All pull requests

Pull requests

• Add LFM3-Audio speech-to-speech mode (no STT, no TTS)

  #766·jackboyla opened Oct 28, 2026Contributor·

  ·

• Skip redundant Transformers streamer draining after normal completion

  #763·jackboyla opened Oct 28, 2026Contributor·

  ·
  1

• Share STT language detection lists while preserving Orukeet Greek coverage

  #761·salignatmoandal opened Oct 26, 2026Contributor·5 tasks done·

  ·

• Share TTS response latency timing through a base TTS handler

  #759·omkar-334 opened Oct 26, 2026·

  ·
  1

• Deliver Kitten audio chunks promptly and stop synthesis on cancellation

  #756·Deep-unlearning opened Oct 23, 2026Collaborator·

  ·

• Add Granite Speech TurboCTC recognition backend

  #755·Deep-unlearning opened Oct 23, 2026Collaborator·

  ·

• Add Moonshine speech recognition backend

  #754·Deep-unlearning opened Oct 23, 2026Collaborator·

  ·

• Add a ConvFill LLM backend example

  #750·andimarafioti opened Oct 22, 2026Member·

  ·

• Report cumulative LLM time and rounds for tool-assisted replies

  #748·jackboyla opened Oct 21, 2026Contributor·

  ·
  1

• Expose speech language capabilities before session setup

  #744·jackboyla opened Oct 21, 2026Contributor·

  ·
  1

• fix: resolve the diarization and viseme stages onto an available NPU

  #741·li-lizhe opened Oct 20, 2026Contributor·

  ·
  1

• Reject ~name config paths on Windows too

  #736·Hotragn opened Oct 19, 2026Contributor·

  ·

• Keep combining marks in remove_unspeechable

  #733·Jackmeson1 opened Oct 18, 2026·

  ·
  1

• Fall back instead of emitting no audio when Kokoro gets an unknown voice

  #731·jeffjhunter opened Oct 17, 2026·

  ·
  4

• Fix session tools being dropped when a response carries an empty tools list

  #730·jeffjhunter opened Oct 17, 2026·

  ·

• Record Supertonic first-audio latency on the turn tracker

  #727·salignatmoandal opened Oct 16, 2026Contributor·1 of 2 tasks·

  ·

• Expose VAD threshold and barge-in controls in the packaged audio client

  #722·modelpath-dev opened Oct 15, 2026·

  ·
  1

  2

• [#681] feat: serve named pipelines concurrently

  #715·tavallaie opened Oct 12, 2026Contributor·

  ·
  1

• Add optional Open WebUI local integration example

  #691·andimarafioti opened Oct 9, 2026Member·

  ·
  1

• Accept rocm as a device and fall back where a backend is CUDA only

  #685·LaelaZorana opened Oct 8, 2026·

  ·
  1

• Share OpenAI-compatible LLM clients with explicit runtime cleanup

  #673·tavallaie opened Oct 7, 2026Contributor·

  ·
  1

  1

• Cancel speaker echo in the packaged audio client

  #671·jackboyla opened Oct 7, 2026Contributor·

  ·
  39

• Separate Realtime language coverage from session hint support

  #670·Deep-unlearning opened Oct 7, 2026Collaborator·

  ·
  1

• Serialize spoken text and its tool call as one assistant turn

  #664·coffee-the-dev opened Oct 6, 2026·

  ·
  1

• Open a turn at 192 ms, keep 384 ms for barge-in

  #599·ServeurpersoCom opened Sep 25, 2026Contributor·

  ·
  1