,9 +110,7 @@ export function PersonalWechatWindowsSendDialog({
    onOpenPersonalWechatSettings()
  }

-  const handleSend = async (
-    filePath: string
-  ): Promise<{ success: boolean; error?: string }> => {
+  const handleSend = async (filePath: string): Promise<{ success: boolean; error?: string }> => {
    setSendBusy(true)
    try {
      const response = await window.api.sendGeneratedTtsVoice({
@@ -137,15 +142,28 @@ export function PersonalWechatWindowsSendDialog({
  const handleSendReportImage = async (): Promise<void> => {
    if (!initialImage || !status?.canSendImage || sendBusy) return
    setSendBusy(true)
+    setSendError(null)
+    setSendSuccess(null)
    try {
+      await persistPostfixText()
      const response = await window.api.sendPersonalWechatMessage({
        type: 'image',
        to: targetId,
        isGroup: isGroupChat,
-        filePath: initialImage.path
+        filePath: initialImage.path,
+        postfixText
      } satisfies PersonalWechatSendRequest)
      setStatus(response.status)
-      if (!response.success) return
+      if (!response.success) {
+        setSendError(response.error || '日报图片发送失败')
+        return
+      }
+      if (response.postfixError) {
+        setSendError(`日报图片已发送，但${response.postfixError}`)
+        setSendSuccess('日报图片发送成功')
+      } else {
+        setSendSuccess(postfixText.trim() ? '日报图片和后置词发送成功' : '日报图片发送成功')
+      }
      setMessages((current) => [
        ...current,
        {
@@ -156,6 +174,8 @@ export function PersonalWechatWindowsSendDialog({
          outgoing: true
        }
      ])
+    } catch (error) {
+      setSendError(error instanceof Error ? error.message : String(error))
    } finally {
      setSendBusy(false)
    }
@@ -183,7 +203,7 @@ export function PersonalWechatWindowsSendDialog({
        <DialogHeader className="flex-row items-center justify-between space-y-0 pr-10">
          <div>
            <span className="text-[11px] font-bold tracking-normal text-primary">实验性功能</span>
-              <DialogTitle className="mt-0.5 text-[19px] leading-[26px] tracking-normal">
+            <DialogTitle className="mt-0.5 text-[19px] leading-[26px] tracking-normal">
              文字转语音
            </DialogTitle>
          </div>
@@ -201,7 +221,7 @@ export function PersonalWechatWindowsSendDialog({
        </div>

        <div className="personal-wechat-send-target">
-            <span>{isGroupChat ? '发送到群聊' : '发送给联系人'}</span>
+          <span>{isGroupChat ? '发送到群聊' : '发送给联系人'}</span>
          <strong>{displayName}</strong>
          <code>{targetId}</code>
        </div>
@@ -246,7 +266,7 @@ export function PersonalWechatWindowsSendDialog({
                    key={message.id}
                    className={`personal-wechat-message-bubble ${message.outgoing ? 'is-outgoing' : ''}`}
                  >
-                      <span className="personal-wechat-message-kind">语音</span>
+                    <span className="personal-wechat-message-kind">语音</span>
                    <span>{message.text || message.fileName}</span>
                  </div>
                ))}
@@ -255,6 +275,16 @@ export function PersonalWechatWindowsSendDialog({
            {initialImage && status.canSendImage ? (
              <section className="personal-wechat-composer" aria-label="日报图片发送">
                <p>已准备日报图片：{initialImage.name}</p>
+                <ReportImagePostfixInput
+                  value={postfixText}
+                  onChange={setPostfixText}
+                  onBlur={() =>
+                    void persistPostfixText().catch((error) =>
+                      setSendError(error instanceof Error ? error.message : '发送后置词保存失败')
+                    )
+                  }
+                  disabled={sendBusy}
+                />
                <Button size="sm" onClick={() => void handleSendReportImage()} disabled={sendBusy}>
                  {sendBusy ? '发送中…' : '发送日报图片'}
                </Button>
@@ -271,6 +301,16 @@ export function PersonalWechatWindowsSendDialog({
                busy={sendBusy}
              />
            ) : null}
+            {sendError ? (
+              <div className="personal-wechat-global-error" role="alert">
+                {sendError}
+              </div>
+            ) : null}
+            {sendSuccess ? (
+              <div className="personal-wechat-global-success" role="status">
+                {sendSuccess}
+              </div>
+            ) : null}
          </>
        ) : null}
      </DialogContent>
@@ -0,0 +1,32 @@
+import { DEFAULT_REPORT_IMAGE_POSTFIX_TEXT } from '../../../../shared/personal-wechat'
+
+export function ReportImagePostfixInput({
+  value,
+  onChange,
+  onBlur,
+  disabled
+}: {
+  value: string
+  onChange: (value: string) => void
+  onBlur: () => void
+  disabled: boolean
+}): React.ReactElement {
+  return (
+    <label className="flex flex-col gap-1.5 text-sm">
+      <span className="font-medium">发送后置词</span>
+      <input
+        aria-label="发送后置词"
+        className="h-9 rounded-md border border-border bg-background px-3 text-sm outline-none focus:border-primary disabled:opacity-60"
+        value={value}
+        maxLength={200}
+        disabled={disabled}
+        onChange={(event) => onChange(event.target.value)}
+        onBlur={onBlur}
+        placeholder={DEFAULT_REPORT_IMAGE_POSTFIX_TEXT}
+      />
+      <span className="text-xs text-muted-foreground">
+        图片确认发送成功后，会发送一条文本消息；留空则只发图片。
+      </span>
+    </label>
+  )
+}
@@ -34,12 +34,12 @@ export function ExportIcon({ className }: IconProps): React.ReactElement {
  )
 }

-export function MoreIcon({ className }: IconProps): React.ReactElement {
+export function StatsIcon({ className }: IconProps): React.ReactElement {
  return (
    <svg className={className} viewBox="0 0 24 24" aria-hidden="true" focusable="false">
-      <circle cx="5" cy="12" r="1.4" />
-      <circle cx="12" cy="12" r="1.4" />
-      <circle cx="19" cy="12" r="1.4" />
+      <path d="M5.5 19v-6" />
+      <path d="M12 19V5" />
+      <path d="M18.5 19v-9" />
    </svg>
  )
 }
@@ -0,0 +1,36 @@
+import { useCallback, useEffect, useState } from 'react'
+import { DEFAULT_REPORT_IMAGE_POSTFIX_TEXT } from '../../../../shared/personal-wechat'
+
+export function useReportImagePostfixSetting(enabled: boolean): {
+  postfixText: string
+  setPostfixText: (value: string) => void
+  persistPostfixText: () => Promise<void>
+} {
+  const [postfixText, setPostfixText] = useState(DEFAULT_REPORT_IMAGE_POSTFIX_TEXT)
+
+  useEffect(() => {
+    if (!enabled) return
+    const getSettings = window.api.getSettings
+    if (typeof getSettings !== 'function') return
+    let active = true
+    void getSettings()
+      .then((result) => {
+        if (!active) return
+        setPostfixText(
+          String(result.settings.reportImagePostfixText ?? DEFAULT_REPORT_IMAGE_POSTFIX_TEXT)
+        )
+      })
+      .catch(() => undefined)
+    return () => {
+      active = false
+    }
+  }, [enabled])
+
+  const persistPostfixText = useCallback(async (): Promise<void> => {
+    const setSettings = window.api.setSettings
+    if (typeof setSettings !== 'function') return
+    await setSettings({ reportImagePostfixText: postfixText })
+  }, [postfixText])
+
+  return { postfixText, setPostfixText, persistPostfixText }
+}
@@ -27,6 +27,7 @@ import {
  buildContactSearchIndex,
  filterContactSearchIndex
 } from '../../../../shared/contact-search'
+import { loadExportPreferences, saveExportPreferences } from './exportPreferences'

 const ALL_CONTACT_TYPES: ExportContactType[] = ['group', 'user']
 const contactTypeKey = (types: ExportContactType[] | undefined): string =>
@@ -44,6 +45,8 @@ export function ExportWorkspace({
  onCancelExport
 }: ExportWorkspaceProps): React.ReactElement {
  const initialSelection = initialContact || contacts[0] || null
+  const defaultNameMode = initialSelection?.type === 'group' ? 'groupNickname' : 'remark'
+  const [savedPreferences] = useState(() => loadExportPreferences(defaultNameMode))
  const runningAllTask = exportTasks.find(
    (task) => task.scope === 'all' && task.status === 'running'
  )
@@ -64,24 +67,30 @@ export function ExportWorkspace({
  const contactSearchIndex = useMemo(() => buildContactSearchIndex(contacts), [contacts])
  const [activeContactId, setActiveContactId] = useState(initialSelection?.md5 || '')
  const [previewByContact, setPreviewByContact] = useState<Record<string, Message[]>>({})
-  const [range, setRange] = useState<ExportRange>(() => (runningAllTask ? 'all' : 'today'))
-  const [startDate, setStartDate] = useState('')
-  const [endDate, setEndDate] = useState('')
-  const [selectedKinds, setSelectedKinds] = useState<Set<string>>(() => new Set(['text']))
-  const [nameMode, setNameMode] = useState<ExportNameMode>(
-    initialSelection?.type === 'group' ? 'groupNickname' : 'remark'
+  const [range, setRange] = useState<ExportRange>(() => savedPreferences.range)
+  const [startDate, setStartDate] = useState(() => savedPreferences.startDate)
+  const [endDate, setEndDate] = useState(() => savedPreferences.endDate)
+  const [selectedKinds, setSelectedKinds] = useState<Set<string>>(
+    () => new Set(savedPreferences.selectedKinds)
+  )
+  const [nameMode, setNameMode] = useState<ExportNameMode>(() => savedPreferences.nameMode)
+  const [includeMedia, setIncludeMedia] = useState(() => savedPreferences.includeMedia)
+  const [includeVoiceTranscripts, setIncludeVoiceTranscripts] = useState(
+    () => savedPreferences.includeVoiceTranscripts
  )
-  const [includeMedia, setIncludeMedia] = useState(true)
-  const [includeVoiceTranscripts, setIncludeVoiceTranscripts] = useState(true)
  const [voiceModelStatus, setVoiceModelStatus] = useState<VoiceModelStatus | null>(null)
-  const [includeAvatars, setIncludeAvatars] = useState(true)
-  const [preferOriginal, setPreferOriginal] = useState(true)
-  const [fallbackThumbnail, setFallbackThumbnail] = useState(true)
-  const [keepMissing, setKeepMissing] = useState(true)
-  const [format, setFormat] = useState<ExportFormat>(() => runningAllTask?.format || 'csv')
-  const [zip, setZip] = useState(() => runningAllTask?.zip === true)
-  const [fileName, setFileName] = useState('')
-  const [outputDirectory, setOutputDirectory] = useState('')
+  const [includeAvatars, setIncludeAvatars] = useState(() => savedPreferences.includeAvatars)
+  const [preferOriginal, setPreferOriginal] = useState(() => savedPreferences.preferOriginal)
+  const [fallbackThumbnail, setFallbackThumbnail] = useState(
+    () => savedPreferences.fallbackThumbnail
+  )
+  const [keepMissing, setKeepMissing] = useState(() => savedPreferences.keepMissing)
+  const [format, setFormat] = useState<ExportFormat>(
+    () => runningAllTask?.format || savedPreferences.format
+  )
+  const [zip, setZip] = useState(() => runningAllTask?.zip ?? savedPreferences.zip)
+  const [fileName, setFileName] = useState(() => savedPreferences.fileName)
+  const [outputDirectory, setOutputDirectory] = useState(() => savedPreferences.outputDirectory)
  const [status, setStatus] = useState<ExportStatus>('idle')
  const [jobId, setJobId] = useState('')
  const [progress, setProgress] = useState<ExportJobProgress | null>(null)
@@ -92,6 +101,42 @@ export function ExportWorkspace({
  const [taskCenterOpen, setTaskCenterOpen] = useState(false)
  const selectionLimit = 5

+  React.useEffect(() => {
+    saveExportPreferences({
+      format,
+      range,
+      startDate,
+      endDate,
+      selectedKinds: Array.from(selectedKinds) as ExportRequest['kinds'],
+      nameMode,
+      includeMedia,
+      includeVoiceTranscripts,
+      includeAvatars,
+      preferOriginal,
+      fallbackThumbnail,
+      keepMissing,
+      zip,
+      fileName,
+      outputDirectory
+    })
+  }, [
+    endDate,
+    fallbackThumbnail,
+    fileName,
+    format,
+    includeAvatars,
+    includeMedia,
+    includeVoiceTranscripts,
+    keepMissing,
+    nameMode,
+    outputDirectory,
+    preferOriginal,
+    range,
+    selectedKinds,
+    startDate,
+    zip
+  ])
+
  React.useEffect(() => {
    if (selectedContacts.length > 0) return
    const candidate = initialContact || contacts[0]
@@ -206,7 +251,6 @@ export function ExportWorkspace({
  const handleSelectContact = (contact: Contact): void => {
    if (exportAll) {
      setExportAll(false)
-      setRange('today')
    }
    if (!selectionMode) {
      setSelectedContacts([contact])
@@ -236,7 +280,6 @@ export function ExportWorkspace({
    setExportAll(true)
    setAllContactTypes([...ALL_CONTACT_TYPES])
    setSelectionMode(false)
-    setRange('all')
    setStatus('idle')
  }

@@ -417,6 +460,7 @@ export function ExportWorkspace({
    setSelectedKinds(new Set(['text']))
    setNameMode(contact?.type === 'group' ? 'groupNickname' : 'remark')
    setIncludeMedia(true)
+    setIncludeVoiceTranscripts(true)
    setIncludeAvatars(true)
    setPreferOriginal(true)
    setFallbackThumbnail(true)
@@ -424,6 +468,7 @@ export function ExportWorkspace({
    setFormat('csv')
    setZip(false)
    setFileName('')
+    setOutputDirectory('')
    setStatus('idle')
    setJobId('')
    setProgress(null)
@@ -492,7 +537,7 @@ export function ExportWorkspace({
        selectionMode={selectionMode}
        exportContactCount={exportContacts.length}
        format={format}
-        range={range}
+        range={exportAll ? 'all' : range}
        startDate={startDate}
        endDate={endDate}
        selectedKinds={selectedKinds}
@@ -0,0 +1,115 @@
+import type { ExportFormat, ExportMessageKind, ExportNameMode } from '../../../../shared/export'
+import type { ExportRange } from './exportTypes'
+
+export interface ExportPreferences {
+  format: ExportFormat
+  range: ExportRange
+  startDate: string
+  endDate: string
+  selectedKinds: ExportMessageKind[]
+  nameMode: ExportNameMode
+  includeMedia: boolean
+  includeVoiceTranscripts: boolean
+  includeAvatars: boolean
+  preferOriginal: boolean
+  fallbackThumbnail: boolean
+  keepMissing: boolean
+  zip: boolean
+  fileName: string
+  outputDirectory: string
+}
+
+export const EXPORT_PREFERENCES_STORAGE_KEY = 'tracememo_export_preferences'
+
+const formats: ExportFormat[] = ['html', 'csv', 'json', 'markdown']
+const ranges: ExportRange[] = ['all', 'today', 'threeDays', 'sevenDays', 'custom']
+const messageKinds: ExportMessageKind[] = [
+  'text',
+  'image',
+  'video',
+  'voice',
+  'sticker',
+  'file',
+  'share',
+  'location',
+  'system'
+]
+const nameModes: ExportNameMode[] = ['groupNickname', 'remark', 'wechatNickname']
+
+export function loadExportPreferences(defaultNameMode: ExportNameMode): ExportPreferences {
+  const defaults: ExportPreferences = {
+    format: 'csv',
+    range: 'today',
+    startDate: '',
+    endDate: '',
+    selectedKinds: ['text'],
+    nameMode: defaultNameMode,
+    includeMedia: true,
+    includeVoiceTranscripts: true,
+    includeAvatars: true,
+    preferOriginal: true,
+    fallbackThumbnail: true,
+    keepMissing: true,
+    zip: false,
+    fileName: '',
+    outputDirectory: ''
+  }
+
+  try {
+    const stored = window.localStorage.getItem(EXPORT_PREFERENCES_STORAGE_KEY)
+    if (!stored) return defaults
+    const value = JSON.parse(stored) as Partial<ExportPreferences>
+    return {
+      format: formats.includes(value.format as ExportFormat)
+        ? (value.format as ExportFormat)
+        : defaults.format,
+      range: ranges.includes(value.range as ExportRange)
+        ? (value.range as ExportRange)
+        : defaults.range,
+      startDate: typeof value.startDate === 'string' ? value.startDate : defaults.startDate,
+      endDate: typeof value.endDate === 'string' ? value.endDate : defaults.endDate,
+      selectedKinds: Array.isArray(value.selectedKinds)
+        ? [
+            ...new Set(
+              value.selectedKinds.filter((kind): kind is ExportMessageKind =>
+                messageKinds.includes(kind as ExportMessageKind)
+              )
+            )
+          ]
+        : defaults.selectedKinds,
+      nameMode: nameModes.includes(value.nameMode as ExportNameMode)
+        ? (value.nameMode as ExportNameMode)
+        : defaults.nameMode,
+      includeMedia:
+        typeof value.includeMedia === 'boolean' ? value.includeMedia : defaults.includeMedia,
+      includeVoiceTranscripts:
+        typeof value.includeVoiceTranscripts === 'boolean'
+          ? value.includeVoiceTranscripts
+          : defaults.includeVoiceTranscripts,
+      includeAvatars:
+        typeof value.includeAvatars === 'boolean' ? value.includeAvatars : defaults.includeAvatars,
+      preferOriginal:
+        typeof value.preferOriginal === 'boolean' ? value.preferOriginal : defaults.preferOriginal,
+      fallbackThumbnail:
+        typeof value.fallbackThumbnail === 'boolean'
+          ? value.fallbackThumbnail
+          : defaults.fallbackThumbnail,
+      keepMissing:
+        typeof value.keepMissing === 'boolean' ? value.keepMissing : defaults.keepMissing,
+      zip: typeof value.zip === 'boolean' ? value.zip : defaults.zip,
+      fileName: typeof value.fileName === 'string' ? value.fileName : defaults.fileName,
+      outputDirectory:
+        typeof value.outputDirectory === 'string' ? value.outputDirectory : defaults.outputDirectory
+    }
+  } catch {
+    return defaults
+  }
+}
+
+export function saveExportPreferences(preferences: ExportPreferences): void {
+  try {
+    window.localStorage.setItem(EXPORT_PREFERENCES_STORAGE_KEY, JSON.stringify(preferences))
+  } catch {
+    // Export remains usable when browser storage is unavailable.
