Instead of chmod on every app start, catch EACCES (error 13) when
spawning sidecar or ffmpeg, fix permissions, then retry once:
- sidecar spawn: catches permission denied, runs set_executable_permissions
on the sidecar dir, retries spawn
- ffmpeg: catches permission denied, chmod +x ffmpeg and ffprobe, retries
Zero overhead on normal launches. Only fixes permissions when actually needed.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The chmod fix only ran after fresh extraction, but existing sidecar
dirs extracted by older versions still lacked execute permissions.
Now set_executable_permissions() runs on EVERY app launch (both the
early-return path for existing dirs and after fresh extraction).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Previously only the main sidecar binary got chmod 755. Now all files
in the extraction directory get execute permissions — covers ffmpeg,
ffprobe, and any other bundled binaries. Applied in three places:
- sidecar/mod.rs: after local extraction
- commands/sidecar.rs: after download extraction
- commands/media.rs: removed single-file fix (now handled globally)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The bundled ffmpeg in the sidecar extract dir lacked execute permissions.
Now sets chmod 755 on Unix when find_ffmpeg locates the bundled binary.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Changed from folder picker (can only select existing folders) to save
dialog where the user can type a new name. The typed name becomes the
project folder, created automatically if it doesn't exist. Any file
extension the user types is stripped (e.g. "My Project.vtn" becomes
the folder "My Project/").
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Save behavior:
- Save on v2 project: saves in place (no dialog)
- Save on v1 project: auto-migrates to folder structure next to the
original .vtn (creates ProjectName/ folder with .vtn + audio.wav)
- Save on unsaved project: opens folder picker (Save As)
- Save As: always opens folder picker for a new location
Added projectIsV2 state to track project format version.
Split "Save Project" button into "Save" + "Save As".
Extracted saveToFolder() helper for shared save logic.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Changed save dialog from file picker (.vtn) to folder picker. The
project name is derived from the folder name. Files are created
inside the chosen folder:
Folder/
Folder.vtn
audio.wav
Also: save-in-place for already-saved projects (Ctrl+S just saves,
no dialog). Extracted buildProjectData() helper for reuse.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
audio_file, source_file, audio_wav are all optional with serde defaults.
v1 projects have audio_file, v2 projects have source_file + audio_wav.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Projects now save as folders containing .vtn + audio.wav:
My Transcript/
My Transcript.vtn
audio.wav
Audio handling:
- Always extract to 22kHz mono WAV on import (all formats, not just video)
- Prevents WebAudio crash from decoding large MP3/FLAC/OGG to PCM in memory
- WAV saved alongside .vtn on project save (moved from temp)
- Sidecar still uses original file (does its own conversion)
Project format v2:
- source_file: original import path (for re-extraction)
- audio_wav: relative path to extracted WAV (portable)
Re-link on open:
- If audio.wav exists → load directly
- If missing but source exists → re-extract automatically
- If both missing → dialog to locate file via file picker
- V1 project migration: extracts WAV on first open
New Rust commands: check_file_exists, copy_file, create_dir
extract_audio: now accepts optional output_path, uses 22kHz sample rate
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Files >1 hour are split into 5-minute chunks. Previously each chunk
showed "Starting transcription..." making it look like a restart.
Now shows "Chunk 3/12: Starting transcription..." and
"Chunk 3/12: Transcribing segment 5 (42% of audio)..."
Also skips the "Loading model..." message for chunks after the first
since the model is already loaded.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
AIChatPanel had its own hardcoded configMap with the old llama-server
URL (localhost:8080) and field names (local_model_path). Every chat
message reconfigured the provider with these wrong values, overriding
the correct settings applied at startup.
Fix: replace the duplicate with a call to the shared configureAIProvider().
Also strip trailing slashes from ollama_url before appending /v1 to
prevent double-slash URLs (http://localhost:11434//v1).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Cache loaded audio in _sf_load() — previously the entire WAV file was
re-read from disk for every 10s crop call. For a 3-hour file with
1000+ chunks, this meant ~345GB of disk reads. Now read once, cached.
- Better progress messages for long files: show elapsed time in m:ss
format, warn "(180min audio, this may take a while)" for files >10min
- Increased progress poll interval from 2s to 5s (less noise)
- Better time estimate: use 0.8x audio duration (was 0.5x)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Cancel button on the progress overlay during transcription
- Clicking Cancel shows confirmation: "Processing is incomplete. If you
cancel now, the transcription will need to be started over."
- "Continue Processing" dismisses the dialog, "Cancel Processing" stops
- Cancel clears partial results (segments, speakers) and resets UI
- Pipeline results are discarded if cancelled during processing
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
AI provider:
- Extract configureAIProvider() from saveSettings for reuse
- Call it on app startup after sidecar is ready (was only called on Save)
- Call it after first-time sidecar download completes
- Sidecar now receives correct Ollama URL/model immediately
Video extraction:
- Hide ffmpeg console window on Windows (CREATE_NO_WINDOW flag)
- Show "Extracting audio from video..." overlay with spinner during extraction
- UI stays responsive while ffmpeg runs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Video files (MP4, MKV, etc.) are now processed with ffmpeg to extract
audio to a temp WAV file before loading into wavesurfer. This prevents
the WebView crash caused by trying to fetch multi-GB files into memory.
- New extract_audio Tauri command uses ffmpeg (sidecar-bundled or system)
- Frontend detects video extensions and extracts audio automatically
- User-friendly error if ffmpeg is not installed with install instructions
- Reverted wavesurfer MediaElement approach in favor of clean extraction
- Added FFmpeg install guide to USER_GUIDE.md
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The configure action registered the provider but never called
set_active(), so the sidecar kept using the old/default provider.
Also updated the local provider default from localhost:8080 to
localhost:11434/v1 (Ollama). Added debug logging for configure.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Diarization: Audio.crop patch now pads short segments with zeros to
match the expected duration. pyannote batches embeddings with vstack
which requires uniform tensor sizes — the last segment of a file can
be shorter than the 10s window.
CI: Reordered sidecar workflow to check for python/ changes FIRST,
before bumping version or configuring git. All subsequent steps are
gated on has_changes. This prevents unnecessary version bumps and
build runs when only app code changes.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Runs daily at 6am UTC and on manual dispatch. Separately tracks app
releases (v*) and sidecar releases (sidecar-v*), keeping the latest
5 of each and deleting older ones along with their tags.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
base_url was being set to 'http://localhost:11434/v1' by the frontend,
then LocalProvider appended another '/v1', resulting in '/v1/v1'.
Now the provider uses base_url directly (frontend already appends /v1).
Also fixed health check to hit Ollama root instead of /health.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous patch only replaced Audio.__call__ (segmentation), but
pyannote also calls Audio.crop during speaker embedding extraction.
crop loads a time segment of audio — patched to load full file via
soundfile then slice the tensor to the requested time range.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
