Run Controls
Run the voice agent lab to simulate wake word detection, transcript assembly, agent task execution, and TTS handoff.
Transcript
No voice turn yet.
Assistant Reply
No reply emitted yet.
AI WebGPU Lab App
Voice Agent Lab Demo
`app-voice-agent-lab` exposes an integrated local voice-to-agent surface. It runs one deterministic wake-word, transcript, task-routing, and TTS handoff flow in a single app-shaped result.
This is the first internal demo surface that keeps voice roundtrip and browser-agent task metrics in the same raw result before real local provider wiring lands.
Task Deck
Metrics
Environment
Activity Log
Fixture Metadata
Loading fixture...
Schema-Aligned Result Draft
{
"status": "pending"
}
Demo Notes
- Voice turn, task routing, and response handoff are fixed to one local-only fixture.
- Roundtrip and task metrics stay together so the app can inherit both voice and agent regression contracts.
- The same result schema can later absorb real STT, planner, DOM policy, and TTS providers.