feat(voice): pluggable voice backend with Gemini Live & Qwen Realtime#401

Open

Overbaker wants to merge 21 commits into

tiann:mainfrom

Overbaker:feat/pluggable-voice-backend

Commits on Apr 21, 2026

feat(voice): pluggable voice backend with Gemini Live API support
Overbaker
committed
feat(voice): switch default language to Chinese
Overbaker
committed
fix(voice): use inline Blob URL for AudioWorklet instead of ?url import
Overbaker
committed
fix(voice): switch to gemini-2.5-flash-native-audio-latest model
Overbaker
committed
feat(voice): add Qwen Realtime as third voice backend option
Overbaker
committed
fix(pwa): add skipWaiting + clientsClaim for immediate SW activation
Overbaker
committed
fix(voice): Qwen WebSocket proxy + switch to qwen3-omni-flash-realtime
Overbaker
committed
fix(voice): switch default TTS to qwen-realtime + increase socket buffer
Overbaker
committed
fix(voice): fix Gemini Live barge-in and tool call issues
Overbaker
committed
fix(voice): greeting via system prompt to preserve tool call ability
Overbaker
committed
fix(voice): address PR review — add JWT auth to WS proxy, stop leaking API keys
Overbaker
committed
fix(voice): gate voice button on backend discovery readiness
Overbaker
committed
fix(voice): fix Qwen URL duplication, defer ready until session registered, update tests
Overbaker
committed
fix(voice): use request origin for proxy URL, connect worklet to sink, serialize tool calls
Overbaker
committed
fix(voice): create playback AudioContext in user gesture for mobile autoplay
Overbaker
committed
fix(voice): store playback AudioContext in state and close on cleanup
Overbaker
committed
fix(pwa,voice): remove forced SW activation, guard voice debug logs
Overbaker
committed
fix(voice): reject startup promise on early WS close, fix debug flag
Overbaker
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat(voice): pluggable voice backend with Gemini Live & Qwen Realtime#401

feat(voice): pluggable voice backend with Gemini Live & Qwen Realtime#401
Overbaker wants to merge 21 commits into
tiann:mainfrom
Overbaker:feat/pluggable-voice-backend

Commits on Apr 21, 2026

feat(voice): pluggable voice backend with Gemini Live API support

feat(voice): switch default language to Chinese

fix(voice): use inline Blob URL for AudioWorklet instead of ?url import

fix(voice): switch to gemini-2.5-flash-native-audio-latest model

feat(voice): add Qwen Realtime as third voice backend option

fix(pwa): add skipWaiting + clientsClaim for immediate SW activation

fix(voice): Qwen WebSocket proxy + switch to qwen3-omni-flash-realtime

fix(voice): switch default TTS to qwen-realtime + increase socket buffer

fix(voice): fix Gemini Live barge-in and tool call issues

fix(voice): greeting via system prompt to preserve tool call ability

fix(voice): address PR review — add JWT auth to WS proxy, stop leaking API keys

fix(voice): gate voice button on backend discovery readiness

fix(voice): fix Qwen URL duplication, defer ready until session registered, update tests

fix(voice): use request origin for proxy URL, connect worklet to sink, serialize tool calls

fix(voice): create playback AudioContext in user gesture for mobile autoplay

fix(voice): store playback AudioContext in state and close on cleanup

fix(pwa,voice): remove forced SW activation, guard voice debug logs

fix(voice): reject startup promise on early WS close, fix debug flag

Commits on Apr 22, 2026

fix(voice): queue Gemini proxy messages during connect, fix Qwen startup hang

fix(voice): pin browser wsUrl to proxy, separate ElevenLabs language from Gemini/Qwen

fix(voice): apply initial mic mute state after recorder starts