🤖 Consent Agent

How it works

This app turns your Reachy Mini into a voice-interactive agent powered entirely by local hardware.

1. Reachy listens continuously via the built-in microphone
2. Speech is detected using voice activity detection
3. Audio is transcribed locally with Whisper
4. A response is generated with Qwen3-14B via vLLM
5. Speech is synthesized with Piper TTS
6. Reachy speaks the response through the built-in speaker

Requirements

• HP ZGX Nano (or NVIDIA GPU with 24GB+ VRAM)
• ZGX AI API Docker container running on the GPU
• Reachy Mini (Wireless) on the same network

Configuration

Set the ZGX_API_URL environment variable on the robot if the API is not at http://localhost:8090.