feat: 完善群聊日报语音转写缓存

支持空内容的微信语音消息并兼容历史缓存转写结果
将系统通知标记为微信系统消息，排除活跃成员与发言排行

tests/unit/group-report.test.ts

@@ -1,6 +1,11 @@
 import { describe, expect, it } from 'vitest'
-import { parseGroupDailyReport } from '../../src/renderer/src/utils/group-report'
+import {
+  buildGroupReportInput,
+  parseGroupDailyReport
+} from '../../src/renderer/src/utils/group-report'
 import { summaryContent } from '../../src/renderer/src/utils/group-report-facts'
+import { summarySender } from '../../src/renderer/src/utils/group-report-facts'
+import { selectHeroParticipantNames } from '../../src/shared/group-report'
 import type { GroupReportMetadata } from '../../src/shared/group-report'
 import type { Message } from '../../src/shared/types'
 
@@ -25,6 +30,12 @@ const media = {
 }
 
 describe('group report parsing', () => {
+  it('keeps only distinct real participants in the hero avatar list', () => {
+    expect(
+      selectHeroParticipantNames(['濑岛田井卫', '测试群昵称', '濑岛田井卫', '', '  '])
+    ).toEqual(['濑岛田井卫', '测试群昵称'])
+  })
+
   it('includes a cached voice transcript in the report input content', () => {
     const message: Message = {
       id: 'voice-1',
@@ -40,6 +51,68 @@ describe('group report parsing', () => {
     expect(summaryContent(message)).toContain('今晚八点确认发布。')
   })
 
+  it('includes a voice transcript when legacy cached messages have no contentData', () => {
+    const message: Message = {
+      id: 'voice-legacy',
+      from: 'member',
+      type: '语音',
+      datetime: '2026-08-11 10:00:00',
+      content: '[语音消息]',
+      isSender: false,
+      voiceTranscript: '能不能听见这个语音？'
+    }
+
+    expect(summaryContent(message)).toBe('[语音] 能不能听见这个语音？')
+  })
+
+  it('passes legacy voice transcripts into the daily-report model prompt', async () => {
+    const message: Message = {
+      id: 'voice-prompt',
+      from: 'member',
+      type: '语音',
+      datetime: '2026-08-11 10:00:00',
+      content: '[语音消息]',
+      isSender: false,
+      name: '测试成员',
+      voiceTranscript: '试一下好不好使？'
+    }
+
+    const input = await buildGroupReportInput([message], null, true, 'full')
+
+    expect(input.prompt).toContain('[语音] 试一下好不好使？')
+  })
+
+  it('labels system notices separately and excludes them from active members', async () => {
+    const systemMessage: Message = {
+      id: 'system-1',
+      from: 'system',
+      type: '系统消息',
+      datetime: '2026-08-11 09:59:00',
+      content: '由于账号安全原因，无法加入当前群聊。',
+      isSender: false,
+      contentData: {
+        type: 'system',
+        content: '由于账号安全原因，无法加入当前群聊。'
+      }
+    }
+    const memberMessage: Message = {
+      id: 'member-1',
+      from: 'member',
+      type: '普通文本',
+      datetime: '2026-08-11 10:00:00',
+      content: '收到',
+      name: '测试成员',
+      isSender: false
+    }
+
+    expect(summarySender(systemMessage, null, true)).toBe('微信系统消息')
+    const input = await buildGroupReportInput([systemMessage, memberMessage], null, true, 'full')
+
+    expect(input.metadata.activeUsers).toBe(1)
+    expect(input.topSpeakers).toEqual([{ name: '测试成员', count: 1 }])
+    expect(input.prompt).toContain('微信系统消息：由于账号安全原因，无法加入当前群聊。')
+  })
+
   it('falls back to topic keywords when the model omits top-level keywords', () => {
     const report = parseGroupDailyReport(
       JSON.stringify({

tests/unit/message-parser.test.ts

@@ -8,6 +8,7 @@ describe('message parser', () => {
       md5: '0123456789abcdef0123456789abcdef'
     })
     expect(parseMessageContent('voice fixture', 34)).toEqual({ type: 'voice' })
+    expect(parseMessageContent('', 34)).toEqual({ type: 'voice' })
     expect(
       parseMessageContent(
         '<emoji md5="abcdefabcdefabcdefabcdefabcdefab" cdnurl="https://fixture.invalid/a" />',

tests/unit/voice-message-reference.test.ts

@@ -74,6 +74,20 @@ describe('daily report voice transcription', () => {
     })
   })
 
+  it('normalizes legacy voice messages without contentData for report facts', async () => {
+    const legacy = { ...voice('legacy', 4), contentData: undefined }
+    const result = await transcribeVoiceMessages([legacy], {
+      getModelStatus: vi.fn(async () => status('ready')),
+      recognize: vi.fn(async () => ({ success: true, transcript: '日报语音内容' })),
+      onProgress: vi.fn()
+    })
+
+    expect(result[0]).toMatchObject({
+      voiceTranscript: '日报语音内容',
+      contentData: { type: 'voice' }
+    })
+  })
+
   it('does not require the model when every transcript is cached', async () => {
     const getModelStatus = vi.fn(async () => status('missing'))
     const recognize = vi.fn()
@@ -88,6 +102,26 @@ describe('daily report voice transcription', () => {
     expect(recognize).not.toHaveBeenCalled()
   })
 
+  it('hydrates persisted transcript cache before invoking recognition', async () => {
+    const getModelStatus = vi.fn(async () => status('missing'))
+    const getCachedTranscript = vi.fn(async () => ({
+      state: 'transcribed' as const,
+      transcript: '持久化缓存内容'
+    }))
+    const recognize = vi.fn()
+    const result = await transcribeVoiceMessages([voice('persisted', 8)], {
+      getModelStatus,
+      getCachedTranscript,
+      recognize,
+      onProgress: vi.fn()
+    })
+
+    expect(result[0].voiceTranscript).toBe('持久化缓存内容')
+    expect(getCachedTranscript).toHaveBeenCalledOnce()
+    expect(getModelStatus).not.toHaveBeenCalled()
+    expect(recognize).not.toHaveBeenCalled()
+  })
+
   it('stops before recognition when the local model is not ready', async () => {
     const recognize = vi.fn()
     await expect(