+  }
+}
@@ -0,0 +1,365 @@
+import React, { useCallback, useEffect, useMemo, useState } from 'react'
+import type { Contact } from '../../../../shared/types'
+import {
+  GROUP_STATS_RANGE_OPTIONS,
+  formatActiveMemberStats,
+  formatMemberLineName,
+  formatSilentMemberStats,
+  formatStatsDateTime,
+  resolveGroupStatsRangeStart,
+  type GroupMemberStatsResult,
+  type GroupStatsRangeKey
+} from '../../../../shared/group-stats'
+import { Button, Dialog, DialogContent, DialogHeader, DialogTitle, Tabs, TabsContent, TabsList, TabsTrigger } from '../ui'
+
+interface GroupMemberStatsDialogProps {
+  open: boolean
+  onOpenChange: (open: boolean) => void
+  contact: Contact
+  /** 跳到「设置 · 本地索引」，让用户能把没追平的索引处理掉。 */
+  onOpenLocalIndexSettings?: () => void
+}
+
+type CopyTarget = 'none' | 'active' | 'silent'
+
+/**
+ * 群发言统计面板（单群）。
+ *
+ * - **结果**由 main 的 `GroupStatsService` 给出，组件只展示，不自己拼业务文本；
+ * - **可复制的文本**由 `shared/group-stats` 的 formatter 生成，保证「界面上看到的」
+ *   与「复制出去的」不会各说各话；
+ * - **名单不截断**：数字与列表必须能对上，否则用户会以为统计漏了人。
+ *   列表靠 CSS 限高滚动，不靠丢数据。
+ */
+export function GroupMemberStatsDialog({
+  open,
+  onOpenChange,
+  contact,
+  onOpenLocalIndexSettings
+}: GroupMemberStatsDialogProps): React.ReactElement {
+  const [rangeKey, setRangeKey] = useState<GroupStatsRangeKey>('30d')
+  const [customStart, setCustomStart] = useState('')
+  const [customEnd, setCustomEnd] = useState('')
+  const [result, setResult] = useState<GroupMemberStatsResult | null>(null)
+  const [loading, setLoading] = useState(false)
+  const [error, setError] = useState<string | null>(null)
+  const [copied, setCopied] = useState<CopyTarget>('none')
+  const [notice, setNotice] = useState<string | null>(null)
+  /** 发送能力必须由 main 判定：renderer 侧的 platform 常量会误判 macOS Intel。 */
+  const [canSendText, setCanSendText] = useState(false)
+  const [sending, setSending] = useState(false)
+
+  const timeWindow = useMemo(() => {
+    if (rangeKey === 'custom') {
+      const start = customStart ? new Date(`${customStart}T00:00:00`).getTime() : NaN
+      const end = customEnd ? new Date(`${customEnd}T23:59:59.999`).getTime() : NaN
+      return { startTime: start, endTime: end }
+    }
+    const endTime = Date.now()
+    return { startTime: resolveGroupStatsRangeStart(rangeKey, endTime), endTime }
+  }, [rangeKey, customStart, customEnd])
+
+  const rangeReady = Number.isFinite(timeWindow.startTime) && Number.isFinite(timeWindow.endTime)
+
+  useEffect(() => {
+    if (!open || !rangeReady || !contact.md5) return
+    let disposed = false
+    setLoading(true)
+    setError(null)
+    window.api
+      .getGroupMemberStats({
+        userMd5: contact.md5,
+        startTime: timeWindow.startTime,
+        endTime: timeWindow.endTime
+      })
+      .then((data) => {
+        if (!disposed) setResult(data)
+      })
+      .catch((loadError: unknown) => {
+        if (!disposed) setError(loadError instanceof Error ? loadError.message : String(loadError))
+      })
+      .finally(() => {
+        if (!disposed) setLoading(false)
+      })
+    return () => {
+      disposed = true
+    }
+  }, [open, rangeReady, contact.md5, timeWindow.startTime, timeWindow.endTime])
+
+  useEffect(() => {
+    if (!open) return
+    let disposed = false
+    window.api
+      .getPersonalWechatSenderStatus()
+      .then((status) => {
+        if (!disposed) setCanSendText(status?.canSendText === true)
+      })
+      .catch(() => {
+        if (!disposed) setCanSendText(false)
+      })
+    return () => {
+      disposed = true
+    }
+  }, [open])
+
+  const silentText = useMemo(() => (result ? formatSilentMemberStats(result) : ''), [result])
+  const activeText = useMemo(() => (result ? formatActiveMemberStats(result) : ''), [result])
+
+  const copy = useCallback(async (text: string, target: CopyTarget) => {
+    if (!text) return
+    try {
+      await navigator.clipboard.writeText(text)
+      setCopied(target)
+      setNotice(null)
+      setTimeout(() => setCopied('none'), 2000)
+    } catch {
+      setError('复制失败，请手动选择文本')
+    }
+  }, [])
+
+  const send = useCallback(
+    async (text: string) => {
+      if (!text || sending) return
+      setSending(true)
+      setNotice(null)
+      try {
+        await window.api.sendPersonalWechatMessage({
+          to: contact.m_nsUsrName,
+          isGroup: true,
+          type: 'text',
+          text
+        })
+        setNotice('已发送到当前群')
+      } catch (sendError: unknown) {
+        setError(sendError instanceof Error ? sendError.message : '发送失败')
+      } finally {
+        setSending(false)
+      }
+    },
+    [contact.m_nsUsrName, sending]
+  )
+
+  const stale = result !== null && result.freshness !== 'fresh'
+  const groupTitle = contact.m_nsNickName || contact.m_nsUsrName
+  /** 备注与群名相同时不重复展示，避免标题里出现两遍同一个人。 */
+  const groupRemark = contact.remark && contact.remark !== groupTitle ? contact.remark : ''
+  /** 选「全部」时 startTime 为 0，改用本机第一条消息的真实时间作为起点。 */
+  const rangeStart = result
+    ? result.startTime > 0
+      ? result.startTime
+      : result.firstMessageTime
+    : null
+
+  return (
+    <Dialog open={open} onOpenChange={onOpenChange}>
+      <DialogContent className="max-w-3xl">
+        <DialogHeader>
+          <DialogTitle>群发言统计 · {groupTitle}</DialogTitle>
+          {groupRemark ? (
+            <p className="group-stats-subtitle">群备注：{groupRemark}</p>
+          ) : null}
+        </DialogHeader>
+
+        <div className="group-stats-panel">
+          <div className="group-stats-ranges">
+            {GROUP_STATS_RANGE_OPTIONS.map((option) => (
+              <Button
+                key={option.key}
+                variant={rangeKey === option.key ? 'default' : 'outline'}
+                size="sm"
+                onClick={() => setRangeKey(option.key)}
+              >
+                {option.label}
+              </Button>
+            ))}
+            {rangeKey === 'custom' ? (
+              <div className="group-stats-custom-range">
+                <input
+                  type="date"
+                  aria-label="开始日期"
+                  value={customStart}
+                  onChange={(event) => setCustomStart(event.target.value)}
+                />
+                <span>至</span>
+                <input
+                  type="date"
+                  aria-label="结束日期"
+                  value={customEnd}
+                  onChange={(event) => setCustomEnd(event.target.value)}
+                />
+              </div>
+            ) : null}
+          </div>
+
+          {!rangeReady ? (
+            <p className="group-stats-hint">请选择完整的开始与结束日期。</p>
+          ) : null}
+          {stale ? (
+            <div className="group-stats-stale">
+              <p role="alert" className="group-stats-warning">
+                本地索引尚未完全同步，以下结果可能不完整。
+              </p>
+              {onOpenLocalIndexSettings ? (
+                <Button
+                  variant="outline"
+                  size="sm"
+                  onClick={() => {
+                    // 先关面板再跳转：否则设置页会被这个 Dialog 挡在后面。
+                    onOpenChange(false)
+                    onOpenLocalIndexSettings()
+                  }}
+                >
+                  去建立索引
+                </Button>
+              ) : null}
+            </div>
+          ) : null}
+          {error ? (
+            <p role="alert" className="group-stats-error">
+              {error}
+            </p>
+          ) : null}
+          {notice ? <p className="group-stats-notice">{notice}</p> : null}
+
+          {loading ? <p className="group-stats-hint">统计中…</p> : null}
+
+          {result && !loading ? (
+            <>
+              <div className="group-stats-overview">
+                <div className="group-stats-metric">
+                  <div className="group-stats-metric-label">当前群成员</div>
+                  <div className="group-stats-metric-value">{result.memberCount}</div>
+                </div>
+                <div className="group-stats-metric">
+                  <div className="group-stats-metric-label">发过言</div>
+                  <div className="group-stats-metric-value">{result.activeMemberCount}</div>
+                </div>
+                <div className="group-stats-metric">
+                  <div className="group-stats-metric-label">没发言</div>
+                  <div className="group-stats-metric-value">{result.silentMemberCount}</div>
+                </div>
+              </div>
+
+              <p className="group-stats-range-text">
+                统计区间：
+                {rangeStart
+                  ? `${formatStatsDateTime(rangeStart)} ～ ${formatStatsDateTime(result.endTime)}`
+                  : `全部历史 ～ ${formatStatsDateTime(result.endTime)}`}
+                {result.startTime <= 0 && rangeStart ? '（本机该群第一条消息起）' : null}
+              </p>
+
+              {/* 两个名单用 tab 分开：一次只渲染一个列表，滚动压力减半，
+                  而且「复制/发送」跟它作用的内容处在同一个 tab 里，不会看错对象。 */}
+              <Tabs defaultValue="active" className="group-stats-tabs">
+                <TabsList className="group-stats-tabs-list">
+                  <TabsTrigger value="active">
+                    发言排行（{result.activeMemberCount}）
+                  </TabsTrigger>
+                  <TabsTrigger value="silent">
+                    未发言统计（{result.silentMemberCount}）
+                  </TabsTrigger>
+                </TabsList>
+
+                <TabsContent value="active">
+                  <div className="group-stats-section">
+                    <div className="group-stats-section-head">
+                      <h3>
+                        发言成员排行
+                        <span className="group-stats-section-count">
+                          （共 {result.activeMemberCount} 人）
+                        </span>
+                      </h3>
+                      <div className="group-stats-section-actions">
+                        <Button
+                          variant="outline"
+                          size="sm"
+                          onClick={() => void copy(activeText, 'active')}
+                        >
+                          {copied === 'active' ? '已复制' : '复制'}
+                        </Button>
+                      </div>
+                    </div>
+                    {result.activeMembers.length === 0 ? (
+                      <p className="group-stats-empty">该区间内没有人发言。</p>
+                    ) : (
+                      <ol className="group-stats-list">
+                        {result.activeMembers.map((member, index) => (
+                          <li key={member.senderId} className="group-stats-row">
+                            <span className="group-stats-row-name">
+                              <span className="group-stats-row-index">{index + 1}.</span>
+                              {formatMemberLineName(member)}
+                            </span>
+                            <span className="group-stats-row-meta">
+                              {member.messageCount} 条 ·{' '}
+                              {formatStatsDateTime(member.lastMessageTime)}
+                            </span>
+                          </li>
+                        ))}
+                      </ol>
+                    )}
+                  </div>
+                </TabsContent>
+
+                <TabsContent value="silent">
+                  <div className="group-stats-section">
+                    <div className="group-stats-section-head">
+                      <h3>
+                        未发言成员
+                        <span className="group-stats-section-count">
+                          （共 {result.silentMemberCount} 人）
+                        </span>
+                      </h3>
+                      <div className="group-stats-section-actions">
+                        <Button
+                          variant="outline"
+                          size="sm"
+                          onClick={() => void copy(silentText, 'silent')}
+                        >
+                          {copied === 'silent' ? '已复制' : '复制'}
+                        </Button>
+                        {canSendText ? (
+                          <Button
+                            variant="default"
+                            size="sm"
+                            disabled={sending}
+                            onClick={() => void send(silentText)}
+                          >
+                            {sending ? '发送中…' : '发送到当前群'}
+                          </Button>
+                        ) : null}
+                      </div>
+                    </div>
+                    {result.silentMembers.length === 0 ? (
+                      <p className="group-stats-empty">全部成员在该区间内都发过言。</p>
+                    ) : (
+                      <ol className="group-stats-list">
+                        {result.silentMembers.map((member, index) => (
+                          <li key={member.senderId} className="group-stats-row">
+                            <span className="group-stats-row-name">
+                              <span className="group-stats-row-index">{index + 1}.</span>
+                              {formatMemberLineName(member)}
+                            </span>
+                          </li>
+                        ))}
+                      </ol>
+                    )}
+                  </div>
+                </TabsContent>
+              </Tabs>
+
+              <p className="group-stats-notes">
+                {result.unattributedMessages > 0
+                  ? `另有 ${result.unattributedMessages} 条消息无法归属到具体成员（未计入任何人）。`
+                  : null}
+                {result.excludedSystemMessages > 0
+                  ? `已排除 ${result.excludedSystemMessages} 条系统消息。`
+                  : null}
+                {result.limitations.join(' ')}
+              </p>
+            </>
+          ) : null}
+        </div>
+      </DialogContent>
+    </Dialog>
+  )
+}
@@ -54,6 +54,13 @@ function NavIcon({ page }: NavIconProps): React.ReactElement {
          <path d="M5.5 15.5v3h13v-3" />
        </svg>
      )
+    case 'automation':
+      // 闪电：强调「命中即自动执行」，与退群监控的箭头区分开。
+      return (
+        <svg viewBox="0 0 24 24" aria-hidden="true" focusable="false">
+          <path d="M13 4.5 7.5 13h4l-.5 6.5L16.5 11h-4z" />
+        </svg>
+      )
    case 'agent-hub':
      return (
        <svg viewBox="0 0 24 24" aria-hidden="true" focusable="false">
@@ -3,6 +3,7 @@ export type AppPage =
  | 'search'
  | 'report'
  | 'exit-monitor'
+  | 'automation'
  | 'agent-hub'
  | 'export'
  | 'api'
@@ -18,6 +19,7 @@ export const PRIMARY_NAV_ITEMS: NavigationItem[] = [
  { id: 'search', label: '问问微信' },
  { id: 'report', label: '日报' },
  { id: 'exit-monitor', label: '退群监控' },
+  { id: 'automation', label: '自动化' },
  { id: 'agent-hub', label: 'Agent' },
  { id: 'export', label: '导出' },
