mirror of
https://wget.la/https://github.com/Wxw-Gu/WechatExplorer
synced 2026-10-10 01:42:49 +08:00
feat: 图片文字索引按时间分段优先处理最近图片
- 首次索引先处理最近 7 天,再依次回溯 30 天 / 近一年 / 更早历史 - 完成后新到的图片单独补齐,不受历史回填影响 - 覆盖度增加时间维度,可区分「最近已完整」与「更早仍在补齐」
This commit is contained in:
@@ -29,7 +29,9 @@ import {
|
||||
normalizeMessageIdentity
|
||||
} from '../../shared/local-query-api'
|
||||
import {
|
||||
IMAGE_TEXT_BACKFILL_TIER_LABEL,
|
||||
describeImageTextCoverage,
|
||||
describeImageTextCoveredRanges,
|
||||
imageTextCoverageState,
|
||||
type ImageTextIndexCoverage
|
||||
} from '../../shared/image-text-index'
|
||||
@@ -144,6 +146,17 @@ export function buildImageOcrCoverage(
|
||||
? `(图片数量统计于 ${formatLocalMinute(coverage.countedAt)})`
|
||||
: ''
|
||||
const base = describeImageTextCoverage(coverage)
|
||||
/**
|
||||
* recent-first 之后必须把"哪段时间能下确定性结论"一起给出。
|
||||
*
|
||||
* 否则模型只看一个总百分比:30% 时它会以为连最近一周都不可信(过度保守没坏处),
|
||||
* 但 99% 时它会以为"去年也能放心下结论"(这就把索引缺口说成了事实空缺)。
|
||||
*/
|
||||
const tiers = coverage.tiers ?? []
|
||||
const coveredRanges = tiers
|
||||
.filter((entry) => entry.state === 'complete')
|
||||
.map((entry) => IMAGE_TEXT_BACKFILL_TIER_LABEL[entry.tier])
|
||||
const rangeNote = tiers.length ? describeImageTextCoveredRanges(coverage) : ''
|
||||
return {
|
||||
state,
|
||||
totalImageMessages: coverage.totalImageMessages,
|
||||
@@ -154,10 +167,11 @@ export function buildImageOcrCoverage(
|
||||
failed: coverage.failed,
|
||||
pending: coverage.pending,
|
||||
...(coverage.countedAt ? { countedAtLabel: formatLocalMinute(coverage.countedAt) } : {}),
|
||||
...(coveredRanges.length ? { coveredRanges } : {}),
|
||||
summary:
|
||||
state === 'complete'
|
||||
? `${base}${countedNote}`
|
||||
: `${base}${countedNote}${IMAGE_OCR_ZERO_RESULT_CAUTION}`
|
||||
: `${base}${countedNote}${rangeNote}${IMAGE_OCR_ZERO_RESULT_CAUTION}`
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user