Speaches reliability release (0.3.4)
Bug Fixes & Improvements
- 🔧 Disabled Speaches VAD (Voice Activity Detection) by default for better Wyoming protocol compatibility
- In testing this actually speeds up response times
- 🐛 Fixed issue where empty transcripts weren't being sent properly for recognition by Home Assistant
- ♻️ Improved resource management with async context managers for STT and TTS clients
Documentation
- 📚 Updated sequence diagram to better illustrate streaming vs non-streaming transcription flow
This release focuses on stability improvements and better compatibility with the Wyoming protocol, particularly for users of the Speaches backend.
Full Changelog: v0.3.3...v0.3.4