  { id: 'api', label: 'API' },
@@ -4,7 +4,13 @@ import { Checkbox } from '../ui'

 interface MessageTypeSelectorProps {
  value: SummaryMessageType[]
-  counts: Record<SummaryMessageType, number>
+  /**
+   * 每类消息的条数。
+   *
+   * 可选：`自动化 → 定时日报` 的编辑器没有真实消息统计，
+   * 不传就不显示数字，而不是显示一排假的 0。
+   */
+  counts?: Record<SummaryMessageType, number>
  disabled: boolean
  onChange: (value: SummaryMessageType[]) => void
 }
@@ -49,7 +55,7 @@ export function MessageTypeSelector({
              <b>{option.label}</b>
              <small>{option.description}</small>
            </span>
-            <em>{counts[option.value]}</em>
+            {counts ? <em>{counts[option.value]}</em> : null}
          </label>
        ))}
      </div>
@@ -0,0 +1,342 @@
+import React, { useEffect, useMemo, useState } from 'react'
+import {
+  Button,
+  DropdownMenu,
+  DropdownMenuContent,
+  DropdownMenuItem,
+  DropdownMenuTrigger,
+  Progress,
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue
+} from '../ui'
+import {
+  describeImageTextCoverage,
+  imageTextCoverageState,
+  imageTextPhaseLabel,
+  imageTextProcessedPercent,
+  imageTextStateLabel,
+  imageTextStateTone
+} from '../../../../shared/image-text-index'
+import { useImageTextIndexStatus } from './hooks/useImageTextIndexStatus'
+import {
+  IndexStatusBadge,
+  type IndexStatusTone
+} from '../../features/settings/components/IndexStatusBadge'
+
+interface ImageTextIndexSummaryProps {
+  dbReady: boolean
+  onNotice: (message: string) => void
+}
+
+/**
+ * 「图片文字索引」卡片（**设置页宽版**）。
+ *
+ * 为什么不是直接复用 `ImageTextIndexCard`：那个组件是给**问问微信侧栏**写的，
+ * 而侧栏最窄时正文只有约 145px（见 `search.scss` 的注释）。
+ * 四列 metric 布局塞进 145px 会直接崩掉 —— 反过来让侧栏继续用表格布局又浪费设置页的宽度。
+ * 所以两者**共享全部数据来源与文案函数**（`useImageTextIndexStatus` +
+ * `shared/image-text-index`），只有**版式**不同。
+ *
+ * 状态徽章、统计数字、进度条都走公共组件/函数，不存在第二套计算口径。
+ */
+export function ImageTextIndexSummary({
+  dbReady,
+  onNotice
+}: ImageTextIndexSummaryProps): React.ReactElement {
+  const {
+    status,
+    count,
+    counting,
+    pending,
+    running,
+    paused,
+    established,
+    refreshCount,
+    start,
+    pause,
+    resume,
+    cancel,
+    resetFailures,
+    repair
+  } = useImageTextIndexStatus({ dbReady, onNotice })
+  const [confirming, setConfirming] = useState(false)
+  const [confirmCount, setConfirmCount] = useState<number | null>(null)
+  /**
+   * 处理范围（天）。`0` = 全部历史。
+   *
+   * 保留这个选择器是因为它来自真实业务能力，不是装饰：全量回填几万张图没法用来排查问题，
+   * 先跑「最近 1 天」才能证明链路真的通了。
+   */
+  const [rangeDays, setRangeDays] = useState('0')
+  const sinceMs = useMemo(() => {
+    const days = Number(rangeDays)
+    return Number.isFinite(days) && days > 0 ? Date.now() - days * 24 * 60 * 60 * 1000 : undefined
+  }, [rangeDays])
+
+  useEffect(() => {
+    if (!dbReady) return
+    void refreshCount(sinceMs)
+  }, [dbReady, sinceMs, refreshCount])
+
+  const coverage = status?.coverage ?? null
+  const progress = status?.progress ?? null
+  const coverageState = coverage ? imageTextCoverageState(coverage) : 'not_built'
+  const percent = coverage
+    ? imageTextProcessedPercent(coverage.processed, coverage.totalImageMessages)
+    : 0
+  const interrupted = paused || progress?.state === 'cancelled'
+  const phaseLabel = progress?.currentPhase ? imageTextPhaseLabel(progress.currentPhase) : null
+
+  const labelInput = {
+    progressState: progress?.state,
+    running,
+    paused,
+    established,
+    coverageState,
+    percent
+  }
+  const stateLabel = imageTextStateLabel(labelInput)
+  const tone = imageTextStateTone(labelInput) as IndexStatusTone
+
+  const detectedImages = count?.totalImageMessages ?? coverage?.totalImageMessages ?? null
+  const nothingCounted =
+    count !== null && count.scannedConversations === 0 && count.failedConversations > 0
+  const failureCount = coverage?.failed ?? 0
+
+  const requestStart = async (): Promise<void> => {
+    const fresh = await refreshCount(sinceMs)
+    setConfirmCount(fresh?.totalImageMessages ?? detectedImages)
+    setConfirming(true)
+  }
+
+  return (
+    <section className="local-index-card" aria-label="图片文字索引状态">
+      <header className="local-index-card-head">
+        <div className="local-index-card-title">
+          <h3>图片文字索引</h3>
+          <p>本地 OCR 提取图片文字，让截图、报价图也能被搜到</p>
+        </div>
+        <IndexStatusBadge label={stateLabel} tone={tone} />
+      </header>
+
+      {/* 未建立：先给一个可核对的图片量，再让用户决定跑不跑 */}
+      {!established && !running ? (
+        <>
+          <div className="local-index-stats">
+            <div className="local-index-stat">
+              <strong className="local-index-stat-value">
+                {counting
+                  ? '统计中…'
+                  : nothingCounted
+                    ? '无法统计'
+                    : detectedImages === null
+                      ? '—'
+                      : detectedImages.toLocaleString()}
+              </strong>
+              <span className="local-index-stat-label">检测到图片消息</span>
+            </div>
+          </div>
+          <div className="local-index-inline">
+            <span className="local-index-stat-label">处理范围</span>
+            <Select value={rangeDays} onValueChange={setRangeDays}>
+              <SelectTrigger className="h-7 w-[110px] text-xs">
+                <SelectValue />
+              </SelectTrigger>
+              <SelectContent>
+                <SelectItem value="0">全部历史</SelectItem>
+                <SelectItem value="1">近 1 天</SelectItem>
+                <SelectItem value="7">近 7 天</SelectItem>
+                <SelectItem value="30">近 30 天</SelectItem>
+              </SelectContent>
+            </Select>
+          </div>
+          {nothingCounted ? (
+            <p className="local-index-note is-error">
+              {`无法统计本账号的图片消息（${count?.error || '读取消息表失败'}）。这不代表没有图片，可以重新统计再试一次。`}
+            </p>
+          ) : null}
+        </>
+      ) : null}
+
+      {/* 运行中 / 暂停：原地切进度，按钮按真实状态显隐，不允许重复点 */}
+      {(running || paused) && progress ? (
+        <div className="local-index-progress">
+          <div className="local-index-progress-head">
+            <span>识别图片文字</span>
+            <span className="local-index-progress-count">
+              {progress.processed.toLocaleString()} / {progress.totalImageMessages.toLocaleString()}
+            </span>
+          </div>
+          {/* 复用项目 `Progress`（更新页 / 导出任务中心同一组件），不另起一套轨道样式。 */}
+          <Progress
+            value={Math.min(100, Math.max(0, progress.percent))}
+            aria-label="图片文字索引进度"
+          />
+          <p className="local-index-note">
+            {phaseLabel ?? '正在识别…'}
+            {` · 识别出文字 ${progress.indexed.toLocaleString()} · 无文字 ${progress.empty.toLocaleString()}`}
+          </p>
+        </div>
+      ) : null}
+
+      {/* 已建立：四个可核对的数字 */}
+      {established && !running && !paused && coverage ? (
+        <>
+          <div className="local-index-stats is-four">
+            <div className="local-index-stat">
+              <strong className="local-index-stat-value">
+                {coverage.indexed.toLocaleString()}
+              </strong>
+              <span className="local-index-stat-label">已识别文字</span>
+            </div>
+            <div className="local-index-stat">
+              <strong className="local-index-stat-value">{coverage.empty.toLocaleString()}</strong>
+              <span className="local-index-stat-label">无文字图片</span>
+            </div>
+            <div className="local-index-stat">
+              <strong className="local-index-stat-value">
+                {coverage.missing.toLocaleString()}
+              </strong>
+              <span className="local-index-stat-label">图片已清理</span>
+            </div>
+            <div className="local-index-stat">
+              <strong className="local-index-stat-value">{coverage.failed.toLocaleString()}</strong>
+              <span className="local-index-stat-label">识别失败</span>
+            </div>
+          </div>
+        </>
+      ) : null}
+
+      {progress?.state === 'error' ? (
+        <p className="local-index-note is-error">
+          {progress.lastError || '图片文字索引建立失败，可以稍后重试。'}
+        </p>
+      ) : null}
+      {coverage?.systemicFailure === true ? (
+        <p className="local-index-note is-error">
+          {`${(coverage.failed ?? 0).toLocaleString()} 条处理失败，成功识别 0 条 —— 当前无法搜索图片中的文字。`}
+        </p>
+      ) : null}
+      {paused ? (
+        <p className="local-index-note">已暂停。已识别的结果都保留了，点「继续」会从断点接着做。</p>
+      ) : null}
+      {!dbReady ? (
+        <p className="local-index-note is-error">请先连接微信数据，然后再建立图片文字索引。</p>
+      ) : null}
+
+      {/* footer：左侧覆盖量说明，右侧操作。按钮不再悬浮在 stats 右下角。 */}
+      <div className="local-index-footer">
+        {established && !running && !paused && coverage ? (
+          <p className="local-index-note">{describeImageTextCoverage(coverage)}</p>
+        ) : null}
+        <div className="local-index-actions">
+          {running ? (
+            <>
+              <Button
+                size="sm"
+                variant="outline"
+                disabled={pending !== null}
+                onClick={() => void pause()}
+              >
+                {pending === 'pause' ? '暂停中…' : '暂停'}
+              </Button>
+              <Button
+                size="sm"
+                variant="outline"
+                disabled={pending !== null}
+                onClick={() => void cancel()}
+              >
+                {pending === 'cancel' ? '取消中…' : '取消'}
+              </Button>
+            </>
+          ) : null}
+
+          {/* 中断过就一定要有「继续」：checkpoint 是保留的，续做不会从头开始。 */}
+          {interrupted ? (
+            <Button size="sm" disabled={pending !== null} onClick={() => void resume()}>
+              {pending === 'resume' ? '继续中…' : '继续'}
+            </Button>
+          ) : null}
+
+          {!running && !interrupted ? (
+            <>
+              {/* 修复类操作各自只在很窄的场景有用，收进「更多」；
+                平铺出来会让用户面对四个都带「索引」字样的按钮。 */}
+              {established || failureCount > 0 ? (
+                <DropdownMenu>
+                  <DropdownMenuTrigger asChild>
+                    <Button
+                      size="sm"
+                      variant="outline"
+                      aria-label="更多操作"
+                      disabled={pending !== null}
+                    >
+                      ···
+                    </Button>
+                  </DropdownMenuTrigger>
+                  <DropdownMenuContent align="end">
+                    {established ? (
+                      <DropdownMenuItem disabled={pending !== null} onSelect={() => void repair()}>
+                        图片内容搜不到？修复搜索索引
+                      </DropdownMenuItem>
+                    ) : null}
+                    {failureCount > 0 ? (
+                      <DropdownMenuItem
+                        disabled={pending !== null}
+                        onSelect={() => void resetFailures()}
+                      >
+                        {`重试识别失败的图片（${failureCount.toLocaleString()} 张）`}
+                      </DropdownMenuItem>
+                    ) : null}
+                  </DropdownMenuContent>
+                </DropdownMenu>
+              ) : null}
+              <Button
+                size="sm"
+                disabled={!dbReady || pending !== null || counting}
+                onClick={() => void requestStart()}
+              >
+                {/* 统计图片数期间按钮是禁用的；这时如实说"统计中"，
+                  而不是灰着一个写着「更新索引」却点不动的按钮。 */}
+                {counting ? '统计中…' : established ? '更新索引' : '建立图片文字索引'}
+              </Button>
+            </>
+          ) : null}
+        </div>
+      </div>
+
+      {/* 确认弹窗复用 `confirming` 状态；保持与侧栏卡片一致的二次确认语义。 */}
+      {confirming ? (
+        <div className="local-index-confirm" role="dialog" aria-label="建立图片文字索引">
+          <p>
+            当前账号检测到约{' '}
+            <strong>
+              {confirmCount === null ? '未知数量' : confirmCount.toLocaleString()} 条图片消息
+            </strong>
+            。
+          </p>
+          <p>
+            识别仅在本机进行，原始图片不会因为本地识别而自动上传；可能需要较长时间，可以暂停稍后继续。
+          </p>
+          <div className="local-index-actions">
+            <Button size="sm" variant="outline" onClick={() => setConfirming(false)}>
+              取消
+            </Button>
+            <Button
+              size="sm"
+              onClick={() => {
+                setConfirming(false)
+                void start({ ...(sinceMs ? { sinceMs } : {}) })
+              }}
+            >
+              开始索引
+            </Button>
+          </div>
+        </div>
+      ) : null}
+    </section>
+  )
+}
@@ -0,0 +1,142 @@
+import React from 'react'
+import { useKnowledgeStatus } from './hooks/useKnowledgeStatus'
+import {
+  formatBytes,
+  formatIndexDate,
+  formatKnowledgeProcessed,
+  knowledgeIsStale,
+  knowledgeStateLabel
+} from './searchFormatters'
+import {
+  IndexStatusBadge,
+  type IndexStatusTone
+} from '../../features/settings/components/IndexStatusBadge'
+import { Button } from '../ui'
+
+interface KnowledgeIndexCardProps {
+  dbReady: boolean
+  onNotice: (message: string) => void
+}
+
+/**
+ * 「聊天记录索引」卡片 —— 问问微信背后那份索引。
+ *
+ * 布局与服务端语义分得很开：
+ * - 数据全部来自 `useKnowledgeStatus`（与问问微信页**同一个 hook**，不存在第二套状态源）；
+ * - 文案全部来自 `searchFormatters`（那边的词汇表有约束：区分「正在追新」与「正在补齐历史」、
+ *   不可查询时吃掉「可用」前缀、禁止笼统的「已同步」）；
+ * - 这里只负责**信息层级**：标题 → 状态徽章 → 三个核心数字 → 操作。
+ *
+ * 状态刻意从标题里拆出来单独做成徽章：混在标题行里时既没有视觉权重差、也扫不到。
+ */
+export function KnowledgeIndexCard({
+  dbReady,
+  onNotice
+}: KnowledgeIndexCardProps): React.ReactElement {
+  const {
+    knowledgeStatus,
+    knowledgeSyncing,
+    syncStarting,
+    cancelRequested,
+    startKnowledgeSync,
+    cancelKnowledgeSync
+  } = useKnowledgeStatus({ dbReady, onNotice })
+
+  const indexed = knowledgeStatus?.indexedMessageCount ?? 0
+  const usable = indexed > 0 || (knowledgeStatus?.indexedChunkCount ?? 0) > 0
+  const stale = knowledgeIsStale(knowledgeStatus)
+  const failed = knowledgeStatus?.state === 'error'
+
+  /** 徽章色只用主题 token，不硬编码彩虹色。 */
+  const tone: IndexStatusTone = failed
+    ? 'error'
+    : !usable
+      ? 'idle'
+      : stale || knowledgeSyncing
+        ? 'warn'
+        : 'ok'
+
+  const cancellable = knowledgeStatus?.pass?.cancellable === true
+
+  return (
+    <section className="local-index-card" aria-label="聊天记录索引状态">
+      <header className="local-index-card-head">
+        <div className="local-index-card-title">
+          <h3>聊天记录索引</h3>
+          <p>用于搜索、问问微信和群聊统计</p>
+        </div>
+        <IndexStatusBadge label={knowledgeStateLabel(knowledgeStatus)} tone={tone} />
+      </header>
+
+      {usable ? (
+        <div className="local-index-stats">
+          <div className="local-index-stat">
+            <strong className="local-index-stat-value">{indexed.toLocaleString()}</strong>
+            <span className="local-index-stat-label">已收录消息</span>
