Releases: livekit/agents
livekit-agents@0.8.8
v0.7.x to v0.8.x migration guide
Patch Changes
-
fix uninitialized SpeechHandle error on interruption - #665 (@theomonnom)
-
voiceassistant: avoid stacking assistant replies when allow_interruptions=False - #667 (@theomonnom)
-
fix: disconnect event may now have a arguments - #668 (@theomonnom)
livekit-plugins-nltk@0.7.2
Patch Changes
- fix another semver break - #659 (@theomonnom)
livekit-plugins-browser@0.0.2
Patch Changes
- livekit-plugins-browser: prepare for release - #659 (@theomonnom)
livekit-agents@0.8.7
Patch Changes
- voiceassistant: fix llm not having the full chat context on bad interruption timing - #659 (@theomonnom)
livekit-plugins-silero@0.6.4
Patch Changes
-
silero: adjust vad activation threshold - #639 (@theomonnom)
-
silero: fix vad padding & static audio - #631 (@theomonnom)
livekit-plugins-rag@0.2.2
livekit-plugins-openai@0.8.1
Patch Changes
-
add support for Ollama, Perplexity, Fireworks, Octo, Together, and Groq LLMs through the OpenAI API - #611 (@nbsp)
-
Support OpenAI Assistants API as a beta feature under
livekit.plugins.openai.beta
- #601 (@keepingitneil)
Add _metadata to ChatCtx and ChatMessage which can be used (in the case of OpenAI assistants) for bookeeping to sync local state with remote, OpenAI state
livekit-plugins-nltk@0.7.1
Patch Changes
- Revert "nltk: fix broken punkt download" - #630 (@theomonnom)
livekit-plugins-deepgram@0.6.5
Patch Changes
- deepgram: fallback to nova-2-general when the language isn't supported - #623 (@theomonnom)
livekit-plugins-cartesia@0.4.1
Patch Changes
- Switch Cartesia to a sentence tokenizer and keep the same context id throughout. - #608 (@keepingitneil)
Propagate segment_id through the basic sentence tokenizer