{"full_name":"livekit/agents","owner":"livekit","name":"agents","shard":"1","url":"https://github.com/livekit/agents","description":"A framework for building realtime voice AI agents 🤖🎙️📹 ","homepage":"https://docs.livekit.io/agents","topics":["ai","real-time","voice","video","agents","openai"],"tag_pages":["agents","ai","openai","real-time","video","voice"],"language":"Python","license":"Apache-2.0","avatar":"https://avatars.githubusercontent.com/u/69438833?v=4","created_at":"2023-10-19T23:00:55Z","pushed_at":"2026-11-15T18:14:51Z","metrics":{"stars":15237,"forks":4143,"watchers":100,"issues":2339,"prs":6156,"releases":383,"issues_open":254,"issues_closed":2085,"prs_open":714,"prs_closed":1337,"prs_merged":4105,"discussions":0},"growth":{"day":{"stars":10,"forks":5,"base_day":"2026-11-15"},"week":{"stars":161,"forks":85,"base_day":"2026-11-04"}},"history":{"dates":["2026-11-03","2026-11-04","2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[15068,15076,15161,15176,15193,15202,15214,15227,15237],"forks":[4054,4058,4113,4118,4125,4126,4132,4138,4143],"watchers":[101,101,null,100,100,100,100,100,100],"issues_open":[null,null,null,253,254,253,252,252,254],"prs_open":[null,null,null,692,694,692,694,694,714]},"boards":{},"same_owner":[{"full_name":"livekit/livekit","name":"livekit","shard":"1","boards":{}}],"issues":[{"number":3863,"title":"AudioStreamDecoder - error decoding audio","comments":26,"created_at":"2025-11-10T15:50:01Z"},{"number":3515,"title":"`min_interruption_duration` seems to be ignored over `min_interruption_words`","comments":26,"created_at":"2025-09-26T14:50:08Z"},{"number":391,"title":"Make VoiceAssistant support multi Participants","comments":25,"created_at":"2024-06-25T09:24:54Z"},{"number":3375,"title":"Whatsapp calling API support","comments":16,"created_at":"2025-09-08T09:24:41Z"},{"number":3598,"title":"How do you synchronize extra audio like sound effects with TTS?","comments":12,"created_at":"2025-10-07T22:42:01Z"},{"number":2907,"title":"Backchannel Prediction","comments":12,"created_at":"2025-07-15T13:46:23Z"},{"number":3450,"title":"Completely shutdown STT or agents speaking","comments":11,"created_at":"2025-09-18T05:57:33Z"},{"number":7215,"title":"AsyncTools can't continuously update the user while also treating tool output as a normal output","comments":9,"created_at":"2026-09-10T20:02:50Z"}],"discussions":[],"latest_release":{"name":"livekit-agents@1.8.12","tag":"livekit-agents@1.8.12","published_at":"2026-11-13T17:37:40Z","body":"## Delegation recovery\r\n\r\nYou can recover caller answers that GPT-Live transcribes but never delegates. After a quiet grace period, a detector checks missed turns and starts backend delegation when needed, with a note for answers spoken over the agent.\r\n\r\n```python\r\nGPTLiveModel(delegation_recovery=True)\r\n```\r\n\r\n## New providers and models\r\n\r\n- **Lokutor**: Conv 2.0 STT and Versa 2.0 TTS across nine languages.\r\n- **xAI**: `xai/stt-2` through Inference, with word alignment for adaptive interruption.\r\n- **Google**: Gemini 4 Carbon, Gemini 4 Hydrogen (also through Inference), and `gemini-decision-1` through the Decisions API.\r\n- **Perplexity**: Flash and standard Decider v1.3, v1.2, and latest aliases.\r\n- **AssemblyAI**: `assemblyai-stt-3-7-realtime-agent` streaming STT.\r\n- **Baseten**: `streaming-v2` transcription protocol.\r\n- **Turn detector**: updated local v1-mini model with turn-length input.\r\n\r\n## Breaking changes\r\n\r\n- **GPT-Live**: `delegation_created` listeners now also receive backend delegations; filter `mode=\"responses\"` where needed.\r\n- **NVIDIA**: TTS output is restricted to `LINEAR_PCM`; update other encoding configurations.\r\n\r\n## Behavior changes\r\n\r\n- **Dependencies**: Google now requires `google-genai>=3.2`, Phonic requires `phonic>=0.33.4`, and local inference requires `>=0.3.0,<0.4`.\r\n- **Jobs**: uncaught entrypoint failures report failure; job-triggered session closes report `JOB_SHUTDOWN`.\r\n- **Manual turns**: committed snapshots survive clears and subsequent VAD activity.\r\n- **ElevenLabs**: dialogue models now honor `similarity_boost`.\r\n\r\nInterruptions, tool delivery, handoffs, telemetry authentication, speech connections, and answering-machine detection are more reliable.","is_prerelease":false},"readme":true,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}