Repository navigation
KnowGrove 2.8.35
自动逐字稿可读性
- 录音、本地音视频导入、独立媒体链接、浏览器/手机收集、原稿续跑和新采集的网页内嵌媒体,统一先保存原稿,再补标点与自然段,无需单独开启。
- 严格保留原词、重复、原有标点、数字和英文术语;代码、录音标记、时间轴和文章原文不被改写。原始转录保留校验备份。
- 旧稿可运行“修复当前逐字稿标点”,有原稿备份、取消和并发编辑保护;已可读的逐字稿不会重复调用 AI。
- 任务界面明确显示标点处理阶段和部分完成状态,进度不回退;验证失败时保留原话,不静默接受改写。
需选定的 AI 服务正常可用。本次不会自动切换默认服务,也不会修正 ASR 识别错字。旧网页中没有受管范围的内嵌媒体需重新采集,不猜测修改文章内容。
Automatic transcript readability
All new media routes share punctuation and paragraph restoration after a safe raw backup. Original words, technical tokens and protected metadata are verified; existing transcripts have a separate safe repair command. Ordinary article prose is unchanged. A working selected provider is required; this is not ASR word correction.
Verification
Public source: 413 tests, lint/type/review checks, production build, independent exact-head review, PR/main/release CI, and native empty-vault enable/reload/command verification. All three released assets match the tested candidate and have verified GitHub build provenance for the 2.8.35 tag. No dependency, workflow, budget, provider-default or user-data export changes.