Add PCM voice pipeline: server-side STT/TTS, Smart Turn, web UI voice mode#971

Closed

pbranchu wants to merge 24 commits intoRightNow-AI:mainfrom

pbranchu:feat/voice-pcm

Commits on Apr 4, 2026

Add voice channel adapter with WebSocket server

Philippe Branchu
and
claude
committed
Add per-channel system prompt injection via function parameters

Philippe Branchu
and
claude
committed
Fix cargo fmt formatting

Philippe Branchu
and
claude
committed
Add voice pipeline: PCM mode, Smart Turn ONNX, STT/TTS providers, web UI voice mode

Philippe Branchu
and
claude
committed
Fix Smart Turn: catch ort panic when libonnxruntime.so is missing

Philippe Branchu
and
claude
committed
Fix Smart Turn startup hang: spawn_blocking + timeout fallback

Philippe Branchu
and
claude
committed
Route voice WebSocket through main API port; fix UI button icons

Philippe Branchu
and
claude
committed
Fix voice UI: move call button to header, add Permissions-Policy for mic

Philippe Branchu
and
claude
committed
Add voice identity handshake and optional speaker diarization

Philippe Branchu
and
claude
committed
Add voice barge-in: interrupt agent speech on loud mic input

Philippe Branchu
and
claude
committed
Fix CSP: allow blob: in script-src for AudioWorklet

Philippe Branchu
and
claude
committed
Fix CSP violations: blob: for AudioWorklet, nonce inline script, fix voiceTranscript Alpine expression

Philippe Branchu
and
claude
committed
Fix silence fallback: use sleep_until(absolute) instead of sleep(duration)

Philippe Branchu
and
claude
committed
Fix build: restore rmcp-based mcp.rs, fix non-exhaustive struct ctor

Philippe Branchu
and
claude
committed
Fix PCM mode: Hello frame before binary was routing sessions to text mode

Philippe Branchu
and
claude
committed
Fix voice: spurious dispatch, multiple sessions, missing transcript

Philippe Branchu
and
claude
committed
Fix voice transcript: show clean text in chat, use Alpine-safe mutations

Philippe Branchu
and
claude
committed
Fix voice: discard stale responses when user speaks during thinking

Philippe Branchu
and
claude
committed
Voice: inject discarded response as context when user interrupts thinking

Philippe Branchu
and
claude
committed
cargo fmt

Philippe Branchu
and
claude
committed
Fix voice config: use api_key_env instead of literal api_key

Philippe Branchu
and
claude
committed
Voice UI: gate phone button on server STT+TTS capability

Philippe Branchu
and
claude
committed
Remove personal name from doc comment example

Philippe Branchu
and
claude
committed
Fix Clippy warnings in voice and tts

Philippe Branchu
and
claude
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add PCM voice pipeline: server-side STT/TTS, Smart Turn, web UI voice mode#971

Add PCM voice pipeline: server-side STT/TTS, Smart Turn, web UI voice mode#971
pbranchu wants to merge 24 commits intoRightNow-AI:mainfrom
pbranchu:feat/voice-pcm

Commits on Apr 4, 2026

Add voice channel adapter with WebSocket server

Add per-channel system prompt injection via function parameters

Fix cargo fmt formatting

Add voice pipeline: PCM mode, Smart Turn ONNX, STT/TTS providers, web UI voice mode

Fix Smart Turn: catch ort panic when libonnxruntime.so is missing

Fix Smart Turn startup hang: spawn_blocking + timeout fallback

Route voice WebSocket through main API port; fix UI button icons

Fix voice UI: move call button to header, add Permissions-Policy for mic

Add voice identity handshake and optional speaker diarization

Add voice barge-in: interrupt agent speech on loud mic input

Fix CSP: allow blob: in script-src for AudioWorklet

Fix CSP violations: blob: for AudioWorklet, nonce inline script, fix voiceTranscript Alpine expression

Fix silence fallback: use sleep_until(absolute) instead of sleep(duration)

Fix build: restore rmcp-based mcp.rs, fix non-exhaustive struct ctor

Fix PCM mode: Hello frame before binary was routing sessions to text mode

Fix voice: spurious dispatch, multiple sessions, missing transcript

Fix voice transcript: show clean text in chat, use Alpine-safe mutations

Fix voice: discard stale responses when user speaks during thinking

Voice: inject discarded response as context when user interrupts thinking

cargo fmt

Fix voice config: use api_key_env instead of literal api_key

Voice UI: gate phone button on server STT+TTS capability

Remove personal name from doc comment example

Fix Clippy warnings in voice and tts