soundfile needs collect_all() to include libsndfile native library —
hiddenimports alone wasn't enough, causing 'No module named soundfile'
in the frozen sidecar. This is needed for the pyannote Audio patch
that bypasses torchaudio/torchcodec.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- README: Updated to reflect current architecture (decoupled app/sidecar),
Ollama as local AI, CUDA support, split CI workflows
- USER_GUIDE.md: Complete how-to including first-time setup, transcription
workflow, speaker detection setup, Ollama configuration, export formats,
keyboard shortcuts, and troubleshooting
- CONTRIBUTING.md: Dev setup, project structure, conventions, CI/CD overview
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add min-width: 0 to flex container (allows shrinking for wrap)
- Add overflow-x: hidden to prevent horizontal scroll
- Add white-space: pre-wrap to segment text
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- AI Provider: "Local (llama-server)" changed to "Ollama" with URL and
model fields (defaults to localhost:11434, llama3.2)
- Ollama connects via its OpenAI-compatible API (/v1 endpoint)
- Removed empty "Local AI" tab
- Renamed "Developer" tab to "Debug"
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
torchaudio 2.10 unconditionally delegates load() to torchcodec, ignoring
the backend parameter. Since torchcodec is excluded from PyInstaller,
this broke our pyannote Audio monkey-patch.
Fix: replace torchaudio.load() with soundfile.read() + torch.from_numpy().
soundfile handles WAV natively (audio is pre-converted to WAV), has no
torchcodec dependency, and is already a transitive dependency.
Also added soundfile to PyInstaller hiddenimports.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CSP: Add blob: to connect-src/img-src/media-src for wavesurfer.js audio
playback. Add http://tauri.localhost to default-src for devtools.
pyannote: sys.modules block didn't work — pyannote still uses AudioDecoder
unconditionally. New approach: monkey-patch Audio.__call__ in diarize.py
to use torchaudio.load() directly, bypassing the broken torchcodec path.
Patch runs once before pipeline loading.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- DevTools off by default (no more auto-open on launch)
- New "Developer" tab in Settings with a checkbox to toggle devtools
- Toggle takes effect immediately (opens/closes inspector)
- Setting persists: devtools restored on next launch if enabled
- toggle_devtools Tauri command wraps window.open/close_devtools
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CSP: Add connect-src for ipc.localhost and asset.localhost so Tauri IPC
commands and local file loading (waveform, audio playback) work.
pyannote: Block torchcodec in sys.modules at startup so pyannote.audio
falls back to torchaudio for audio decoding. pyannote has a bug where
it uses AudioDecoder unconditionally even when torchcodec import fails.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Enable Tauri devtools feature so right-click Inspect works in release
- Open devtools automatically on launch for debugging
- Add log_frontend command: frontend can write to ~/.voicetonotes/frontend.log
- Sidecar logs go to %LOCALAPPDATA%/com.voicetonotes.app/sidecar.log
- Frontend logs go to %USERPROFILE%/.voicetonotes/frontend.log
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
torchcodec is partially bundled but non-functional (missing FFmpeg DLLs),
causing pyannote.audio to try AudioDecoder which fails with NameError.
Excluding it forces pyannote to fall back to torchaudio for audio loading.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add git pull --rebase before push in both version bump workflows to
handle concurrent pushes from parallel workflows
- Add explicit python/ change detection in sidecar workflow (Gitea may
not support paths filter), skip all jobs if no python changes
- Gate all sidecar build jobs on has_changes output
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Sidecar now has its own version (1.0.0) and release lifecycle:
- Sidecar tags: sidecar-v1.0.0, sidecar-v1.0.1, etc.
- App tags: v0.2.x (unchanged)
- Sidecar workflow triggers only on python/** changes or manual dispatch
- App release no longer bumps python/pyproject.toml
Sidecar version tracked via sidecar-version.txt in app data dir:
- resolve_sidecar_path() reads version from file instead of CARGO_PKG_VERSION
- download_sidecar() fetches latest sidecar-v* release from Gitea API
- check_sidecar_update() compares local vs remote sidecar versions
- Version file written after successful download
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CI split:
- release.yml: version bump + lightweight app builds (no Python/sidecar)
- build-sidecar.yml: builds CPU + CUDA sidecar variants per platform,
uploads as separate release assets, runs in parallel with app builds
- Sidecar workflow uses retry loop to find release (race with version bump)
Fixes:
- Add reqwest "json" feature for .json() method
- Add explicit type annotations for reqwest Response and bytes::Bytes
- Reuse client instance for download (was using reqwest::get directly)
Bundle targets: deb, rpm, nsis, msi, dmg (all formats, app is small now)
Windows upload finds both *.msi and *-setup.exe
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Major refactor: sidecar is no longer bundled in the installer. Instead,
it's downloaded on first launch with a setup screen offering CPU vs CUDA
choice. This solves the 2GB+ installer size limit and decouples app/sidecar.
Backend:
- New commands: check_sidecar, download_sidecar, check_sidecar_update
- Streaming download with progress events via reqwest
- Added reqwest + futures-util dependencies
- Removed sidecar.zip from bundle resources
- Restored NSIS target (no longer size-constrained)
CI:
- Each platform builds both CPU and CUDA sidecar variants (except macOS: CPU only)
- Sidecar zips uploaded as separate release assets
- Asset naming: sidecar-{os}-{arch}-{variant}.zip
Frontend:
- SidecarSetup.svelte: first-launch setup with CPU/CUDA radio choice,
progress bar, error/retry handling
- Update banner on launch if newer sidecar version available
- Conditional rendering: setup screen → main app flow
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add rpm to bundle targets and install rpm on Linux CI
- Upload both .deb and .rpm from Linux build
- Install 7-Zip via choco if not already available on Windows runner
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace Compress-Archive (2GB limit) with 7z for sidecar packaging
- Remove NSIS from bundle targets — NSIS has a 2GB per-file limit that
breaks with CUDA-sized sidecar.zip; MSI (WiX) handles large files
by splitting into multiple CABs
- Update Windows upload to look for .msi only
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Default torch on PyPI is CPU-only on Windows. Must use PyTorch's own
package index (cu126) to get CUDA-enabled wheels. This also pins the
CUDA version on Linux for deterministic builds.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Windows and Linux sidecar builds now use --with-cuda for GPU acceleration
(macOS stays CPU-only — Apple Silicon uses Metal, not CUDA)
- Windows upload switched from --data-binary to -T streaming for 2GB+ files
- Add cleanup_old_sidecars() that removes stale sidecar-* directories on
startup, keeping only the current version
- Add NSIS uninstall hook to remove sidecar data dir on Windows uninstall
(user data in ~/.voicetonotes is preserved)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- transcribe: catch model load failures on CUDA and retry with CPU
- hardware detect: test CUDA runtime actually works (torch.zeros on cuda)
before recommending GPU, since CPU-only builds detect CUDA via driver
but lack cublas/cuDNN libraries
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Exclude ctranslate2.converters from PyInstaller bundle — these modules
import torch at module level causing circular import crashes, and are
only needed for model conversion (never used at runtime)
- Defer all heavy ML imports to first handler call instead of startup,
so the sidecar can send its ready message without loading torch/whisper
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Fix is_running() to check actual process liveness via try_wait()
instead of just checking if the handle exists
- Auto-restart sidecar on pipe errors (broken pipe, closed stdout)
with one retry attempt
- Hide sidecar console window on Windows (CREATE_NO_WINDOW flag)
- Log sidecar stderr to sidecar.log file for crash diagnostics
- Include exit status in error message when sidecar fails to start
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
