- Rename middle-click-reader to read-aloud service
- Change hotkey from Ctrl+middle-click to Alt+R
- Replace edge-tts with Piper TTS for local neural voices
- Update desktop and service files
- Add piper-tts dependency
- Update tests and setup scripts
This is a comprehensive refactoring that transforms the dictation service from a
complex multi-mode application into two clean, focused features:
1. Voice dictation with system tray icon
2. On-demand read-aloud via Ctrl+middle-click
## Key Changes
### Dictation Service Enhancements
- Add GTK/AppIndicator3 system tray icon for visual status
- Remove all notification spam (dictation start/stop/status)
- Icon states: microphone-muted (OFF) → microphone-high (ON)
- Click tray icon to toggle dictation (same as Alt+D)
- Simplify ai_dictation_simple.py by removing conversation mode
### Read-Aloud Service Redesign
- Replace automatic clipboard reader with on-demand Ctrl+middle-click
- New middle_click_reader.py service
- Works anywhere: highlight text, Ctrl+middle-click to read
- Uses Edge-TTS (Christopher voice) with mpv playback
- Lock file prevents feedback with dictation service
### Conversation Mode Removed
- Delete all VLLM/conversation code (VLLMClient, ConversationManager, TTS)
- Archive 5 old implementations to archive/old_implementations/
- Remove conversation-related scripts and services
- Clean separation of concerns for future reintegration if needed
### Dependencies Cleanup
- Remove: openai, aiohttp, pyttsx3, requests (conversation deps)
- Keep: PyGObject, pynput, sounddevice, vosk, numpy, edge-tts
- Net reduction: 4 packages removed, 6 core packages retained
### Testing Improvements
- Add test_dictation_service.py (8 tests) ✅
- Add test_middle_click.py (11 tests) ✅
- Fix test_run.py to use correct model path
- Total: 19 unit tests passing
- Delete obsolete test files (test_suite, test_vllm_integration, etc.)
### Documentation
- Add CHANGES.md with complete changelog
- Add docs/MIGRATION_GUIDE.md for upgrading
- Add README.md with quick start guide
- Update docs/README.md with current features only
- Add justfile for common tasks
### New Services & Scripts
- Add middle-click-reader.service (systemd)
- Add scripts/setup-middle-click-reader.sh
- Add desktop files for autostart
- Remove toggle-conversation.sh (obsolete)
## Impact
**Code Quality**
- Net change: -6,007 lines (596 added, 6,603 deleted)
- Simpler architecture, easier maintenance
- Better test coverage (19 tests vs mixed before)
- Cleaner separation of concerns
**User Experience**
- No notification spam during dictation
- Clean visual status via tray icon
- Full control over read-aloud (no unwanted readings)
- Better performance (fewer background processes)
**Privacy**
- No conversation data stored
- No VLLM connection needed
- All processing local except Edge-TTS text
## Migration Notes
Users upgrading should:
1. Run `uv sync` to update dependencies
2. Restart dictation.service to get tray icon
3. Run scripts/setup-middle-click-reader.sh for new read-aloud
4. Remove old read-aloud.service if present
See docs/MIGRATION_GUIDE.md for details.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Replace exact text preview with word count in notifications
- Change 'Speaking' notification to show 'Dictating... (X words)' instead of text
- Change 'Dictation' notification to show 'Text typed successfully (X words)' instead of preview
- Improves privacy by not displaying sensitive dictation content in notifications
- Fix state detection priority: dictation now takes precedence over conversation
- Fix critical bug: event loop was created but never started, preventing async coroutines from executing
- Optimize audio processing: reorder AcceptWaveform/PartialResult checks
- Switch to faster Vosk model: vosk-model-en-us-0.22-lgraph for 2-3x speed improvement
- Reduce block size from 8000 to 4000 for lower latency
- Add filtering to remove spurious 'the', 'a', 'an' words from start/end of transcriptions
- Update toggle-dictation.sh to properly clean up conversation lock file
- Improve batch audio processing for better responsiveness