+          </div>
+          <div className="local-index-stat">
+            <strong className="local-index-stat-value">
+              {formatBytes(knowledgeStatus?.databaseBytes ?? 0)}
+            </strong>
+            <span className="local-index-stat-label">占用空间</span>
+          </div>
+          <div className="local-index-stat">
+            <strong className="local-index-stat-value">
+              {knowledgeStatus?.indexLatestAt
+                ? formatIndexDate(knowledgeStatus.indexLatestAt)
+                : '—'}
+            </strong>
+            <span className="local-index-stat-label">最后更新</span>
+          </div>
+        </div>
+      ) : (
+        <p className="local-index-note">
+          还没有建立索引。建立后「问问微信」才能跨会话检索历史消息。
+        </p>
+      )}
+
+      {failed && knowledgeStatus?.lastError ? (
+        <p className="local-index-note is-error">{knowledgeStatus.lastError}</p>
+      ) : null}
+
+      {!dbReady ? (
+        <p className="local-index-note is-error">请先连接微信数据，然后再建立聊天记录索引。</p>
+      ) : null}
+
+      {/* footer：左侧状态说明，右侧操作。按钮不再悬浮在 stats 右下角。 */}
+      <div className="local-index-footer">
+        <p className="local-index-note">
+          {knowledgeSyncing && knowledgeStatus?.pass
+            ? formatKnowledgeProcessed(knowledgeStatus)
+            : failed
+              ? '上次同步没有完成，可以重新同步。'
+              : stale
+                ? '有较新的聊天记录尚未入索引，点「同步最新记录」追平。'
+                : '索引数据仅保存在本机，可随时重新建立。'}
+        </p>
+        <div className="local-index-actions">
+          {knowledgeSyncing && cancellable ? (
+            <Button
+              size="sm"
+              variant="outline"
+              disabled={cancelRequested}
+              onClick={() => void cancelKnowledgeSync()}
+            >
+              {cancelRequested ? '正在取消…' : '取消同步'}
+            </Button>
+          ) : null}
+          {/* 用项目 `Button` 而不是自写 `<button>` + 自己的主色样式：
+            `hsl(var(--tm-primary))` 在这套主题变量下不生效（实测背景回退成了浏览器默认灰），
+            而复用组件同时还省掉了一套与全站不一致的按钮外观。 */}
+          <Button
+            size="sm"
+            disabled={!dbReady || syncStarting || (knowledgeSyncing && !cancellable)}
+            onClick={() => void startKnowledgeSync()}
+          >
+            {knowledgeSyncing ? '同步中…' : usable ? '同步最新记录' : '建立索引'}
+          </Button>
+        </div>
+      </div>
+    </section>
+  )
+}
@@ -214,11 +214,11 @@ export function AgentHubWorkspace({
          <ul>
            <li>
              <i />
-              本机 HTTP 通信
+              在主进程内运行，不开放本地端口
            </li>
            <li>
              <i />
-              入站请求鉴权
+              对话记录可回看
            </li>
            <li>
              <i />
@@ -58,6 +58,67 @@ export const API_ENDPOINTS: ApiEndpoint[] = [
      { key: 'q', label: '待解析标识', required: true, placeholder: '昵称、wxid 或 md5' }
    ]
  }),
+  endpoint('app-capabilities', {
+    name: '应用能力',
+    description: '查看数据库、Query、Automation 与运行环境的能力状态。'
+  }),
+  endpoint('group-exit-monitor', {
+    name: '退群监控状态',
+    description: '查看监控开关、监控群范围、事件数量和运行状态。'
+  }),
+  endpoint('group-exit-monitor-update', {
+    name: '配置退群监控',
+    description: '设置监控群范围或启用/关闭退群监控，不会修改退群通知 Automation。',
+    body: true
+  }),
+  endpoint('group-exit-events', {
+    name: '退群事件历史',
+    description: '按群和时间窗口查询退群事件，最多返回 200 条。',
+    parameters: [
+      { key: 'conversationId', label: '群会话 ID', placeholder: 'xxx@chatroom' },
+      { key: 'since', label: '开始时间', placeholder: '2026-10-01T00:00:00+07:00' },
+      { key: 'until', label: '结束时间', placeholder: '2026-10-02T23:59:59+07:00' },
+      { key: 'limit', label: '数量上限', placeholder: '50，最大 200' }
+    ]
+  }),
+  endpoint('group-member-stats', {
+    name: '群成员活跃统计',
+    description: '查询指定群在时间窗口内的活跃成员、沉默成员和数据完整性。',
+    parameters: [
+      { key: 'conversationId', label: '群会话 ID', required: true, placeholder: 'xxx@chatroom' },
+      { key: 'start', label: '开始时间', required: true, placeholder: '2026-10-01T00:00:00+07:00' },
+      { key: 'end', label: '结束时间', required: true, placeholder: '2026-10-02T23:59:59+07:00' }
+    ]
+  }),
+  endpoint('automations', {
+    name: '自动化规则',
+    description: '列出自动化规则，可按类型和启用状态筛选。',
+    parameters: [
+      { key: 'type', label: '规则类型', placeholder: 'daily_report / scheduled_report / leave_notification' },
+      { key: 'enabled', label: '启用状态', placeholder: 'true 或 false' }
+    ]
+  }),
+  endpoint('automation-create', {
+    name: '创建自动化规则',
+    description: '创建一条默认停用的自动化规则。',
+    body: true
+  }),
+  endpoint('automation-validate', {
+    name: '校验自动化规则',
+    description: '检查规则字段、目标标识和预计影响，不保存也不执行。',
+    body: true
+  }),
+  endpoint('automation-executions', {
+    name: '自动化执行记录',
+    description: '分页上限内查询执行结果，可按规则、状态和时间筛选。',
+    parameters: [
+      { key: 'ruleId', label: '规则 ID', placeholder: 'ruleId' },
+      { key: 'status', label: '状态', placeholder: 'running / success / failed' },
+      { key: 'since', label: '开始时间', placeholder: '2026-10-01T00:00:00+07:00' },
+      { key: 'until', label: '结束时间', placeholder: '2026-10-02T23:59:59+07:00' },
+      { key: 'limit', label: '数量上限', placeholder: '50，最大 200' }
+    ]
+  }),
  endpoint('report', {
    name: '群聊日报导出',
    description: '通过内置模板导出群聊日报 HTML 与 PNG。',
@@ -1,4 +1,4 @@
-export type ApiMethod = 'GET' | 'POST'
+export type ApiMethod = 'GET' | 'POST' | 'PATCH'
 export type { ApiTokenStatus } from '../../../../../shared/local-api-auth'

 export interface ApiParameter {
@@ -4,8 +4,19 @@ export function buildApiUrl(
  path: string,
  params: Record<string, string>
 ): string {
-  const url = new URL(path, `http://${host}:${port}`)
+  let resolvedPath = path
+  const pathParameterKeys = new Set(
+    Object.keys(params).filter((key) => path.includes(`{${key}}`))
+  )
  Object.entries(params).forEach(([key, value]) => {
+    const normalized = value.trim()
+    if (normalized && resolvedPath.includes(`{${key}}`)) {
+      resolvedPath = resolvedPath.replace(`{${key}}`, encodeURIComponent(normalized))
+    }
+  })
+  const url = new URL(resolvedPath, `http://${host}:${port}`)
+  Object.entries(params).forEach(([key, value]) => {
+    if (pathParameterKeys.has(key)) return
    if (value.trim()) url.searchParams.set(key, value.trim())
  })
  return url.toString()
@@ -0,0 +1,57 @@
+import * as React from 'react'
+import { SegmentedControl, SegmentedControlItem } from '../../components/ui'
+import { AUTOMATION_RULE_TYPE_LABELS, type AutomationRuleType } from '../../../../shared/automation'
+
+/**
+ * Tab 上能出现的规则类型。
+ *
+ * `scheduled_report` 是真实的 `AutomationRuleType`；这个别名只为让调用点的
+ * 语义更清楚 —— 它等价于 `AutomationRuleType`。
+ */
+export type AutomationRuleTabType = AutomationRuleType
+
+/**
+ * Tab 顺序 = 触发方式的演进：消息触发 → 时间触发 → 系统事件触发。
+ */
+const RULE_TABS: AutomationRuleTabType[] = [
+  'daily_report',
+  'scheduled_report',
+  'leave_notification'
+]
+
+const TAB_LABELS: Record<AutomationRuleTabType, string> = AUTOMATION_RULE_TYPE_LABELS
+
+/**
+ * AutomationRuleTypeTabs —— 编辑器上方的**规则类型切换**。
+ *
+ * 它切的是"现在在编辑哪一类规则"，**不是**"把当前这条规则改成另一类"。
+ * 文案必须说清这一点：叫「规则类型」时，用户会以为它是当前规则的类型选择器，
+ * 于是"切一下再保存"就意外改到了另一条规则上。
+ *
+ * 两项都是**真规则**：切换只改前端选中态，不碰任何已保存的配置。
+ */
+
+export function AutomationRuleTypeTabs({
+  value,
+  onValueChange
+}: {
+  value: AutomationRuleTabType
+  onValueChange: (next: AutomationRuleTabType) => void
+}): React.ReactElement {
+  return (
+    <div className="automation-rule-type-tabs">
+      <span className="automation-rule-type-label">切换规则类型</span>
+      <SegmentedControl
+        value={value}
+        onValueChange={(next) => onValueChange(next as AutomationRuleTabType)}
+        aria-label="切换规则类型"
+      >
+        {RULE_TABS.map((type) => (
+          <SegmentedControlItem key={type} value={type}>
+            {TAB_LABELS[type]}
+          </SegmentedControlItem>
+        ))}
+      </SegmentedControl>
+    </div>
+  )
+}
@@ -0,0 +1,798 @@
+import * as React from 'react'
+import {
+  BUILTIN_DAILY_REPORT_RULE_ID,
+  BUILTIN_LEAVE_NOTIFICATION_RULE_ID,
+  calculateNextRunAt,
+  normalizeScheduledReportConfig,
+  type AutomationExecution,
+  type AutomationRule,
+  type AutomationRuleDraft,
+  type AutomationStatusSummary
+} from '../../../../shared/automation'
+import {
+  Button,
+  SegmentedControl,
+  SegmentedControlItem,
+  Spinner,
+  useToast
+} from '../../components/ui'
+import { RuleListPanel } from './RuleListPanel'
+import { RuleEditorPanel } from './RuleEditorPanel'
+import { ExecutionLogPanel } from './ExecutionLogPanel'
+import { ExecutionDetailDrawer } from './ExecutionDetailDrawer'
+import {
+  AutomationRuleTypeTabs,
+  type AutomationRuleTabType
+} from './AutomationRuleTypeTabs'
+import { ScheduledReportEditor } from './ScheduledReportEditor'
+import { ScheduledReportRuleList } from './ScheduledReportRuleList'
+import { ScheduledReportNotificationPanel } from './ScheduledReportNotificationPanel'
+import {
+  createDraftFromRule,
+  createEmptyScheduledReportDraft,
+  draftToAutomationRuleDraft,
+  formatNextRunAt,
+  resolveGroupDisplayName,
+  type ScheduledReportDraft
+} from './model/scheduled-report-model'
+import {
+  LeaveNotificationEditor,
+  type LeaveNotificationSaveInput
+} from './LeaveNotificationEditor'
+import type { LeaveNotificationContactOption } from './LeaveNotificationTargetPicker'
+import { UNAVAILABLE_STATUS, automationApi, isAutomationApiAvailable, type AutomationGroupOption } from './model/api'
+
+/**
+ * AutomationWorkspace —— 自动化一级菜单。
+ *
+ * 两个 tab：规则 / 执行日志。顶部三块状态（监听 / 发送能力 / 今日执行）
+ * **全部来自 main 的真实能力**，渲染层不做任何平台猜测 ——
+ * 粗平台判断在这里是错的：装了 macOS 但没绑定个人微信，照样发不出去。
+ *
+ * 「规则」内部再分一层**规则类型**：@我生成日报 / 退群通知。
+ * 两种类型现在都是**真实规则**，保存都会落盘。
+ */
+
+type AutomationTab = 'rules' | 'logs'
+
+/** 由「退群监控」深链过来时的请求：直接打开指定规则类型。 */
+export interface AutomationOpenRuleRequest {
+  ruleType: AutomationRuleTabType
+  /** 每次点击都要能重新触发，所以带一个自增/时间戳。 */
+  requestId: number
+}
+
+export interface AutomationWorkspaceProps {
+  dbReady: boolean
+  /** 发送能力不可用时，引导用户去设置页。 */
+  onOpenSendSettings?: () => void
+  /** 「管理监控群聊 →」：跳到退群监控的管理群聊页（纯导航）。 */
+  onOpenExitMonitorGroups?: () => void
+  /** 「更改模型」：复用现有的模型设置入口（定时日报的模型配置）。 */
+  onOpenModelSettings?: () => void
+  /** 「微信异常通知」依赖 Agent Hub：未就绪时给一个可操作的去处。 */
+  onOpenAgentHub?: () => void
+  /** 深链请求（来自退群监控页的「退群通知自动化」入口）。 */
+  openRuleRequest?: AutomationOpenRuleRequest | null
+}
+
+interface EditorState {
+  mode: 'create' | 'edit'
+  rule: AutomationRule | null
+  /** 用 Tab 类型：定时日报只在本层存在，不进真实 `AutomationRule`。 */
+  ruleType: AutomationRuleTabType
+}
+
+export function AutomationWorkspace({
+  dbReady,
+  onOpenSendSettings,
+  onOpenExitMonitorGroups,
+  onOpenModelSettings,
+  onOpenAgentHub,
+  openRuleRequest
+}: AutomationWorkspaceProps): React.ReactElement {
+  const { toast } = useToast()
+  const apiAvailable = React.useMemo(() => isAutomationApiAvailable(), [])
+
+  const [tab, setTab] = React.useState<AutomationTab>('rules')
+  const [status, setStatus] = React.useState<AutomationStatusSummary>(UNAVAILABLE_STATUS)
+  const [rules, setRules] = React.useState<AutomationRule[]>([])
+  const [groups, setGroups] = React.useState<AutomationGroupOption[]>([])
+  const [executions, setExecutions] = React.useState<AutomationExecution[]>([])
+  const [sendableContacts, setSendableContacts] = React.useState<LeaveNotificationContactOption[]>([])
+  const [monitoredGroupCount, setMonitoredGroupCount] = React.useState(0)
+  /** 已监控群聊清单（含显示名）—— 退群通知「通知群聊」二次勾选的候选项。 */
+  const [monitoredGroups, setMonitoredGroups] = React.useState<AutomationGroupOption[]>([])
+  const [loading, setLoading] = React.useState(true)
+  const [clearing, setClearing] = React.useState(false)
+  const [saving, setSaving] = React.useState(false)
+  const [busyRuleId, setBusyRuleId] = React.useState<string | null>(null)
+  const [editor, setEditor] = React.useState<EditorState | null>(null)
+  const [selectedExecution, setSelectedExecution] = React.useState<AutomationExecution | null>(null)
+  const [pendingDelete, setPendingDelete] = React.useState<AutomationRule | null>(null)
+
+  const reload = React.useCallback(async (): Promise<void> => {
+    const [
+      nextStatus,
+      nextRules,
+      nextGroups,
+      nextExecutions,
+      nextContacts,
+      nextMonitored,
+      nextMonitoredGroups
+    ] = await Promise.all([
+      automationApi.getStatus(),
+      automationApi.listRules(),
+      automationApi.listGroups(),
+      automationApi.listExecutions(100),
+      // 「指定好友」候选、已监控群聊数量与清单：都来自既有能力，不在自动化里另存一份。
+      automationApi.listSendableContacts(),
+      automationApi.getMonitoredGroupCount(),
+      automationApi.getMonitoredGroups()
+    ])
+    setStatus(nextStatus)
+    setRules(nextRules)
+    setGroups(nextGroups)
+    setExecutions(nextExecutions)
+    setSendableContacts(nextContacts)
+    setMonitoredGroupCount(nextMonitored)
+    setMonitoredGroups(nextMonitoredGroups)
+  }, [])
+
+  React.useEffect(() => {
+    let disposed = false
+    setLoading(true)
+    void reload().finally(() => {
+      if (!disposed) setLoading(false)
+    })
+    return () => {
+      disposed = true
+    }
+  }, [reload])
+
+  // 切到日志 tab 时刷新一次：规则跑完后台不会主动推事件给渲染层，
+  // 用户点进来看到的是上一次加载的快照才是真的误导。
+  React.useEffect(() => {
+    if (tab !== 'logs') return
+    void reload()
+  }, [tab, reload])
+
+  /**
+   * 「退群监控 → 退群通知自动化」的深链。
+   *
+   * 落到「规则 → 退群通知」这一层，而不是只跳到自动化首页 ——
+   * 否则用户还得自己再找一次。请求对象本身在父层是稳定的，
+   * 所以依赖整个对象；requestId 变化就代表又点了一次。
+   */
+  React.useEffect(() => {
+    if (!openRuleRequest) return
+    setTab('rules')
+    setEditor({ mode: 'edit', rule: null, ruleType: openRuleRequest.ruleType })
+  }, [openRuleRequest])
+
+  /**
+   * 当前编辑目标（日报）—— **UI 显示与保存共用这一个**。
+   *
+   * 这里保留"回落到内置日报"的兜底，但它同时驱动下面的「正在编辑：X」，
+   * 所以即使回落发生，用户也能看到真实目标，不会再出现"以为在改 A、实际改了 B"。
+   */
+  const dailyReportTarget =
+    editor?.ruleType === 'daily_report'
+      ? (editor.rule ?? rules.find((item) => item.id === BUILTIN_DAILY_REPORT_RULE_ID) ?? null)
+      : null
+
+  /** 退群通知是 singleton 规则，永远从规则表里取（不存在则由 main 侧播种）。 */
+  const leaveNotificationRule =
+    rules.find((item) => item.id === BUILTIN_LEAVE_NOTIFICATION_RULE_ID) ?? null
+
+
+  /** 「指定好友」目标的显示名（首页卡片用；拿不到时留空，不伪造）。 */
+  const leaveContactName = React.useMemo((): string => {
+    const target = leaveNotificationRule?.leaveNotification?.target
+    if (!target || target.type !== 'contact') return ''
+    return sendableContacts.find((contact) => contact.id === target.contactId)?.name ?? ''
+  }, [leaveNotificationRule, sendableContacts])
+
+  /**
+   * 记住"上一次在日报这一类里编辑的那条"。
+   *
+   * 切类型只是换一类规则来编辑，切回来时应当回到刚才那条 ——
+   * 否则用户从"某条日报"切走再切回，会被悄悄带到内置日报上。
+   */
+  const lastDailyTargetRef = React.useRef<AutomationRule | null>(null)
+
+  /** 切到「@我生成日报」时的编辑目标：上次那条 → 内置日报 → 新建。 */
+  const resolveDailyTarget = React.useCallback((): AutomationRule | null => {
+    return (
+      lastDailyTargetRef.current ??
