What's Changed
- fix(cartesia): redact API keys from websocket handshake errors by @LHMQ878 in #6740
- fix: respect non-retryable STT errors by @chenghao-mou in #6759
- fix: fallback adapters report the active instance's model/provider by @Fyko in #6690
- fix(inference): don't crash the session on null usage token counts by @biztex in #6763
- fix(voice): stop the interrupt walk at a queued speech that disallows interruptions by @biztex in #6644
- fix(voice): don't raise when interrupting a cancelled or finished speech by @longcw in #6767
- fix: isolate bundled VAD inference from the default executor by @chenghao-mou in #6770
- fix(xai): align realtime Voice Agent defaults and API coverage by @damien-xai in #6755
- feat(spatius): default audio format to ogg_opus by @3DRX in #6749
- fix: don't count discarded room audio as played by @chenghao-mou in #6728
- feat(azure): replace unknown tts model string with host/region + voice by @alvxrog in #6775
- feat: expose cache_creation_tokens in LLM metrics by @chasleslr in #6655
- fix: treat a silent agent turn as silence, not a run_input error by @theomonnom in #6779
- fix: stop dropping session host responses by @theomonnom in #6781
- fix(llm): never give a const field the null default sentinel by @u9g in #6688
- fix: hold pending replies at playout launch by @chenghao-mou in #6745
- fix(inference): keep a stream retryable until it emits generation by @u9g in #6697
- test: stabilize text email collection eval by @chenghao-mou in #6791
- feat(evals): run judges at low inference priority by @u9g in #6794
- feat(simulation): pin text simulations to the low inference class by @theomonnom in #6721
- fix(google): allow explicit project for Speech-to-Text by @uuzzrm in #6810
- fix(silero): update bundled VAD checkpoint to v6.2 by @longcw in #6816
- fix(voice): export uncommitted user_turn spans on close by @chenghao-mou in #6592
- feat(telemetry): include recording options in session reports by @chenghao-mou in #6547
- fix: respect agent endpointing delay during reply hold-off by @chenghao-mou in #6820
- docs: require confirmation for non-patch version bumps by @chenghao-mou in #6821
- examples: switch homepage agent to Fish Audio with expressive mode by @theomonnom in #6822
- fix: stabilize LiveKit pronunciation with Fish Audio by @chenghao-mou in #6825
- fix(deps): pin openai below 3 by @longcw in #6831
- livekit-agents@1.6.10 by @github-actions[bot] in #6834
New Contributors
- @Fyko made their first contribution in #6690
- @damien-xai made their first contribution in #6755
- @alvxrog made their first contribution in #6775
- @chasleslr made their first contribution in #6655
Full Changelog: https://github.com/livekit/agents/compare/livekit-agents@1.6.9...livekit-agents@1.6.10