All pull requests

Pull requests

• fix(soniox): end speech after a fully retracted segment

  #2804·rosetta-livekit-bot[bot] opened Oct 21, 2026Bot·

  ·
  3

• fix(voice): don't cut an AgentTask's speech on shutdown()

  #2803·rosetta-livekit-bot[bot] opened Oct 21, 2026Bot·

  ·
  3

• feat(inworld): cancel the TTS context when a stream is closed

  #2802·ianbbqzy opened Oct 21, 2026Contributor·

  ·
  4

• Add audio_sample_rate and model args for LemonSlice avatars

  #2801·jp-lemon opened Oct 21, 2026Contributor·4 of 8 tasks·

  ·
  5

• fix(inference): Drop the sampling params gpt-6 rejects while it reasons

  #2800·adrian-cowham opened Oct 20, 2026Contributor·

  ·
  7

• feat(inworld): let the server split streamed text on inworld-tts-2

  #2799·ianbbqzy opened Oct 20, 2026Contributor·

  ·
  1

• feat(elevenlabs): enable manual STT flushing

  #2798·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  4

• fix(sarvam): keep stt streams open across flushes

  #2797·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  4

• feat(azure): add DecisionModel for Microsoft-Decision-1

  #2796·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  6

• fix(voice): keep preemptive generation when a transcript only changes formatting

  #2794·u9g opened Oct 20, 2026Contributor·6 of 8 tasks·

  ·
  1

• fix(voice): keep the session host open for finalize_simulation after the agent ends the call

  #2793·u9g opened Oct 20, 2026Contributor·

  ·
  4

• fix(voice): add say() text to the realtime chat context when it is spoken

  #2791·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  5

• fix(realtime): treat credit_balance_exhausted as a fatal OpenAI error

  #2790·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  1

• fix(inference): emit every parallel tool call sent in one chunk

  #2789·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  3

• fix(tokenize): don't hold a CJK reply until the end in word streams

  #2788·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  4

• fix(meta): don't delay Muse STT audio by the websocket connect time

  #2786·rosetta-livekit-bot[bot] opened Oct 20, 2026Bot·

  ·
  3

• fix(elevenlabs): don't crash when an STT stream closes while connecting

  #2785·sarankumar1325 opened Oct 20, 2026·6 of 8 tasks·

  ·
  1

  3

• fix(voice): cancel pending speech readers on interruption

  #2784·devp19 opened Oct 20, 2026·9 tasks done·

  ·
  2

• feat(voice): add silent and toolChoice to ctx.update()

  #2783·longcw opened Oct 20, 2026Contributor·

  ·
  6

• fix(voice): stop waitForInactive from spinning while someone speaks

  #2782·longcw opened Oct 20, 2026Contributor·

  ·
  2

• fix(voice): watch an async tool's deferred reply as a Task

  #2781·longcw opened Oct 20, 2026Contributor·

  ·
  3

• feat(finesse): add Finesse TTS plugin

  #2777·iyashjayesh opened Oct 19, 2026·

  ·
  14

• fix(inference): do not reconnect adaptive interruption stream

  #2776·rosetta-livekit-bot[bot] opened Oct 19, 2026Bot·

  ·
  1

• fix(bey): don't retry 4xx when starting a session, and keep the provider's error

  #2775·rosetta-livekit-bot[bot] opened Oct 19, 2026Bot·

  ·
  3

• fix(bey): use the current session endpoint and send 16 kHz audio

  #2774·rosetta-livekit-bot[bot] opened Oct 19, 2026Bot·

  ·
  1