Voice Assistant — Demo

Clínica Dra. Valeria — pick a demo to run against the local server.

Live Voice Widget current

Full voice loop: hold-to-talk → STT (Deepgram) → LLM (Gemini) → TTS (ElevenLabs) with streamed replies and per-sentence playback.

Audio Test Harness dev

Dev-only harness. Record a clip; the server transcodes it with FFmpeg and echoes it back. No LLM, no STT.

API · GET /health — liveness · GET /ws/live-voice — live voice WebSocket · GET /ws/audio-test — transcoder test WebSocket