+      rules.find((item) => item.id === BUILTIN_DAILY_REPORT_RULE_ID) ??
+      null
+    )
+  }, [rules])
+
+  /*
+   * ---------- 定时日报（真实规则，真写入）----------
+   *
+   * `scheduled_report` 是真实的 `AutomationRuleType`，所以这一层
+   * **不做任何数据副本**：列表 = `rules.filter(ruleType==='scheduled_report')`，
+   * 保存 / 启停 / 删除 / 立即执行全部经 `automation:*` IPC 落到
+   * `AutomationRuleStore` 与 `AutomationService`。
+   */
+  const scheduledRules = React.useMemo(
+    () => rules.filter((rule) => rule.ruleType === 'scheduled_report'),
+    [rules]
+  )
+  /** `null` = 看列表；非空 = 在编辑器里（`ruleId` 为 `null` 表示新建）。 */
+  const [scheduledEditor, setScheduledEditor] = React.useState<{ ruleId: string | null } | null>(
+    null
+  )
+  const [scheduledDraft, setScheduledDraft] = React.useState<ScheduledReportDraft | null>(null)
+
+  /** 「正在编辑」显示的名字（与保存目标同源）。 */
+  const editorTargetName = React.useMemo((): string => {
+    if (!editor) return ''
+    if (editor.ruleType === 'leave_notification') return leaveNotificationRule?.name ?? '退群通知'
+    if (editor.ruleType === 'scheduled_report') {
+      return scheduledDraft?.name?.trim() || '新建定时日报'
+    }
+    return dailyReportTarget?.name ?? '新建自动化'
+  }, [editor, leaveNotificationRule, dailyReportTarget, scheduledDraft])
+
+  /**
+   * 首页汇总卡：数量 / 运行中 / 最近一次「下次执行」。
+   * 数据直接来自真实规则 —— 加载中不显示（避免闪一个假的"0 条"）。
+   */
+  const scheduledSummary = React.useMemo(() => {
+    if (loading) return null
+    const now = new Date()
+    const running = scheduledRules.filter((rule) => rule.enabled)
+    const nextRuns = running
+      .map((rule) => normalizeScheduledReportConfig(rule.scheduledReport))
+      .filter((config) => !config.targetNeedsReview)
+      .map((config) => calculateNextRunAt(config.schedule.time, now))
+      .sort()
+    return {
+      total: scheduledRules.length,
+      running: running.length,
+      nextRunLabel: nextRuns.length ? formatNextRunAt(nextRuns[0]) : '—'
+    }
+  }, [scheduledRules, loading])
+
+  /** 群标识 → 显示名（列表与编辑器共用，保证任何地方都不出现裸 id）。 */
+  const resolveScheduledGroupDisplay = React.useCallback(
+    (raw: string): string => resolveGroupDisplayName(raw, groups),
+    [groups]
+  )
+
+  const openScheduledEditor = (rule: AutomationRule | null): void => {
+    setScheduledDraft(rule ? createDraftFromRule(rule) : createEmptyScheduledReportDraft())
+    setScheduledEditor({ ruleId: rule?.id ?? null })
+  }
+
+  const closeScheduledEditor = (): void => {
+    setScheduledEditor(null)
+    setScheduledDraft(null)
+  }
+
+  /** 保存：**真的**创建 / 更新一条 `scheduled_report` 规则。 */
+  const handleScheduledSave = async (draft: ScheduledReportDraft): Promise<void> => {
+    if (!scheduledEditor) return
+    const target = scheduledEditor.ruleId
+      ? (scheduledRules.find((rule) => rule.id === scheduledEditor.ruleId) ?? null)
+      : null
+    // 编辑态但目标不存在 ⇒ 报错，绝不静默新建（与 @我日报同一条红线）。
+    if (scheduledEditor.ruleId && !target) {
+      toast({
+        description: '要编辑的定时日报已不存在，请返回列表重新打开',
+        variant: 'destructive',
+        duration: 3600
+      })
+      return
+    }
+    setSaving(true)
+    const payload = draftToAutomationRuleDraft(draft)
+    const saved =
+      scheduledEditor.ruleId && target
+        ? await automationApi.updateRule(target.id, payload)
+        : await automationApi.createRule(payload)
+    setSaving(false)
+    if (!saved) {
+      toast({ description: '保存失败，请稍后重试', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) =>
+      current.some((item) => item.id === saved.id)
+        ? current.map((item) => (item.id === saved.id ? saved : item))
+        : [...current, saved]
+    )
+    closeScheduledEditor()
+    toast({
+      description: scheduledEditor.ruleId ? `已保存「${saved.name}」` : `已创建「${saved.name}」`,
+      duration: 2800
+    })
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  /** 立即执行：走与 scheduler 完全相同的执行链路（manual trigger）。 */
+  const handleScheduledRunNow = async (rule: AutomationRule): Promise<void> => {
+    setBusyRuleId(rule.id)
+    const result = await automationApi.runScheduledReportRule(rule.id)
+    setBusyRuleId(null)
+    toast({
+      description: result.success
+        ? `「${rule.name}」已执行完成`
+        : result.error || '定时日报执行失败，请查看执行日志',
+      variant: result.success ? undefined : 'destructive',
+      duration: 3600
+    })
+    // 执行结果落在 Automation Execution Log，切到日志 tab 就能看到。
+    void reload()
+  }
+
+  /** 启停：真写 store。 */
+  const handleScheduledToggle = async (rule: AutomationRule, enabled: boolean): Promise<void> => {
+    setBusyRuleId(rule.id)
+    const updated = await automationApi.setRuleEnabled(rule.id, enabled)
+    setBusyRuleId(null)
+    if (!updated) {
+      toast({ description: '切换失败，规则状态未改变', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) => current.map((item) => (item.id === updated.id ? updated : item)))
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  /** 删除：走统一的确认弹层（与其它规则类型一致）。 */
+  const handleScheduledDelete = (rule: AutomationRule): void => {
+    setPendingDelete(rule)
+  }
+
+  const handleToggle = async (rule: AutomationRule, enabled: boolean): Promise<void> => {
+    setBusyRuleId(rule.id)
+    const updated = await automationApi.setRuleEnabled(rule.id, enabled)
+    setBusyRuleId(null)
+    if (!updated) {
+      toast({ description: '切换失败，规则状态未改变', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) => current.map((item) => (item.id === updated.id ? updated : item)))
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  const handleSave = async (draft: AutomationRuleDraft): Promise<void> => {
+    if (!editor) return
+
+    const target = editor.ruleType === 'daily_report' ? dailyReportTarget : editor.rule
+
+    /*
+     * 编辑态但目标不存在 ⇒ **报错，绝不许静默新建**。
+     *
+     * 这里以前写的是 `editor.mode === 'create' || !target ? createRule : updateRule` ——
+     * 把"没有目标"当成了"新建"，于是用户以为在改 A、实际多出一条新规则。
+     * 新建只能由 `mode === 'create'` 决定；`edit` 且无目标是不可恢复的状态，必须说出来。
+     */
+    if (editor.mode === 'edit' && !target) {
+      toast({
+        description: '要编辑的规则已不存在，请返回列表重新打开',
+        variant: 'destructive',
+        duration: 3600
+      })
+      return
+    }
+
+    setSaving(true)
+    let saved: AutomationRule | null = null
+    if (editor.mode === 'create') {
+      saved = await automationApi.createRule(draft)
+    } else if (target) {
+      saved = await automationApi.updateRule(target.id, draft)
+    }
+    setSaving(false)
+    if (!saved) {
+      toast({ description: '保存失败，请稍后重试', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) => {
+      const exists = current.some((item) => item.id === saved.id)
+      return exists
+        ? current.map((item) => (item.id === saved.id ? saved : item))
+        : [...current, saved]
+    })
+    setEditor(null)
+    toast({
+      description: editor.mode === 'create' ? `已创建「${saved.name}」` : `已保存「${saved.name}」`,
+      duration: 2800
+    })
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  /**
+   * 保存「退群通知」规则。
+   *
+   * 走独立的 singleton upsert 通道：这条规则 id 固定，保存永远不会多出一条。
+   * 保存成功后清掉迁移遗留的 `targetNeedsReview`（用户已经做过选择）。
+   */
+  const handleLeaveNotificationSave = async (
+    input: LeaveNotificationSaveInput
+  ): Promise<void> => {
+    // 静默 return 会让"点保存没反应"变成用户眼里的坏掉，必须说出来。
+    if (!leaveNotificationRule) {
+      toast({
+        description: '退群通知规则已不存在，请返回列表重新打开',
+        variant: 'destructive',
+        duration: 3600
+      })
+      return
+    }
+    setSaving(true)
+    const saved = await automationApi.saveLeaveNotificationRule({
+      ...leaveNotificationRule,
+      enabled: input.enabled,
+      leaveNotification: input.config
+    })
+    setSaving(false)
+    if (!saved) {
+      toast({ description: '保存失败，请稍后重试', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) =>
+      current.some((item) => item.id === saved.id)
+        ? current.map((item) => (item.id === saved.id ? saved : item))
+        : [...current, saved]
+    )
+    setEditor(null)
+    toast({ description: `已保存「${saved.name}」`, duration: 2800 })
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  const confirmDelete = async (): Promise<void> => {
+    if (!pendingDelete) return
+    const target = pendingDelete
+    setPendingDelete(null)
+    setBusyRuleId(target.id)
+    const removed = await automationApi.deleteRule(target.id)
+    setBusyRuleId(null)
+    if (!removed) {
+      toast({ description: '删除失败，规则仍然存在', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setRules((current) => current.filter((item) => item.id !== target.id))
+    toast({ description: `已删除「${target.name}」`, duration: 2800 })
+  }
+
+  const handleClearExecutions = async (): Promise<void> => {
+    setClearing(true)
+    const cleared = await automationApi.clearExecutions()
+    setClearing(false)
+    if (!cleared) {
+      toast({ description: '清空失败，请稍后重试', variant: 'destructive', duration: 3200 })
+      return
+    }
+    setExecutions([])
+    void automationApi.getStatus().then(setStatus)
+  }
+
+  const capability = status.sendCapability
+  const capabilityReady = capability.ready && capability.canSendText && capability.canSendImage
+  const showEditor = tab === 'rules' && editor !== null
+
+  return (
+    <div className="automation-page">
+      <header className="automation-page-header">
+        <div>
+          <h1>自动化</h1>
+          <p>当群里出现符合条件的消息时，TraceMemo 会自动执行你配置的动作。</p>
+        </div>
+        {tab === 'rules' && !showEditor ? (
+          <Button
+            onClick={() => {
+              lastDailyTargetRef.current = null
+              setEditor({ mode: 'create', rule: null, ruleType: 'daily_report' })
+            }}
+          >
+            新建自动化
+          </Button>
+        ) : null}
+      </header>
+
+      {!apiAvailable ? (
+        <p className="automation-notice warning">
+          自动化接口尚未就绪（可能正在启动或版本不匹配）。下面的数据可能不是最新的。
+        </p>
+      ) : null}
+      {!dbReady ? (
+        <p className="automation-notice warning">
+          微信数据库尚未连接，规则可以编辑，但无法读取群列表，也不会触发。
+        </p>
+      ) : null}
+
+      <section className="automation-status-bar" aria-label="自动化状态">
+        <div className="automation-status-card">
+          <span className="automation-status-label">消息监听</span>
+          <span className={`automation-status-value ${status.listening ? 'ok' : 'off'}`}>
+            {status.listening ? '运行中' : '未在监听'}
+          </span>
+        </div>
+
+        <div className="automation-status-card">
+          <span className="automation-status-label">发送能力</span>
+          <span className={`automation-status-value ${capabilityReady ? 'ok' : 'warn'}`}>
+            {capabilityReady ? '可以发送文字和图片' : capability.ready ? '发送能力不完整' : '尚未就绪'}
+          </span>
+          <small className="automation-status-note">{capability.message}</small>
+          {!capabilityReady && onOpenSendSettings ? (
+            <Button variant="link" size="sm" onClick={onOpenSendSettings}>
+              去设置发送能力
+            </Button>
+          ) : null}
+        </div>
+
+        <div className="automation-status-card">
+          <span className="automation-status-label">今日执行</span>
+          <span className="automation-status-value">
+            {status.todaySuccesses} / {status.todayExecutions}
+          </span>
+          <small className="automation-status-note">成功 / 总计</small>
+        </div>
+      </section>
+
+      {!showEditor ? (
+        <SegmentedControl
+          value={tab}
+          onValueChange={(value) => setTab(value as AutomationTab)}
+          aria-label="自动化视图"
+        >
+          <SegmentedControlItem value="rules">规则</SegmentedControlItem>
+          <SegmentedControlItem value="logs">执行日志</SegmentedControlItem>
+        </SegmentedControl>
+      ) : null}
+
+      {showEditor && editor ? (
+        <div className="automation-rule-type-shell">
+          {/*
+            切类型 = **换一条规则来编辑**，不是"把这条规则改成另一类"。
+
+            以前这里只改 `ruleType`，把上个类型的 `mode`/`rule` 一起带过来，
+            于是"从退群通知切到日报"会留下 `mode='edit' + rule=null` 的中间态，
+            保存时再隐式回落到内置日报 —— 用户以为在改退群通知，实际改掉了日报。
+            现在切换时**显式重算编辑目标**，并把当前目标名显示出来。
+          */}
+          <AutomationRuleTypeTabs
+            value={editor.ruleType}
+            onValueChange={(next) => {
+              if (next === editor.ruleType) return
+              // 离开定时日报时清掉它的编辑态，下次进来回到列表。
+              closeScheduledEditor()
+              if (next === 'scheduled_report') {
+                // 定时日报是**多条**规则：先给列表，不直接进编辑器。
+                setEditor({ mode: 'edit', rule: null, ruleType: 'scheduled_report' })
+                return
+              }
+              if (next === 'leave_notification') {
+                // 退群通知是 singleton：编辑目标就是它自己，不依赖规则列表。
+                setEditor({ mode: 'edit', rule: null, ruleType: 'leave_notification' })
+                return
+              }
+              // 日报可以有多条：目标必须是**具体某一条**或"新建"，
+              // 绝不允许留下 mode='edit' 却没有目标的状态。
+              const target = resolveDailyTarget()
+              lastDailyTargetRef.current = target
+              setEditor(
+                target
+                  ? { mode: 'edit', rule: target, ruleType: 'daily_report' }
+                  : { mode: 'create', rule: null, ruleType: 'daily_report' }
+              )
+            }}
+          />
+
+          {/*
+            编辑目标必须**可见**。
+            保存的目标与这里显示的名字来自**同一个变量** —— 不允许 UI 显示一个、保存改另一个。
+          */}
+          {editor.ruleType !== 'scheduled_report' || scheduledEditor ? (
+            <p className="automation-editing-target" data-testid="automation-editing-target">
+              正在编辑：<strong>{editorTargetName}</strong>
+            </p>
+          ) : null}
+
+          {editor.ruleType === 'daily_report' ? (
+            /*
+             * 编辑态却没有目标（规则被删/列表刷新后消失）—— 不能让用户对着一个
+             * 标题写着「编辑自动化」的空表单填半天，最后保存时报错。
+             */
+            editor.mode === 'edit' && !dailyReportTarget ? (
+              <div className="automation-notice warning automation-notice-stack">
+                <p>要编辑的规则已不存在（可能已被删除）。</p>
+                <Button variant="outline" size="sm" onClick={() => setEditor(null)}>
+                  返回规则列表
+                </Button>
+              </div>
+            ) : (
+              <RuleEditorPanel
+                mode={editor.mode}
+                rule={dailyReportTarget}
+                groups={groups}
+                saving={saving}
+                onCancel={() => setEditor(null)}
+                onSave={(draft) => void handleSave(draft)}
+              />
+            )
+          ) : editor.ruleType === 'scheduled_report' ? (
