All pull requests

Pull requests

• feat(agents): add DecisionModel for background decisions

  #2735·rosetta-livekit-bot[bot] opened Oct 15, 2026Bot·

  ·
  6

• fix(elevenlabs): place text-to-dialogue character times on the context clock

  #2734·OliverA100 opened Oct 14, 2026·6 of 8 tasks·

  ·
  1

  7

• fix(llm): treat null for an optional tool argument as omitted

  #2732·u9g opened Oct 14, 2026Contributor·

  ·
  3

• feat(aitwin): add Ai Twin plugin for browser speakText()

  #2730·Swaraj07082 opened Oct 14, 2026·8 tasks·

  ·
  6

• fix(anthropic): drop unpaired tool calls and lead user turns with tool_result

  #2729·Apeksha9389 opened Oct 14, 2026·6 of 7 tasks·

  ·
  1

  2

• fix(audio): stop audioFramesFromFile writes from rejecting unhandled after the consumer stops

  #2728·Apeksha9389 opened Oct 14, 2026·6 of 7 tasks·

  ·
  1

  6

• fix(voice): remove waitForOrAbort's abort listener once the race settles

  #2727·Apeksha9389 opened Oct 14, 2026·6 of 7 tasks·

  ·
  1

  1

• feat: add local fallback control to the audio turn detector

  review_effort:medium
  #2726·chenghao-mou opened Oct 14, 2026Member·4 of 6 tasks·

  ·
  1

• fix(soniox): send finalize when the STT stream is flushed

  #2725·MeerHusam opened Oct 14, 2026·

  ·
  1

  7

• fix(rime): use model-specific WebSocket endpoints

  #2723·rosetta-livekit-bot[bot] opened Oct 14, 2026Bot·

  ·
  1

• feat(inference): add language_hint to DeepgramFluxOptions

  #2722·rosetta-livekit-bot[bot] opened Oct 14, 2026Bot·

  ·
  1

• fix(soniox): avoid retaining frames through a shared abort promise

  #2721·RohSungKyun opened Oct 14, 2026·9 of 11 tasks·

  ·
  1

  2

• feat(inference): add openai/gpt-live-transcribe STT

  #2716·rosetta-livekit-bot[bot] opened Oct 13, 2026Bot·

  ·
  5

• fix(agents): keep timed transcript in step when alignments end mid-word

  #2714·denysdios opened Oct 13, 2026·6 of 7 tasks·

  ·
  5

• fix(voice): release tools still running when an interrupted speech times out

  #2712·rosetta-livekit-bot[bot] opened Oct 13, 2026Bot·

  ·
  4

• fix(voice): answer every tool return after ctx.update()

  #2711·rosetta-livekit-bot[bot] opened Oct 13, 2026Bot·

  ·
  5

• fix(agents): avoid retaining stream items on abort signal in readStream (#2658)

  #2710·Speedyzyzz opened Oct 13, 2026·

  ·
  1

  3

• feat: use blingfire as default sentence tokenizer

  #2709·davidzhao opened Oct 12, 2026Member·

  ·
  10

• fix(google): send user audio to Gemini only inside a manual user activity

  #2708·anzemur opened Oct 12, 2026Contributor·

  ·
  5

• fix(google): don't open a user activity for a server-initiated reply

  #2707·anzemur opened Oct 12, 2026Contributor·

  ·
  1

• fix(google): request one reply when generateReply ends a manual user activity

  #2706·anzemur opened Oct 12, 2026Contributor·

  ·
  3

• fix(google): keep images with their text in a user turn after interrupt()

  #2705·anzemur opened Oct 12, 2026Contributor·

  ·
  1

• fix(inworld): create TTS context with the first sentence

  #2704·ianbbqzy opened Oct 12, 2026Contributor·

  ·
  3

• feat(llm): report prompt-cache write tokens as cache_creation_tokens

  #2703·rosetta-livekit-bot[bot] opened Oct 12, 2026Bot·

  ·
  3

• feat(deepgram): make Flux audio chunk duration configurable

  #2702·rosetta-livekit-bot[bot] opened Oct 12, 2026Bot·

  ·
  3