Reachy Mini Interactive Agent

A conversational AI agent for Reachy Mini that uses OpenAI's Realtime API for real-time voice interaction.

Requirements

• Python 3.10+
• Reachy Mini connected via USB-C
• uv package manager

Setup

1. Install dependencies:

   uv sync

2. Set environment variables:

   export OPENAI_API_KEY="your-openai-api-key"
   export HONCHO_API_KEY="your-honcho-api-key"

   Optional:

   export HONCHO_WORKSPACE_NAME="reachy-mini"  # default

3. Connect your Reachy Mini via USB-C.

Run

uv run python main.py

Press Ctrl+C to stop.

How It Works

• Voice: Streams audio continuously to OpenAI; the model decides when to respond
• Vision: Detects faces for gaze tracking
• Memory: Conversations saved to Honcho in real-time
• Interruption: Stops speaking immediately when you start talking

Project Structure

reachy_agent/
├── agent.py           # Main orchestrator
├── audio.py           # Microphone/speaker handling
├── config.py          # Configuration and system prompt
├── openai_realtime.py # OpenAI Realtime API client
├── memory.py          # Honcho integration
├── robot.py           # Reachy Mini control
└── vision.py          # Camera and face detection

Configuration

All configuration is via environment variables. See config.py for defaults.