Voice Labs¶

The first vertical slice is a browser conversation through Platform, LiveKit, Runtime, VAD/turn detection, STT, LLM, and TTS, with durable lifecycle events written back to PostgreSQL. Provider-backed behavior is considered verified only after configured real credentials and real audio have been used.

The local workspace consists of backend.voice-labs (Platform, Runtime, and Evals), frontend.voice-labs (web), and this stack repository (shared infrastructure, configuration, documentation, and acceptance checks). See the root README for setup.