- Export layout changed from clip_NNN group dirs to vid_NNN per-video folders
- Automatic DB migration rewrites old paths and moves files on startup
- Per-video counter with DB cross-check to prevent overwrites
- Changelog popup on version bump with "don't show again" checkbox
- Scan region resize now requires Shift+drag to prevent accidental edits
- Recalculate vid folder and counter on file load
- Add EAT_LARGE embedding model variant
- Update tests for new flat export path structure
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Covers scan result versioning per model, hard negative management
dialog with training toggle, and ghost folder fix.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Calibration: cv=min(3,n_pos,n_neg_sample) could yield cv=1 (ValueError);
replaced with min_class >= 6 guard to skip calibration for tiny datasets
- AST: clarified chunks are already numpy arrays, use list(chunks) directly
- EAT: noted extract_features returns plain tensor (not tuple)
- Multi-layer: explicit notes on _w2v_model_name storing base name,
ml_cfg needed in _extract_w2v_targeted, embeddings_list vs embeddings
- Added AST to _ml_config layer_counts upfront in Task 2
- Added integration test for model switching (no-reload verification)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Swap Task 5/6 order so get_all_export_paths exists before UI uses it
- Remove cosine similarity clamping to preserve anti-correlation signal
- Use os.path.exists instead of os.path.isfile (handles image sequences)
- Add worker cleanup to disconnect stale signals before new scan
- Remove lock from get_all_export_paths (matches read-only convention)
- Always use get_all_export_paths for Current Profile (not current-file-first)
- Filter export paths with os.path.exists for deleted files
- Use abs() for float comparison in tests instead of ==
- Add cancel_flag to ScanWorker and scan_video for interruptible scans
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
15-task plan covering Rust install, Tauri scaffold, mpv sidecar,
API client, stores, UI components, keyboard shortcuts, and packaging.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix line range 38→36 for _frozen_path extraction
- Clarify line ranges for ffmpeg vs annotation functions
- Remove unused imports (_frozen_path, build_annotation_json_path) from main.py import list
- Add step to clean up dead stdlib imports (re, json, sqlite3, tempfile, datetime)
- Add explicit stub router code for stream, markers, export, hidden
- Add server/__init__.py and server/routes/__init__.py content
- Add _DBWorker and FrameGrabber to "keep in main.py" list
- Note optional tracking deps in Dockerfile
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
mpv embedding:
- Replace wid/QOpenGLWidget with QOffscreenSurface + QOpenGLFramebufferObject
+ QPainter readback — works on Wayland/KDE without sub-surface compositing
- Force desktop OpenGL 3.3 core profile before QApplication (fixes black output on GLES)
- Timer-based render polling (16 ms) replaces signal-flood from mpv C thread;
fixes playback animation and scrubbing preview
- Fix AB-loop: set ab-loop-a/b to "no" on stop (0 means infinite in mpv)
Timeline:
- Full redesign: time ruler with adaptive major/minor ticks, playhead triangle
handle, selection region with edge lines, numbered marker badges
- Height 160 px; layout collapsed from 4 rows to 2 below timeline
- Markers appear immediately on export (optimistic update before ffmpeg finishes)
- Right-click marker → context menu to delete from DB
Hotkeys:
- Replace keyPressEvent with QShortcut(ApplicationShortcut) so keys work
regardless of focused widget; MpvWidget gets NoFocus policy
Export:
- WebP: switch lossless→lossy quality 85, compression_level 1 (~10x faster)
- Add -threads 0 for full CPU utilisation during decode/filter
- Remember last export folder across sessions via QSettings
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Ethanfel