+            scheduledEditor && scheduledDraft ? (
+              <ScheduledReportEditor
+                // 换一条规则就重建编辑器，避免把上一条的草稿带过来。
+                key={scheduledEditor.ruleId ?? 'new'}
+                initialDraft={scheduledDraft}
+                groups={groups}
+                sendableContacts={sendableContacts}
+                saving={saving}
+                onBack={closeScheduledEditor}
+                onCancel={closeScheduledEditor}
+                onSave={(draft) => void handleScheduledSave(draft)}
+                {...(onOpenModelSettings ? { onOpenModelSettings } : {})}
+                {...(scheduledEditor.ruleId
+                  ? (() => {
+                      const rule = scheduledRules.find(
+                        (item) => item.id === scheduledEditor.ruleId
+                      )
+                      const legacy = rule?.scheduledReport?.legacyTarget
+                      return legacy ? { legacyTarget: legacy } : {}
+                    })()
+                  : {})}
+                notificationSlot={
+                  <ScheduledReportNotificationPanel
+                    rule={scheduledRules.find((item) => item.id === scheduledEditor.ruleId) ?? null}
+                    {...(onOpenAgentHub ? { onOpenAgentHub } : {})}
+                  />
+                }
+              />
+            ) : (
+              <ScheduledReportRuleList
+                rules={scheduledRules}
+                loading={loading}
+                busyRuleId={busyRuleId}
+                resolveGroupDisplay={resolveScheduledGroupDisplay}
+                // 与另外两个类型的「取消」等价：退出本类型，回到规则面板。
+                // 少了这一个，用户在定时日报列表上就只能靠切 tab 绕出去。
+                onBack={() => setEditor(null)}
+                onCreate={() => openScheduledEditor(null)}
+                onEdit={openScheduledEditor}
+                onRunNow={(rule) => void handleScheduledRunNow(rule)}
+                onToggle={(rule, enabled) => void handleScheduledToggle(rule, enabled)}
+                onDelete={handleScheduledDelete}
+              />
+            )
+          ) : leaveNotificationRule ? (
+            <LeaveNotificationEditor
+              rule={leaveNotificationRule}
+              contacts={sendableContacts}
+              monitoredCount={monitoredGroupCount}
+              monitoredGroups={monitoredGroups}
+              saving={saving}
+              {...(capabilityReady
+                ? {}
+                : {
+                    // 只提示，不阻断编辑。
+                    sendCapabilityWarning:
+                      '当前发送能力未就绪：退群事件仍会被记录，但通知发送会失败。'
+                  })}
+              onOpenMonitoredGroups={onOpenExitMonitorGroups}
+              onCancel={() => setEditor(null)}
+              onSave={(input) => void handleLeaveNotificationSave(input)}
+            />
+          ) : loading ? (
+            /*
+             * **正在读，不是"没有"。**
+             *
+             * 从「退群监控 → 退群通知自动化」深链进来时，编辑器是同步打开的，
+             * 而规则列表还在 IPC 回来的路上 —— 这一帧必然还没有规则。
+             * 旧写法在这里直接渲染「尚未就绪 / 版本不匹配」，把最正常的一种
+             * 加载状态说成了故障，必须区分开。
+             */
+            <div className="automation-loading">
+              <Spinner />
+              <span>正在读取退群通知规则…</span>
+            </div>
+          ) : (
+            /*
+             * 读完了**确实没有**这条规则。
+             *
+             * `builtin-leave-notification` 由主进程 `AutomationRuleStore` 首次加载时自动播种，
+             * 所以走到这里通常意味着：跑着的主进程还是迁移前的旧构建
+             * （渲染层已经升级，`out/main` 没有）。如实说明并给一个重试入口。
+             */
+            <div className="automation-notice warning automation-notice-stack">
+              <p>没有找到内置的退群通知规则。</p>
+              <small>
+                这条规则应由主进程在首次启动时自动创建。如果你刚更新过版本，
+                请重启 TraceMemo —— 已经跑起来的主进程不会自动换成新代码。
+              </small>
+              <Button
+                variant="outline"
+                size="sm"
+                disabled={loading}
+                onClick={() => {
+                  setLoading(true)
+                  void reload().finally(() => setLoading(false))
+                }}
+              >
+                重新加载
+              </Button>
+            </div>
+          )}
+        </div>
+      ) : tab === 'rules' ? (
+        <RuleListPanel
+          rules={rules}
+          groups={groups}
+          loading={loading}
+          busyRuleId={busyRuleId}
+          leaveNotification={
+            leaveNotificationRule
+              ? {
+                  rule: leaveNotificationRule,
+                  monitoredCount: monitoredGroupCount,
+                  contactName: leaveContactName
+                }
+              : null
+          }
+          scheduledReport={scheduledSummary}
+          onManageScheduledReport={() =>
+            setEditor({ mode: 'edit', rule: null, ruleType: 'scheduled_report' })
+          }
+          onToggle={(rule, enabled) => void handleToggle(rule, enabled)}
+          onEdit={(rule) => {
+            // 记住这条，切类型回来时能恢复 —— 不记住就会被悄悄换成内置日报。
+            lastDailyTargetRef.current = rule
+            setEditor({ mode: 'edit', rule, ruleType: rule.ruleType ?? 'daily_report' })
+          }}
+          onDelete={(rule) => setPendingDelete(rule)}
+          onCreate={() => {
+            // 新建是"还没有对象"，所以不记住任何规则。
+            lastDailyTargetRef.current = null
+            setEditor({ mode: 'create', rule: null, ruleType: 'daily_report' })
+          }}
+          onEditLeaveNotification={() =>
+            setEditor({ mode: 'edit', rule: null, ruleType: 'leave_notification' })
+          }
+        />
+      ) : (
+        <ExecutionLogPanel
+          executions={executions}
+          loading={loading}
+          clearing={clearing}
+          onSelect={setSelectedExecution}
+          onClear={() => void handleClearExecutions()}
+        />
+      )}
+
+      <ExecutionDetailDrawer
+        execution={selectedExecution}
+        onClose={() => setSelectedExecution(null)}
+      />
+
+      {pendingDelete ? (
+        <div className="automation-confirm-layer" role="presentation" onClick={() => setPendingDelete(null)}>
+          <div
+            className="automation-confirm"
+            role="alertdialog"
+            aria-modal="true"
+            aria-label="删除自动化"
+            onClick={(event) => event.stopPropagation()}
+          >
+            <h2>删除「{pendingDelete.name}」？</h2>
+            <p>删除后不会再触发，已有的执行记录会保留。</p>
+            <div className="automation-confirm-actions">
+              <Button variant="ghost" onClick={() => setPendingDelete(null)}>
+                取消
+              </Button>
+              <Button variant="destructive" onClick={() => void confirmDelete()}>
+                删除
+              </Button>
+            </div>
+          </div>
+        </div>
+      ) : null}
+    </div>
+  )
+}
@@ -0,0 +1,139 @@
+import * as React from 'react'
+import type { AutomationRuleDraft } from '../../../../shared/automation'
+
+/**
+ * EffectPreview —— 微信式效果模拟预览。
+ *
+ * **纯前端模拟**：不调用真实微信发送、不调日报生成、不调 AI。
+ * 它的全部作用是在用户按下保存之前，把「这条规则到底会做什么」演一遍。
+ *
+ * 与表单**逐项联动**（这是验收点）：
+ * - 改关键词 → 左侧模拟消息里的内容跟着变；
+ * - 改回复文案 → 右侧气泡跟着变；
+ * - 关闭「回复确认」→ 右侧气泡消失；
+ * - 关闭「必须真正 @我」→ 左侧不再显示 `@你`；
+ * - 关闭「生成日报」→ 不再出图片，并说明原因；
+ * - 关闭「发送图片」→ 日报仍会生成但不会进群，说明区分开。
+ */
+
+const SAMPLE_SENDER_NAME = '张三'
+
+function isActionEnabled(draft: AutomationRuleDraft, type: string): boolean {
+  return draft.actions.some((action) => action.type === type && action.enabled)
+}
+
+function replyText(draft: AutomationRuleDraft): string {
+  const action = draft.actions.find((item) => item.type === 'replyText' && item.enabled)
+  return action?.text?.trim() || '（未填写回复内容）'
+}
+
+/** 左侧模拟消息的正文：`@你 今日<关键词>`。关键词为空时给一句通用占位，而不是伪造关键词。 */
+function sampleIncomingText(draft: AutomationRuleDraft): string {
+  const keyword = draft.conditions.keyword.trim()
+  const body = keyword ? `今日${keyword}` : '今天群里有什么新消息'
+  return draft.conditions.requireMentionMe ? `@你 ${body}` : body
+}
+
+export function EffectPreview({ draft }: { draft: AutomationRuleDraft }): React.ReactElement {
+  const mentionMe = draft.conditions.requireMentionMe
+  const hasReply = isActionEnabled(draft, 'replyText')
+  const hasReport = isActionEnabled(draft, 'generateReport')
+  const hasSend = isActionEnabled(draft, 'sendReportImage')
+  const showReportImage = hasReport && hasSend
+
+  return (
+    <section className="automation-preview" aria-label="效果模拟预览">
+      <div className="automation-preview-heading">
+        <h3>效果预览</h3>
+        <span className="automation-preview-badge">仅预览，不会真实发送</span>
+      </div>
+
+      <div className="automation-preview-phone">
+        <div className="automation-preview-chat">
+          {/* 左侧：触发消息 */}
+          <div className="automation-chat-row incoming">
+            <span className="automation-chat-avatar" aria-hidden="true">
+              {SAMPLE_SENDER_NAME.slice(0, 1)}
+            </span>
+            <div className="automation-chat-body">
+              <span className="automation-chat-name">{SAMPLE_SENDER_NAME}</span>
+              <span className="automation-chat-bubble">{sampleIncomingText(draft)}</span>
+            </div>
+          </div>
+
+          {mentionMe ? (
+            <p className="automation-preview-hint">
+              只有当对方<strong>真正</strong> @ 你时才会触发；正文里手打的「@昵称」不算。
+            </p>
+          ) : (
+            <p className="automation-preview-hint">
+              当前不要求 @你，该会话内任何含关键词的消息都会触发。
+            </p>
+          )}
+
+          {/* 右侧：自动回复 */}
+          {hasReply ? (
+            <div className="automation-chat-row outgoing">
+              <div className="automation-chat-body">
+                <span className="automation-chat-name">我</span>
+                <span className="automation-chat-bubble">{replyText(draft)}</span>
+              </div>
+              <span className="automation-chat-avatar self" aria-hidden="true">
+                我
+              </span>
+            </div>
+          ) : (
+            <p className="automation-preview-hint muted">未启用「回复确认」，命中后不会先回一句。</p>
+          )}
+
+          {/* 右侧：日报图片 */}
+          {showReportImage ? (
+            <div className="automation-chat-row outgoing">
+              <div className="automation-chat-body">
+                <span className="automation-chat-name">我</span>
+                <figure className="automation-chat-image">
+                  <div className="automation-chat-image-canvas" aria-hidden="true">
+                    <span className="automation-chat-image-title">群聊日报</span>
+                    <span className="automation-chat-image-line" />
+                    <span className="automation-chat-image-line short" />
+                    <span className="automation-chat-image-line" />
+                    <span className="automation-chat-image-line short" />
+                  </div>
+                  <figcaption>日报图片预览</figcaption>
+                </figure>
+              </div>
+              <span className="automation-chat-avatar self" aria-hidden="true">
+                我
+              </span>
+            </div>
+          ) : (
+            <p className="automation-preview-hint muted">
+              {!hasReport
+                ? '未启用「生成日报」，命中后不会产出日报。'
+                : '日报仍会生成，但未启用发送，不会发到群里。'}
+            </p>
+          )}
+        </div>
+      </div>
+
+      <dl className="automation-preview-facts">
+        <div>
+          <dt>生效范围</dt>
+          <dd>
+            {draft.conditions.conversationIds.length
+              ? `已选 ${draft.conditions.conversationIds.length} 个群`
+              : '所有群聊'}
+          </dd>
+        </div>
+        <div>
+          <dt>触发间隔</dt>
+          <dd>{draft.cooldownSeconds > 0 ? `${draft.cooldownSeconds} 秒` : '不限制'}</dd>
+        </div>
+        <div>
+          <dt>自己发的消息</dt>
+          <dd>{draft.conditions.ignoreSelf ? '不触发' : '也会触发'}</dd>
+        </div>
+      </dl>
+    </section>
+  )
+}
@@ -0,0 +1,144 @@
+import * as React from 'react'
+import { AUTOMATION_EXECUTION_STATUS_LABELS } from '../../../../shared/automation'
+import type { AutomationExecution } from '../../../../shared/automation'
+import { Button } from '../../components/ui'
+import {
+  STEP_STATUS_LABELS,
+  describeSkippedStep,
+  formatClockTime,
+  formatDuration,
+  formatTriggerTime,
+  stepStatusTone
+} from './model/format'
+
+/**
+ * ExecutionDetailDrawer —— 执行详情抽屉（对应 Stitch 设计稿 3 的右半部分）。
+ *
+ * 用户来这里只想知道一件事：**到底坏在哪一步**。
+ * 所以步骤列表是主体，每个失败步骤都要把原因写在脸上；
+ * 「已跳过」的步骤还要说明是「上一步挂了」还是「规则本来就没配」。
+ *
+ * 抽屉而不是居中弹窗：详情是「从列表里钻进去看」，右侧抽屉保留了列表的位置感。
+ */
+
+export interface ExecutionDetailDrawerProps {
+  execution: AutomationExecution | null
+  onClose: () => void
+}
+
+export function ExecutionDetailDrawer({
+  execution,
+  onClose
+}: ExecutionDetailDrawerProps): React.ReactElement | null {
+  const open = execution !== null
+
+  React.useEffect(() => {
+    if (!open) return
+    const handleKeyDown = (event: KeyboardEvent): void => {
+      if (event.key === 'Escape') onClose()
+    }
+    window.addEventListener('keydown', handleKeyDown)
+    return () => window.removeEventListener('keydown', handleKeyDown)
+  }, [open, onClose])
+
+  if (!execution) return null
+
+  return (
+    <div className="automation-drawer-layer" role="presentation" onClick={onClose}>
+      <aside
+        className="automation-drawer"
+        role="dialog"
+        aria-modal="true"
+        aria-label="执行详情"
+        onClick={(event) => event.stopPropagation()}
+      >
+        <header className="automation-drawer-header">
+          <div>
+            <span className="automation-drawer-eyebrow">执行详情</span>
+            <h2>{execution.ruleName}</h2>
+          </div>
+          <Button variant="ghost" size="sm" onClick={onClose}>
+            关闭
+          </Button>
+        </header>
+
+        <div className="automation-drawer-body">
+          <dl className="automation-detail-facts">
+            <div>
+              <dt>触发时间</dt>
+              <dd>{formatTriggerTime(execution.triggerTime)}</dd>
+            </div>
+            <div>
+              <dt>来源</dt>
+              <dd>{execution.sourceDisplayName}</dd>
+            </div>
+            <div>
+              <dt>结果</dt>
+              <dd>
+                <span className={`automation-status-chip ${execution.status}`}>
+                  {AUTOMATION_EXECUTION_STATUS_LABELS[execution.status]}
+                </span>
+              </dd>
+            </div>
+            <div>
+              <dt>总耗时</dt>
+              <dd>{formatDuration(execution.durationMs)}</dd>
