What's Changed
- fix(voice): agent and user state while a tool runs by @longcw in #6937
- chore: teach reviews to catch untagged PII logs by @chenghao-mou in #6923
- feat(agents): serialize the inline AgentTasks awaited from one turn by @u9g in #6862
- fix(openai): read TTS response by content-type, not model name by @abidullahdev in #6930
- fix: avoid duplicate agent speech end by @chenghao-mou in #6938
- feat(speechify): stream with timestamps via /v1/audio/stream/with-timestamps by @luke-speechify in #6926
- fix(openai realtime): keep derived capabilities in sync with update_options by @longcw in #6943
- fix(google): keep unspoken model text out of the realtime transcript by @rosetta-livekit-bot[bot] in #6955
- fix(realtime): cancel the generation when a speech is interrupted by @longcw in #6962
- fix(openai realtime): recover when the server anchors to a deleted item by @longcw in #6964
- fix(voice): place recorded audio where it actually played by @longcw in #6921
- fix(elevenlabs): stop routing TTS audio to an interrupted stream by @longcw in #6936
- fix: raise the shared http session total timeout to 900s by @chenghao-mou in #6941
- feat(sarvam): add realtime streaming STT by @dhruvladia-sarvam in #6562
- feat(sarvam): add glm5.2, gemma4 and sarvam-105b-conversations model support by @dhruvladia-sarvam in #6841
- Add saaras:v4 as the default and latest Sarvam STT model by @dhruvladia-sarvam in #6900
- Bump
livekitto 1.1.15 by @1egoman in #6970 - (xai inference): add LLM models by @tinalenguyen in #6971
- fix(dtmf): drop removed repeat_instructions arg from GetDtmfTask example and docstring by @feizhuzheng in #6972
- Sunset saaras and saarika v2.5 by @dhruvladia-sarvam in #6905
- feat: add the deployed region to every log record by @theomonnom in #6896
- fix(llm): place a backdated chat item by its creation time by @longcw in #6974
- docs(sarvam): correct STT README brand from Saarika to Saaras by @detail-app[bot] in #6979
- feat(elevenlabs): stream eleven_v3/eleven_v3_conversational via text-to-dialogue by @ameyakhare in #6838
- fix(aws): support transcribe streaming config rename by @mikemikimike in #6959
- fix: retry recording upload connection failures by @chenghao-mou in #6977
- fix: preserve cancellation during audio recognition close by @chenghao-mou in #6903
- fix: support adaptive interruption without word timestamps by @chenghao-mou in #6765
- fix: scope interruption verdicts to agent speech by @chenghao-mou in #6957
- fix: attribute playback evidence to its own segment on the same audio output by @chenghao-mou in #6470
- chore(deps): update dependency aiohttp to v3.14.3 [security] by @renovate[bot] in #6692
- fix(workflows): ask before recording a volunteered value by @u9g in #6980
- fix: warn about realtime audio redaction by @chenghao-mou in #6987
- fix(hotel_receptionist): move book_room's sequencing and pricing rules into the tools by @u9g in #6849
- fix(aws): preserve idle-timeout retry in STT stream cleanup by @tanish111 in #6995
- feat(simulation): stamp the run and job ids on session telemetry by @theomonnom in #6997
- (gemini stt): add gemini-3.5-transcribe-live to inference and plugin by @tinalenguyen in #6991
- (speechmatics inference): add linden model by @tinalenguyen in #6989
- Add Palabra STT and TTS plugins by @danlowkey in #6681
- livekit-agents@1.7.1 by @github-actions[bot] in #6999
New Contributors
- @abidullahdev made their first contribution in #6930
- @rosetta-livekit-bot[bot] made their first contribution in #6955
- @feizhuzheng made their first contribution in #6972
- @ameyakhare made their first contribution in #6838
- @mikemikimike made their first contribution in #6959
- @tanish111 made their first contribution in #6995
- @danlowkey made their first contribution in #6681
Full Changelog: https://github.com/livekit/agents/compare/livekit-agents@1.7.0...livekit-agents@1.7.1