13 Pull requests merged by 7 people

• Update events.py with new AWS Nova Sonic voice

  #3048 merged Jul 30, 2025

• enable - gemini-live-2.5-flash-preview

  #3016 merged Jul 30, 2025

• add modalities arg to OAI realtime model with_azure

  #3049 merged Jul 30, 2025

• Fix RuntimeError on scheduling speech and on mark_generation_done

  #3032 merged Jul 29, 2025

• add retry_on_chunk_sent to FallbackAdapter for allowing retry after chunk sent

  #3033 merged Jul 29, 2025

• improve tts error trace and expose cartesia error message

  #3028 merged Jul 29, 2025

• fix aws stt restarting

  #2994 merged Jul 29, 2025

• feat(azure-stt): add phrase list support for keyword boosting

  #3024 merged Jul 28, 2025

• Gladia STT - fix region parameter to gladia stt on query parameters

  #3018 merged Jul 28, 2025

• Fix incorrect streaming flag for neuphonic

  #3036 merged Jul 28, 2025

• add metadata when set tracer provider

  #3034 merged Jul 28, 2025

• pin the hedra avatar sample rate to 16k

  #3027 merged Jul 26, 2025

• fix gemini realtime generate_reply

  #3017 merged Jul 25, 2025

11 Pull requests opened by 10 people

• feature/enable-prometheus-multiprocess-collector

  #3021 opened Jul 25, 2025

• Update readmes for new avatar plugins, don't require api key for simli init

  #3025 opened Jul 25, 2025

• feat: allow custom track name to be used with RoomIO

  #3029 opened Jul 27, 2025

• feat(anthropic): Add support for Claude's extended thinking feature

  #3037 opened Jul 28, 2025

• Add support for API key and secret in STTOptions and SpeechStream

  #3039 opened Jul 29, 2025

• Updated to send out stt results for deepgram that has a minimum amoun…

  #3041 opened Jul 29, 2025

• use tts as fallback when realtime model returns text response

  #3050 opened Jul 30, 2025

• fix using AgentTask in on_enter and on_ext

  #3051 opened Jul 30, 2025

• Add full list of supported languages for Google Cloud STT v2

  #3053 opened Jul 30, 2025

• Add Trillet Antibot Plugin for real-time AI voice deepfake detection

  #3054 opened Jul 31, 2025

• feat: support bithuman cloud avatar

  #3055 opened Jul 31, 2025

19 Issues closed by 6 people

• Fallback Adapter not catching Gemini parsing failures

  #2488 closed Jul 31, 2025

• How can I set up track subscription permissions

  #1417 closed Jul 31, 2025

• Is update_chat_ctx supported for AWS Realtime?

  #2977 closed Jul 30, 2025

• [Google plugin] Feature Request: External VAD for Gemini Live realtime models

  #3047 closed Jul 30, 2025

• Feature request - before_stt_cb

  #1522 closed Jul 30, 2025

• How to change the default value for min_sentence_length for sentence tokenisation (to avoid concurrency limits of tts service)

  #2878 closed Jul 30, 2025

• AttributeError: 'RunContext' object has no attribute 'chat_ctx'

  #3040 closed Jul 30, 2025

• AWS STT not working when allow_interruptions=False in LiveKit AgentSession

  #2961 closed Jul 29, 2025

• Gladia STT v1.2.2 causing API Connection Error

  #3014 closed Jul 29, 2025

• coroutine EOUModelBase.unlikely_threshold' was never awaited

  #3038 closed Jul 29, 2025

• When i use connect --room <room_name> the agent enters the room and sends the initial message but it is not detected , but when i use the same agent in dev or start mode, i recieve the messages but the audio track is not attache

  #1421 closed Jul 29, 2025

• Sentence chunking for Cartesia TTS

  #1180 closed Jul 29, 2025

• Duplicate response with Gemini Live API

  #2884 closed Jul 28, 2025

• DLL load failed while importing lk_blingfire

  #3031 closed Jul 27, 2025

• 1.2.0 -> 1.2.2 "self.session.generate_reply()" Stopped working for Realtime Gemini Models

  #3015 closed Jul 25, 2025

• Record room agent side instead of using Egress (save minutes and money)

  #2993 closed Jul 25, 2025

• Does the TrackEgressRequest work for streaming audio?