+            </div>
+          </dl>
+
+          {execution.errorSummary ? (
+            <p className="automation-detail-error">
+              <strong>失败原因：</strong>
+              {execution.errorSummary}
+            </p>
+          ) : null}
+
+          <section className="automation-detail-steps">
+            <h3>执行步骤</h3>
+            {execution.steps.length === 0 ? (
+              <p className="automation-detail-empty">
+                本次执行没有产生步骤记录（可能在发送能力检查阶段就被拦下了）。
+              </p>
+            ) : (
+              <ol className="automation-step-list">
+                {execution.steps.map((step) => (
+                  <li key={step.key} className={`automation-step ${stepStatusTone(step.status)}`}>
+                    <span className="automation-step-marker" aria-hidden="true" />
+                    <div className="automation-step-body">
+                      <div className="automation-step-title">
+                        <span className="automation-step-label">{step.label}</span>
+                        <span className="automation-step-status">
+                          {STEP_STATUS_LABELS[step.status]}
+                        </span>
+                      </div>
+                      <div className="automation-step-meta">
+                        {step.status === 'skipped' ? (
+                          <span>{describeSkippedStep(step, execution.steps)}</span>
+                        ) : (
+                          <>
+                            <span>{formatClockTime(step.startedAt)}</span>
+                            <span className="automation-step-sep" aria-hidden="true">
+                              →
+                            </span>
+                            <span>{formatClockTime(step.finishedAt)}</span>
+                            <span className="automation-step-duration">
+                              {formatDuration(step.durationMs)}
+                            </span>
+                          </>
+                        )}
+                      </div>
+                      {step.detail ? (
+                        <p className="automation-step-detail">{step.detail}</p>
+                      ) : null}
+                      {step.error ? <p className="automation-step-error">{step.error}</p> : null}
+                    </div>
+                  </li>
+                ))}
+              </ol>
+            )}
+          </section>
+        </div>
+      </aside>
+    </div>
+  )
+}
@@ -0,0 +1,124 @@
+import * as React from 'react'
+import { AUTOMATION_EXECUTION_STATUS_LABELS } from '../../../../shared/automation'
+import type { AutomationExecution } from '../../../../shared/automation'
+import { Button, Spinner } from '../../components/ui'
+import { STEP_STATUS_LABELS, formatDuration, formatTriggerTime, stepStatusTone } from './model/format'
+
+/**
+ * ExecutionLogPanel —— 执行日志列表（对应 Stitch 设计稿 3 的左半部分）。
+ *
+ * 每行是一条**用户层**记录：时间 / 规则名 / 来源 / 结果 / 耗时 / 步骤轨迹。
+ *
+ * 刻意不展示 wxid、localId、serverId、source XML、原始 payload 与图片路径 ——
+ * `AutomationExecution` 里本来就不带这些字段，这里也不允许从别处补。
+ */
+
+export interface ExecutionLogPanelProps {
+  executions: AutomationExecution[]
+  loading: boolean
+  clearing: boolean
+  onSelect: (execution: AutomationExecution) => void
+  onClear: () => void
+}
+
+/** 把 5 个步骤压成一串小圆点，一眼能看出「卡在哪一步」。 */
+function StepTrail({ execution }: { execution: AutomationExecution }): React.ReactElement {
+  if (!execution.steps.length) {
+    return <span className="automation-trail-empty">—</span>
+  }
+  return (
+    <span className="automation-trail" aria-label="执行步骤">
+      {execution.steps.map((step, index) => (
+        <React.Fragment key={step.key}>
+          {index > 0 ? <span className="automation-trail-link" aria-hidden="true" /> : null}
+          <span
+            className={`automation-trail-dot ${stepStatusTone(step.status)}`}
+            title={`${step.label}：${STEP_STATUS_LABELS[step.status]}`}
+            aria-label={`${step.label}：${STEP_STATUS_LABELS[step.status]}`}
+            role="img"
+          />
+        </React.Fragment>
+      ))}
+    </span>
+  )
+}
+
+export function ExecutionLogPanel({
+  executions,
+  loading,
+  clearing,
+  onSelect,
+  onClear
+}: ExecutionLogPanelProps): React.ReactElement {
+  return (
+    <div className="automation-log-panel">
+      <div className="automation-section-heading">
+        <h2>执行日志</h2>
+        <div className="automation-log-heading-actions">
+          <span className="automation-section-note">最近 {executions.length} 条</span>
+          <Button
+            variant="outline"
+            size="sm"
+            onClick={onClear}
+            disabled={clearing || executions.length === 0}
+          >
+            {clearing ? '清空中…' : '清空记录'}
+          </Button>
+        </div>
+      </div>
+
+      {loading ? (
+        <div className="automation-loading">
+          <Spinner />
+          <span>正在读取执行记录…</span>
+        </div>
+      ) : executions.length === 0 ? (
+        <div className="automation-empty-card">
+          <p>暂无执行记录。</p>
+          <small>规则命中并开始执行后，这里会留下完整的步骤轨迹。</small>
+        </div>
+      ) : (
+        <div className="automation-log-table" role="table" aria-label="执行日志">
+          <div className="automation-log-row head" role="row">
+            <span role="columnheader">时间</span>
+            <span role="columnheader">规则</span>
+            <span role="columnheader">来源</span>
+            <span role="columnheader">结果</span>
+            <span role="columnheader">耗时</span>
+            <span role="columnheader">步骤轨迹</span>
+            <span role="columnheader" />
+          </div>
+          {executions.map((execution) => (
+            <div key={execution.executionId} className="automation-log-row" role="row">
+              <span role="cell" className="automation-log-time">
+                {formatTriggerTime(execution.triggerTime)}
+              </span>
+              <span role="cell" className="automation-log-rule">
+                {execution.ruleName}
+              </span>
+              <span role="cell" className="automation-log-source">
+                {execution.sourceDisplayName}
+              </span>
+              <span role="cell">
+                <span className={`automation-status-chip ${execution.status}`}>
+                  {AUTOMATION_EXECUTION_STATUS_LABELS[execution.status]}
+                </span>
+              </span>
+              <span role="cell" className="automation-log-duration">
+                {formatDuration(execution.durationMs)}
+              </span>
+              <span role="cell">
+                <StepTrail execution={execution} />
+              </span>
+              <span role="cell" className="automation-log-action">
+                <Button variant="link" size="sm" onClick={() => onSelect(execution)}>
+                  查看详情
+                </Button>
+              </span>
+            </div>
+          ))}
+        </div>
+      )}
+    </div>
+  )
+}
@@ -0,0 +1,436 @@
+import * as React from 'react'
+import { Button, Input, RadioGroup, RadioGroupItem, Switch } from '../../components/ui'
+import {
+  LEAVE_NOTIFICATION_TARGET_OPTIONS,
+  describeLeaveNotificationNotifyScope,
+  normalizeLeaveNotificationConfig,
+  type AutomationRule,
+  type LeaveNotificationConfig,
+  type LeaveNotificationNotifyScope,
+  type LeaveNotificationTargetType
+} from '../../../../shared/automation'
+import { insertGroupNamePlaceholder } from '../../../../shared/group-exit-monitor'
+import { LeaveNotificationPreview } from './LeaveNotificationPreview'
+import type { AutomationGroupOption } from './model/api'
+import {
+  LeaveNotificationTargetPicker,
+  type LeaveNotificationContactOption
+} from './LeaveNotificationTargetPicker'
+import { LeaveNotificationTemplateEditor } from './LeaveNotificationTemplateEditor'
+
+/**
+ * LeaveNotificationEditor —— 「自动化 → 规则 → 退群通知」。
+ *
+ * 这是**真实配置页**：保存会落盘成一条 singleton 规则（`builtin-leave-notification`），
+ * 由 `AutomationService.handleGroupExit` 在退群事件到来时执行。
+ *
+ * 职责边界在 UI 上一眼可见，是**两层**而不是一层：
+ *   退群监控 = 监测哪些群有人退出（范围由它管）；
+ *   本规则   = ① 这批群里**哪些要通知**（二次勾选）；② 通知**发到哪**。
+ *
+ * ① 就是旧版「退群监控 → 通知群聊」的那份逐群勾选。它曾在自动化迁移里被压成单值目标而丢失，
+ * 导致"规则一启用就全量通知"；`notifyScope` 把它恢复回来，所以这里**有**勾选控件。
+ */
+
+export interface LeaveNotificationSaveInput {
+  enabled: boolean
+  config: LeaveNotificationConfig
+}
+
+export interface LeaveNotificationEditorProps {
+  /** 已持久化的退群通知规则（singleton，永远存在）。 */
+  rule: AutomationRule
+  /** 「指定好友」的可选项（main 侧已过滤，只含可发送的个人联系人）。 */
+  contacts: LeaveNotificationContactOption[]
+  /** 真实已监控群聊数量（来自退群监控，自动化不维护副本）。 */
+  monitoredCount: number
+  /** 真实已监控群聊清单 —— 二次勾选的候选项，与 `monitoredCount` 同一次取数。 */
+  monitoredGroups: AutomationGroupOption[]
+  saving: boolean
+  /** 发送能力不完整时的一句话提示；为 undefined 表示能力正常。 */
+  sendCapabilityWarning?: string
+  /** 「管理监控群聊 →」—— 跳退群监控的管理群聊页，纯导航。 */
+  onOpenMonitoredGroups?: () => void
+  onCancel: () => void
+  onSave: (input: LeaveNotificationSaveInput) => void
+}
+
+/**
+ * 草稿初值。
+ *
+ * 一律过一遍 `normalizeLeaveNotificationConfig`：它本来就负责"补齐缺失字段 +
+ * 清掉互相矛盾的配置"，所以契约新增字段（例如 `notifyScope`）时这里不会漏。
+ * 手拼字面量就会漏。
+ */
+function initialConfig(rule: AutomationRule): LeaveNotificationConfig {
+  return normalizeLeaveNotificationConfig(rule.leaveNotification)
+}
+
+export function LeaveNotificationEditor({
+  rule,
+  contacts,
+  monitoredCount,
+  monitoredGroups,
+  saving,
+  sendCapabilityWarning,
+  onOpenMonitoredGroups,
+  onCancel,
+  onSave
+}: LeaveNotificationEditorProps): React.ReactElement {
+  const [enabled, setEnabled] = React.useState(rule.enabled)
+  const [config, setConfig] = React.useState<LeaveNotificationConfig>(() => initialConfig(rule))
+  const [groupFilter, setGroupFilter] = React.useState('')
+
+  // 切换编辑对象 / 保存后回填时重建草稿，避免把上一份改动带过来。
+  React.useEffect(() => {
+    setEnabled(rule.enabled)
+    setConfig(initialConfig(rule))
+    setGroupFilter('')
+  }, [rule])
+
+  const targetType = config.target.type
+  const selectedContactId = String(config.target.contactId || '')
+
+  const selectedContactName = React.useMemo(
+    () => contacts.find((contact) => contact.id === selectedContactId)?.name ?? '',
+    [contacts, selectedContactId]
+  )
+
+  /** 选了「指定好友」但那个联系人已经不在了（被删 / 不可发送 / 找不到）。 */
+  const contactMissing =
+    targetType === 'contact' && Boolean(selectedContactId) && !selectedContactName
+
+  /**
+   * 通知不是发给「当前群聊」、但内容里又没有群名 → 收件人认不出是哪个群。
+   *
+   * 只在非「当前群聊」时要求：发给群内时，群名是冗余的。
+   */
+  const missingGroupName =
+    targetType !== 'source_chat' && !config.template.includes('{groupName}')
+
+  /** 通知范围（二次勾选）。`config.notifyScope` 是唯一权威，这里只是取短名。 */
+  const notifyScope: LeaveNotificationNotifyScope =
+    config.notifyScope === 'selected' ? 'selected' : 'all'
+
+  const selectedNotifyIds = React.useMemo(
+    () => new Set(config.notifyRoomIds),
+    [config.notifyRoomIds]
+  )
+
+  const visibleMonitoredGroups = React.useMemo(() => {
+    const keyword = groupFilter.trim().toLowerCase()
+    if (!keyword) return monitoredGroups
+    return monitoredGroups.filter((group) => group.name.toLowerCase().includes(keyword))
+  }, [monitoredGroups, groupFilter])
+
+  const selectTarget = (next: LeaveNotificationTargetType): void => {
+    setConfig((current) => ({
+      ...current,
+      // 换目标类型时清掉上一个类型才有的字段，避免存下互相矛盾的配置。
+      // 同时**清除迁移遗留的「待重选」标记** —— 用户已经做了选择。
+      target: next === 'contact' ? { type: next, contactId: current.target.contactId } : { type: next }
+    }))
+  }
+
+  const selectNotifyScope = (next: LeaveNotificationNotifyScope): void => {
+    setConfig((current) => ({ ...current, notifyScope: next }))
+  }
+
+  const toggleNotifyRoom = (roomId: string): void => {
+    setConfig((current) => {
+      const chosen = new Set(current.notifyRoomIds)
+      if (chosen.has(roomId)) chosen.delete(roomId)
+      else chosen.add(roomId)
+      // 按候选项顺序落盘，勾选顺序不影响存下来的结果。
+      const ordered = monitoredGroups
+        .map((group) => group.id)
+        .filter((id) => chosen.has(id))
+      // 候选项还没加载出来时保留原样，避免把用户的勾选静默清空。
+      return {
+        ...current,
+        notifyRoomIds: monitoredGroups.length ? ordered : current.notifyRoomIds
+      }
+    })
+  }
+
+  const handleSave = (): void => {
+    onSave({
+      enabled,
+      config: {
+        ...config,
+        // 勾选集收敛到「当前已监控」这一集合内：否则监控范围缩小之后，
+        // 会留下一批界面上看不见、且永远不可能命中的幽灵勾选。
+        notifyRoomIds: monitoredGroups.length
+          ? monitoredGroups.map((group) => group.id).filter((id) => selectedNotifyIds.has(id))
+          : config.notifyRoomIds,
+        targetNeedsReview: undefined
+      }
+    })
+  }
+
+  return (
+    <div className="automation-editor">
+      <header className="automation-editor-header">
+        <div>
+          <h2>退群通知</h2>
+          <p>当已监控群聊检测到成员退出时，自动发送通知。</p>
+          <div className="automation-leave-status-row">
+            <span className="automation-field-label">启用这条自动化</span>
+            <Switch
+              checked={enabled}
+              onCheckedChange={setEnabled}
+              aria-label="启用这条自动化"
+            />
+          </div>
+        </div>
+        <div className="automation-editor-actions">
+          <Button variant="ghost" onClick={onCancel} disabled={saving}>
+            取消
+          </Button>
+          <Button onClick={handleSave} disabled={saving}>
+            {saving ? '保存中…' : '保存'}
+          </Button>
+        </div>
+      </header>
+
+      {config.targetNeedsReview ? (
+        <p className="automation-notice warning">
+          旧版「通知群聊」是逐群配置的，无法无损转换成新的单一通知目标。已保留你的通知模板，
+          但<strong>在你重新选择通知目标之前不会发送任何通知</strong>。
+        </p>
+      ) : null}
+      {contactMissing ? (
+        <p className="automation-notice warning">
+          之前选择的联系人已不存在或当前无法发送，请重新选择通知目标。
+        </p>
+      ) : null}
+      {sendCapabilityWarning ? (
+        <p className="automation-notice warning">{sendCapabilityWarning}</p>
+      ) : null}
+
+      <div className="automation-editor-body">
+        <div className="automation-editor-form">
+          <section className="automation-section">
+            <div className="automation-section-heading">
+              <h3>1 · 什么时候触发</h3>
+            </div>
+
+            <div className="automation-inline-row">
+              <div>
+                <span className="automation-field-label">触发事件</span>
+                <span className="automation-static-value">检测到群成员退出</span>
+                <small>这是「退群监控」提供的事件，由本规则响应。</small>
+              </div>
