liyachen 95a77bec4d feat: AI frame selection via a vision model (GLM-4.6V-Flash / Z.ai) ... When a frame-selector API key is set, hook/keyframe send the extension's deduped candidate frames and a target count; the plugin asks the vision model to keep the most effect-heavy frames (skip talking-head shots) and stores only those. Robust index parsing + heuristic fallback (even-sample) when unconfigured or on any failure. New "AI 挑帧" settings (Z.ai base URL + GLM-4.6V-Flash prefilled; user supplies the key). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>		2026-06-28 16:23:16 -04:00
..
__mocks__	feat: scaffold vault-autopilot plugin	2026-05-30 14:14:28 -04:00
providers	feat: add analyzeMultiFrame to Gemini API provider	2026-05-30 22:08:16 -04:00
clip-router.test.ts	feat: every video note gets a gallery cover at <video_id>.webp	2026-06-28 15:25:09 -04:00
frame-select.test.ts	feat: AI frame selection via a vision model (GLM-4.6V-Flash / Z.ai)	2026-06-28 16:23:16 -04:00
path-detector.test.ts	feat: CLI binary path auto-detector for Mac/Windows/Linux	2026-05-30 14:34:50 -04:00
processor.test.ts	feat: file processor — read file, validate SOP, call provider	2026-05-30 14:37:46 -04:00
server.test.ts	feat: merge cover/hook/keyframe into one video note via video-note	2026-06-28 10:34:49 -04:00
util.test.ts	feat(embed): buildVideoEmbed accepts an optional end time	2026-06-28 10:24:24 -04:00
video-note.test.ts	feat: fold screenshots into the matching video note	2026-06-28 13:40:44 -04:00