README.md

---
title: Reachy Mini Open Conversation
emoji: 🤖
colorFrom: indigo
colorTo: purple
sdk: docker
pinned: false
short_description: Voice conversation with Reachy Mini using open-source AI
---

# 🤖 Reachy Mini Open Conversation

A voice conversation app powered by fully open-source models:

- **STT**: [faster-whisper](https://github.com/SYSTRAN/faster-whisper) — fast speech-to-text
- **LLM**: [Ollama](https://ollama.com/) — local LLM inference (llama3.2 by default)
- **TTS**: [edge-tts](https://github.com/rany2/edge-tts) — high-quality text-to-speech

## Setup

Set these environment variables (as Space secrets for HF Spaces):

| Variable | Default | Description |
|---|---|---|
| `OLLAMA_BASE_URL` | `http://localhost:11434` | URL of your Ollama server |
| `MODEL_NAME` | `llama3.2` | Ollama model to use |
| `STT_MODEL` | `base` | faster-whisper model size (tiny/base/small/medium/large-v3) |
| `TTS_VOICE` | `en-US-AriaNeural` | edge-tts voice name |

> **Note**: You need a running Ollama server accessible from the Space. Set `OLLAMA_BASE_URL` to point to your remote Ollama instance.