+            </div>
+
+            <div className="automation-inline-row">
+              <div>
+                <span className="automation-field-label">监控来源</span>
+                <span className="automation-static-value">退群监控</span>
+              </div>
+            </div>
+
+            {/*
+              第一层：只读的监控范围摘要 + 唯一入口。
+              「哪些群被监控」完全由退群监控决定，本规则不重复维护一份。
+            */}
+            <div className="automation-leave-groups-card">
+              <div>
+                <span className="automation-field-label">已监控群聊</span>
+                <strong>{monitoredCount} 个群聊</strong>
+                <small>监控范围由「退群监控」管理</small>
+              </div>
+              <Button
+                variant="outline"
+                size="sm"
+                onClick={onOpenMonitoredGroups}
+                disabled={!onOpenMonitoredGroups}
+              >
+                管理监控群聊 →
+              </Button>
+            </div>
+
+            {/*
+              第二层：在上面这批已监控群聊里**二次勾选**哪些要通知。
+              这就是旧版「退群监控 → 通知群聊」的那份逐群勾选 ——
+              没有它，规则一启用就等于给全部已监控群聊发通知。
+            */}
+            <div className="automation-inline-row">
+              <div className="automation-leave-scope">
+                <span className="automation-field-label">通知范围</span>
+                <RadioGroup
+                  value={notifyScope}
+                  onValueChange={(value) =>
+                    selectNotifyScope(value as LeaveNotificationNotifyScope)
+                  }
+                  aria-label="通知范围"
+                  className="automation-leave-scope-options"
+                >
+                  <div className="automation-leave-radio option">
+                    <RadioGroupItem value="all" id="leave-scope-all" />
+                    <div className="automation-leave-radio-body">
+                      <label htmlFor="leave-scope-all">全部已监控群聊</label>
+                      <small>任一被监控的群有人退出都发通知。</small>
+                    </div>
+                  </div>
+                  <div className="automation-leave-radio option">
+                    <RadioGroupItem
+                      value="selected"
+                      id="leave-scope-selected"
+                      disabled={!monitoredGroups.length}
+                    />
+                    <div className="automation-leave-radio-body">
+                      <label htmlFor="leave-scope-selected">仅选中的群聊</label>
+                      <small>只对下面勾选的群发通知。</small>
+                    </div>
+                  </div>
+                </RadioGroup>
+                <small className="automation-leave-scope-summary">
+                  当前覆盖：{describeLeaveNotificationNotifyScope(config, monitoredCount)}
+                </small>
+              </div>
+            </div>
+
+            {notifyScope === 'selected' ? (
+              monitoredGroups.length ? (
+                <div className="automation-field">
+                  <div className="automation-section-heading">
+                    <span className="automation-field-label">勾选要通知的群聊</span>
+                    <span className="automation-section-note">
+                      {selectedNotifyIds.size ? `已选 ${selectedNotifyIds.size} 个` : '尚未勾选'}
+                    </span>
+                  </div>
+                  <Input
+                    value={groupFilter}
+                    onChange={(event) => setGroupFilter(event.target.value)}
+                    placeholder="搜索群聊"
+                    aria-label="搜索群聊"
+                  />
+                  <div className="automation-group-list" role="group" aria-label="通知群聊">
+                    {visibleMonitoredGroups.length === 0 ? (
+                      <p className="automation-group-empty">没有匹配「{groupFilter}」的群聊</p>
+                    ) : (
+                      visibleMonitoredGroups.map((group) => {
+                        const checked = selectedNotifyIds.has(group.id)
+                        return (
+                          <label
+                            key={group.id}
+                            className={`automation-group-option ${checked ? 'selected' : ''}`}
+                          >
+                            <input
+                              type="checkbox"
+                              checked={checked}
+                              onChange={() => toggleNotifyRoom(group.id)}
+                            />
+                            <span>{group.name}</span>
+                          </label>
+                        )
+                      })
+                    )}
+                  </div>
+                  {selectedNotifyIds.size === 0 ? (
+                    <p className="automation-section-hint">
+                      <span aria-hidden="true">ⓘ</span>
+                      一个群都没勾选，这条规则不会发送任何通知。
+                    </p>
+                  ) : null}
+                </div>
+              ) : (
+                <p className="automation-section-hint">
+                  <span aria-hidden="true">ⓘ</span>
+                  还没有设置监控范围。请先到「退群监控 → 管理群聊」选择要监控的群，
+                  再回来勾选通知范围。
+                </p>
+              )
+            ) : null}
+          </section>
+
+          <section className="automation-section">
+            <div className="automation-section-heading">
+              <h3>2 · 通知发送到哪里</h3>
+            </div>
+
+            <RadioGroup
+              value={targetType}
+              onValueChange={(value) => selectTarget(value as LeaveNotificationTargetType)}
+              aria-label="通知发送到哪里"
+              className="automation-leave-targets"
+            >
+              {LEAVE_NOTIFICATION_TARGET_OPTIONS.map((option) => (
+                <div key={option.type} className="automation-leave-radio option">
+                  <RadioGroupItem value={option.type} id={`leave-target-${option.type}`} />
+                  <div className="automation-leave-radio-body">
+                    <label htmlFor={`leave-target-${option.type}`}>{option.label}</label>
+                    <small>{option.description}</small>
+                  </div>
+                </div>
+              ))}
+            </RadioGroup>
+
+            {targetType === 'contact' ? (
+              <div className="automation-field">
+                <span className="automation-field-label">选择好友</span>
+                <LeaveNotificationTargetPicker
+                  contacts={contacts}
+                  selectedId={selectedContactId}
+                  onSelect={(id) =>
+                    setConfig((current) => ({
+                      ...current,
+                      target: { type: 'contact', contactId: id },
+                      targetNeedsReview: undefined
+                    }))
+                  }
+                  ariaLabel="选择好友"
+                  searchPlaceholder="搜索好友"
+                  emptyText="没有可发送的联系人"
+                />
+              </div>
+            ) : null}
+          </section>
+
+          <section className="automation-section">
+            <div className="automation-section-heading">
+              <h3>3 · 通知内容</h3>
+            </div>
+
+            {/*
+              护栏：通知发给「别人」时，不带群名的通知等于「张三退群了」。
+              主进程已替用户补过一次（一次性迁移），这里兜的是"用户后来主动删掉了"的情况 ——
+              只提示 + 一键补上，不强行改用户文本。
+            */}
+            {missingGroupName ? (
+              <div className="automation-notice warning automation-notice-stack">
+                <p>这份内容里没有群聊名，收件人看不出是哪个群退的人。</p>
+                <Button
+                  variant="outline"
+                  size="sm"
+                  onClick={() =>
+                    setConfig((current) => ({
+                      ...current,
+                      template: insertGroupNamePlaceholder(current.template)
+                    }))
+                  }
+                >
+                  补上群聊名
+                </Button>
+              </div>
+            ) : null}
+
+            <LeaveNotificationTemplateEditor
+              template={config.template}
+              onChange={(next) =>
+                setConfig((current) => ({ ...current, template: next }))
+              }
+            />
+          </section>
+        </div>
+
+        <LeaveNotificationPreview
+          config={config}
+          contactName={selectedContactName}
+          monitoredCount={monitoredCount}
+        />
+      </div>
+    </div>
+  )
+}
@@ -0,0 +1,114 @@
+import * as React from 'react'
+import {
+  describeLeaveNotificationNotifyScope,
+  describeMonitoredScope,
+  leaveNotificationTargetLabel,
+  type LeaveNotificationConfig
+} from '../../../../shared/automation'
+import {
+  createGroupExitPreviewSample,
+  renderLeaveNotificationText
+} from '../../../../shared/group-exit-event'
+
+/**
+ * LeaveNotificationPreview —— 右侧「效果预览」。
+ *
+ * 与左侧表单逐项联动：改模板 → 气泡变；改发送目标 → 顶部收件会话与底部「通知目标」一起变。
+ *
+ * **绝不发送**：即使规则已经真实接通，这里也只做渲染 ——
+ * 它拿的是 `GROUP_EXIT_PREVIEW_SAMPLE` 样本事件，不碰 `WechatActionGateway`。
+ * 模板渲染复用 shared 的唯一实现，所以预览与实发不会出现两套结果。
+ */
+
+export interface LeaveNotificationPreviewProps {
+  config: LeaveNotificationConfig
+  /** 「指定好友」已选联系人的显示名；未选 / 不是该目标时传空串。 */
+  contactName: string
+  /** 真实已监控群聊数量（来自退群监控）。 */
+  monitoredCount: number
+}
+
+export function LeaveNotificationPreview({
+  config,
+  contactName,
+  monitoredCount
+}: LeaveNotificationPreviewProps): React.ReactElement {
+  // 样本事件只服务预览；时间取"此刻"，不伪造历史日期。
+  const sample = React.useMemo(() => createGroupExitPreviewSample(Date.now()), [])
+
+  const chatTitle = React.useMemo((): string => {
+    switch (config.target.type) {
+      case 'source_chat':
+        return sample.groupName || '当前群聊'
+      case 'self':
+        return '我'
+      case 'file_transfer':
+        return leaveNotificationTargetLabel('file_transfer')
+      case 'contact':
+        return contactName.trim() || '未选择好友'
+    }
+  }, [config.target.type, contactName, sample.groupName])
+
+  const targetSummary = React.useMemo((): string => {
+    switch (config.target.type) {
+      case 'source_chat':
+        return '发生退群事件的群聊'
+      case 'self':
+        return '我'
+      case 'file_transfer':
+        return leaveNotificationTargetLabel('file_transfer')
+      case 'contact':
+        return contactName.trim() || '未选择好友'
+    }
+  }, [config.target.type, contactName])
+
+  const bubbleText = renderLeaveNotificationText(sample, config.template).trim() || '（模板为空）'
+
+  return (
+    <section className="automation-preview" aria-label="效果模拟预览">
+      <div className="automation-preview-heading">
+        <h3>效果预览</h3>
+        <span className="automation-preview-badge">仅预览，不会真实发送</span>
+      </div>
+
+      <div className="automation-preview-phone">
+        <div className="automation-preview-chat-title" data-testid="leave-preview-recipient">
+          <span>{chatTitle}</span>
+        </div>
+        <div className="automation-preview-chat">
+          {/* 退群监控产生的系统事件，不来自任何消息文本 */}
+          <p className="automation-preview-system">检测到成员退出</p>
+
+          <div className="automation-chat-row outgoing">
+            <div className="automation-chat-body">
+              <span className="automation-chat-name">我</span>
+              <span className="automation-chat-bubble pre">{bubbleText}</span>
+            </div>
+            <span className="automation-chat-avatar self" aria-hidden="true">
+              我
+            </span>
+          </div>
+        </div>
+      </div>
+
+      <dl className="automation-preview-facts">
+        <div>
+          <dt>触发事件</dt>
+          <dd>成员退出</dd>
+        </div>
+        <div>
+          <dt>监控来源</dt>
+          <dd>{describeMonitoredScope(monitoredCount)}</dd>
+        </div>
+        <div>
+          <dt>通知范围</dt>
+          <dd>{describeLeaveNotificationNotifyScope(config, monitoredCount)}</dd>
+        </div>
+        <div>
+          <dt>通知目标</dt>
+          <dd>{targetSummary}</dd>
+        </div>
+      </dl>
+    </section>
+  )
+}
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
电摇小子	713232bd62	feat: 新增 API（capabilities / 退群监控 / 自动化 / 群统计）	2026-10-04 02:31:34 +07:00
电摇小子	e83e85fe64	feat: 持久化聊天记录导出设置	2026-10-02 19:37:09 +07:00
电摇小子	e5a1656644	style: 修复图标样式 #46 Fixes https://github.com/Wxw-Gu/TraceMemo/issues/46	2026-10-01 17:35:47 +07:00
电摇小子	194dd0035b	chore: 构建	2026-09-29 18:02:09 +07:00
电摇小子	c3629255ea	chore: 更新二维码	2026-09-29 16:24:20 +07:00
电摇小子	345a0db49b	chore: 提升版本整理文档整理档案首页按钮	2026-09-29 16:13:37 +07:00
电摇小子	5276bce060	chore: 打包配置测试用例	2026-09-29 13:06:00 +07:00
电摇小子	0abbda1160	Merge branch 'develop_0920' into develop	2026-09-29 10:51:15 +07:00
电摇小子	97c74859c0	fix: 修复联系人 API 因昵称备注未补全导致搜索结果为空 #51 Fixes https://github.com/Wxw-Gu/TraceMemo/issues/51	2026-09-29 10:47:33 +07:00
电摇小子	4d5c6459fe	feat: 定时日报并入自动化，统一规则管理与发送链路发送目标与退群通知统一为来源群 / 自己 / 文件传输助手 / 指定好友日报页补「定时日报已并入自动化」指引条	2026-09-28 17:11:50 +07:00
Wxw-Gu	c835d5c570	feat: 定时日报归到定制化功能下	2026-09-25 17:44:45 +08:00
Wxw-Gu	59872daf52	feat: 退群监控接入自动化能力统一管理, 支持发送个人/文件传输助手/指定群聊	2026-09-24 16:31:07 +08:00
Wxw-Gu	3bd3935042	fix: 修复mac发送初始化卡在'正在检查发送运行时…'	2026-09-24 10:49:34 +08:00
Wxw-Gu	77a6912ade	fix: 修复退群监控列表错误, 界面左侧滚动条	2026-09-24 10:36:10 +08:00
Wxw-Gu	4b83aaed61	Merge branch 'develop' into develop_0920 # Conflicts: # src/main/index.ts # src/preload/index.d.ts # src/preload/index.ts # src/renderer/src/styles/index.scss	2026-09-24 09:41:53 +08:00
Wxw-Gu	7268188e40	feat: mac发送能力重构为libtmwechat - 发送改走本地host, 在微信登录时触发 - 删除旧OneBot及对应UI控制面	2026-09-24 03:01:03 +08:00
Wxw-Gu	069d836dea	feat: 精确会话自动化规则	2026-09-21 11:06:59 +08:00
Wxw-Gu	95384a4a92	feat: 自动化	2026-09-21 00:35:32 +08:00
Wxw-Gu	9b82ca663b	fix: 修复Windows语音消息取错	2026-09-21 00:33:02 +08:00
Wxw-Gu	f5633a29cb	feat: 新增群员统计与退群事件，并把本地索引集中到设置页 - 群员统计：档案中按群统计发言排行与未发言成员，可复制纯文本、按平台能力发送 - 退群事件永久保存，并以本地推断条目并入档案消息流 - 设置新增本地索引页，集中管理聊天记录索引与图片文字索引	2026-09-20 21:47:06 +08:00
Wxw-Gu	766cd94788	feat: 新增群员统计与退群事件，并把本地索引集中到设置页 - 群员统计：档案中按群统计发言排行与未发言成员，可复制纯文本、按平台能力发送 - 退群事件永久保存，并以本地推断条目并入档案消息流 - 设置新增本地索引页，集中管理聊天记录索引与图片文字索引	2026-09-20 21:46:37 +08:00
wuyouMaster	b6d287a299	fix: make packaging checks platform safe	2026-09-20 18:16:41 +08:00
wuyouMaster	4c25236070	fix: ad-hoc sign packaged macOS key helpers and app bundle electron-builder 26 skips macOS signing entirely when no Developer ID identity is configured, so the packaged key helpers can ship unsigned or with a broken signature and macOS kills them even with SIP disabled. afterPack now verifies and ad-hoc re-signs the packaged xkey helpers and the outer app bundle, failing the build when a signature cannot be repaired.	2026-09-20 18:16:41 +08:00