Commit Graph
8 Commits
Author SHA1 Message Date
dev 881b5b4c67 fix: repair photo worker broken code; add pending queue view
- Fix runAiEnhance missing return/closing brace from previous commit
- Remove duplicate code fragments in worker.js
- Add renderPhotoPending() to show pending/processing jobs in worker dashboard
- New pending queue table in worker.html photo section
2026-09-17 17:52:51 +03:00
dev 1145be02f3 feat: persistent photo job queue, worker, history with rollback
- Replace in-memory photoAiJobs Map with DB-backed photo_jobs table
- Background photo worker (worker.js) with retry, backoff, stale reset
  - Handles both AI (Real-ESRGAN) and server-side (sharp) enhancement
  - Controlled via photo_worker_enabled setting
- Photo job history in enhance modal with before/after thumbnails + rollback
- Worker dashboard: photo jobs section with status counts, recent table,
  compare slider for before/after, rollback from worker UI
- New endpoints: /api/photo-jobs/status|wake|enabled|requeue-failed,
  /api/entries/:id/photo/jobs (history), .../rollback
- swapEntryPhotoFiles logs every mutation to photo_jobs table
2026-09-17 17:36:56 +03:00
dev 049df61e55 feat: photo quality improvements
- webcam capture resolution/quality configurable in admin settings (defaults 640x480 / 0.92)
- enhance photo modal in journal: original vs preview with sliders (brightness, contrast, saturation, sharpen) and auto-levels button
- new endpoint PUT /api/entries/:id/photo/enhance replaces photo, cleans old file and thumb
- sharper HEIC conversion (0.92) and webp thumbnails (85)
- worker: fail explicitly on empty AI response
2026-09-17 15:07:43 +03:00
dev 57e48a9d4f fix(ai): adaptive max_tokens per backend (reasoning profiles get 1024+, native llama.cpp stays compact; per-profile max_tokens override); include entry texts in ai_status SSE notify so journal updates live without reload; audit log action filter; links pagination; settings validation & UI improvements 2026-09-17 14:12:36 +03:00
dev efe2a42d1d feat(settings): AI model profiles management — native llama.cpp + custom OpenAI-compatible providers, default model selection, connection test 2026-09-17 11:54:08 +03:00
dev f8036fae79 Fix: backup/restore now includes group_photos and entry_photos tables; increase AI request timeout to 120s (configurable via AI_REQUEST_TIMEOUT_MS) 2026-09-11 23:12:16 +03:00
dev 275ed46dfb Add HF model auto-download with configurable AI env, and share link visibility toggles 2026-09-11 13:12:13 +03:00
dev b7e798b867 Add AI auto-check worker, share link message/link fields, and update journal/settings UI 2026-09-11 10:29:49 +03:00