All pull requests

Pull requests

• feat(eot): feed the local mini model its 2.0 s window and turn length

  review_effort:high
  #2850·chenghao-mou opened Oct 24, 2026Member·6 of 8 tasks·

  ·
  1

• fix: preserve hyphenation exception boundaries

  #2849·Apeksha9389 opened Oct 24, 2026Contributor·6 of 7 tasks·

  ·
  5

• chore(deps): update dependency zod to v4.6.6 [security]

  #2848·renovate[bot] opened Oct 24, 2026Bot·1 task·

  ·
  3

• feat(beta): add TranslationSession for speech-to-speech translation

  #2845·rosetta-livekit-bot[bot] opened Oct 23, 2026Bot·

  ·
  8

• (wip) hosted logs handoff

  #2844·andrewnitu opened Oct 23, 2026Contributor·8 tasks·

  ·
  1

• feat(inference): claim aligned transcripts for AssemblyAI TTS

  #2843·rosetta-livekit-bot[bot] opened Oct 23, 2026Bot·1 task·

  ·
  3

• (wip) otel attributes to stdout + add agent_name

  #2842·andrewnitu opened Oct 23, 2026Contributor·6 of 8 tasks·

  ·
  1

• fix(assemblyai): make redaction and partial-turn options constructor-only

  #2840·juzer-pam opened Oct 23, 2026Contributor·6 of 8 tasks·

  ·
  5

• fix(google): default the LLM to gemini-3.8-flash

  #2839·rosetta-livekit-bot[bot] opened Oct 23, 2026Bot·

  ·
  1

• fix(voice): resolve replace keys split across chunks to the longest match

  #2836·swayamg20 opened Oct 23, 2026Contributor·

  ·
  1

• fix(voice): pace unspaced Chinese, Japanese and Thai transcripts to the audio

  #2835·swayamg20 opened Oct 23, 2026Contributor·

  ·
  1

  4

• fix(agents): Python port - expose participant disconnect details

  #2834·paulheinrichs-jb opened Oct 23, 2026Contributor·1 of 8 tasks·

  ·
  8

• fix(voice): keep the synchronized transcript when playback finishes after clearBuffer

  #2833·swayamg20 opened Oct 23, 2026Contributor·

  ·
  1

  1

• fix(ipc): stop text turn detection from waiting on a dead inference process

  #2832·swayamg20 opened Oct 23, 2026Contributor·

  ·
  1

  3

• fix(tts): keep StreamPacer buffered tail after end_input

  #2831·rosetta-livekit-bot[bot] opened Oct 23, 2026Bot·

  ·
  6

• fix(anthropic): send parallel_tool_calls when no tool_choice is set

  #2830·rosetta-livekit-bot[bot] opened Oct 23, 2026Bot·

  ·
  1

• fix(deepgram): Drop rejected language hints, add nova-3 multi, and fix the nova-3-general fallback

  #2828·adrian-cowham opened Oct 22, 2026Contributor·

  ·
  11

• chore(deps): update dependency vite to v7.3.7 [security]

  #2827·renovate[bot] opened Oct 22, 2026Bot·1 task·

  ·
  1

• feat(inference): Type and gate language_hint for Deepgram Nova-3

  #2826·adrian-cowham opened Oct 22, 2026Contributor·

  ·
  9

• telemetry: send a session summary in the recording header

  #2824·jmcclanahan opened Oct 22, 2026Contributor·7 of 9 tasks·

  ·
  12

• fix(deepgram): advance startTimeOffset when the Nova STT stream reopens its socket

  #2823·praveen4star opened Oct 22, 2026Contributor·6 of 8 tasks·

  ·
  10

• feat(anam): use lk.playback_started to track turn state

  #2821·rosetta-livekit-bot[bot] opened Oct 22, 2026Bot·

  ·
  1

• fix(ipc): attach the supervised process 'error' listener before initialize()

  #2820·mohhh-ok opened Oct 22, 2026·6 of 8 tasks·

  ·
  1

  2

• fix(voice): end the call after the turn when no reply is asked for

  #2818·rosetta-livekit-bot[bot] opened Oct 22, 2026Bot·

  ·
  4

• fix(voice): snapshot final text before async flush in transcription output

  #2816·aniketkrs opened Oct 22, 2026·

  ·
  1

  6