  #3012 closed Jul 25, 2025

• Feature Request: Allow Selective Track Subscription for VoiceAssistant

  #1399 closed Jul 24, 2025

• Offer manual controls to signal end of turn

  #1026 closed Jul 24, 2025

10 Issues opened by 9 people

• TypeError: RealtimeModel.__init__() got an unexpected keyword argument 'modalities' {"pid": 52358, "job_id": "AJ_Bpwup9frPTzB"}

  #3046 opened Jul 29, 2025

• 🚀 Feature Request: Add Support for Deepgram Voice Agent API with Function Calling

  #3045 opened Jul 29, 2025

• `AgentTask should only be awaited inside on_enter` crash despite being in `on_enter` function

  #3044 opened Jul 29, 2025

• Google TTS gives error - no audio frames were pushed for text

  #3043 opened Jul 29, 2025

• How get language detect by nova-3

  #3042 opened Jul 29, 2025

• How to perform call tool during agent speech?

  #3035 opened Jul 28, 2025

• How to insert filler words in the TTS queue before the main response audio?

  #3030 opened Jul 27, 2025

• Handling Phone Calls During Pod Upgrades in Kubernetes

  #3020 opened Jul 25, 2025

• Allow dynamic control of interruptions during conversation phases

  #3019 opened Jul 25, 2025

• LangGraph RemoteGraph raises AttributeError with HumanMessage input state

  #3011 opened Jul 24, 2025

26 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Add Rime TTS example for generating real-time audio from text

  #2972 commented on Jul 29, 2025 • 1 new comment

• Add Ultravox Realtime API Plugin

  #2992 commented on Jul 30, 2025 • 0 new comments

• update/mistralai-plugins

  #2887 commented on Jul 30, 2025 • 0 new comments

• Fix LangGraph LLMAdapter streaming node filtering

  #2838 commented on Jul 25, 2025 • 0 new comments

• Rebase on latest main upstream

  #2812 commented on Jul 28, 2025 • 0 new comments

• Add Fireworks AI STT and Realtime Voice Agent plugin

  #2687 commented on Jul 28, 2025 • 0 new comments

• Speaker ID using Speechmatics STT

  #2625 commented on Jul 30, 2025 • 0 new comments

• TTS Optimization to generate speech lazily

  #2981 commented on Jul 31, 2025 • 0 new comments

• AssertionError in turn_detector: “end_of_utterance prediction should always return a result”

  #2465 commented on Jul 31, 2025 • 0 new comments

• Bug Report: Egress Recording Stops Capturing Agent Voice Mid-Conversation

  #2716 commented on Jul 31, 2025 • 0 new comments

• Agent Can hear SIP Participant but cannot hear Browser Participant

  #2046 commented on Jul 31, 2025 • 0 new comments

• Feature Request: Recording Audio Inside The Agent

  #2310 commented on Jul 30, 2025 • 0 new comments

• Frequent av.error.EOFError kills turn-taking — agent answers only on 2nd prompt

  #2823 commented on Jul 30, 2025 • 0 new comments

• Agent State Not Updated In Real-Time When Using Tavus Avatar

  #3007 commented on Jul 30, 2025 • 0 new comments

• Feature Request: Deepgram Voice Agent API

  #2615 commented on Jul 29, 2025 • 0 new comments

• Backchannel Prediction

  #2907 commented on Jul 29, 2025 • 0 new comments

• Google Cloud TTS isn't working while running

  #2951 commented on Jul 29, 2025 • 0 new comments

• LLM gives no answer

  #2928 commented on Jul 29, 2025 • 0 new comments

• ERROR livekit.agents - worker failed

  #2975 commented on Jul 28, 2025 • 0 new comments

• Feature Request: Utterance timestamps in the ChatContext or Transcript

  #2326 commented on Jul 27, 2025 • 0 new comments

• issue with _on_metrics_collected

  #2879 commented on Jul 26, 2025 • 0 new comments

• Call Transfer Not Working as Expected

  #1305 commented on Jul 26, 2025 • 0 new comments

• Event Loop availability in initialize/prewarm

  #2585 commented on Jul 25, 2025 • 0 new comments

• Support for Google TTS Chirp 3 Custom Voice

  #3013 commented on Jul 25, 2025 • 0 new comments

• Multilingual Agent (STT and TTS); is that possible with LiveKit?

  #1335 commented on Jul 24, 2025 • 0 new comments

• LangGraph LLMAdapter streams intermediate LLM outputs instead of only final responses

  #2836 commented on Jul 24, 2025 • 0 new comments