MoA Chat - Documentation

MoA Chat is an experimental project inspired by the MoA (Mixture of Agents) technique where multiple AI models collaborate to produce better answers.

🚀 Features

Connects multiple LLMs simultaneously
Uses a fourth LLM as an aggregator
Fully open source and self-hostable
Built using Python and Flask
Deployed easily on Hugging Face Spaces

⚙️ How to Self-Host

To clone this project and run it manually:

git clone https://huggingface.co/spaces/UntilDot/Flask
cd Flask
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py

If you wish to use Docker (advanced users):

docker build -t moa-chat .
docker run -d -p 7860:7860 --env-file .env moa-chat

🔑 API Keys

The system uses secrets for security. You must define these keys in your Hugging Face Space settings or your .env file:

OPENROUTER_API_KEY=your-openrouter-key
GROK_API_KEY=your-grok-key
TOGETHER_API_KEY=your-together-key
GROQ_API_KEY=your-groq-key

🛠️ Adding More Models

To add new models, edit llm/model_config.json like this:

{
  "providers": {
    "openrouter": {
      "url": "https://openrouter.ai/api/v1/chat/completions",
      "key_env": "OPENROUTER_API_KEY"
    }
  },
  "models": {
    "deepseek/deepseek-chat-v3-0324:free": "openrouter"
  }
}

Each model must point to a provider. Each provider defines its own endpoint and required secret.

📖 License

This project is licensed under Apache 2.0.

MoA Chat - Documentación

MoA Chat es un proyecto experimental inspirado en la técnica MoA (Mixture of Agents), donde varios modelos de IA colaboran para generar mejores respuestas.

🚀 Características

Conecta múltiples LLMs simultáneamente
Utiliza un cuarto LLM como agregador
Completamente open source y autohospedable
Construido con Python y Flask
Desplegado fácilmente en Hugging Face Spaces

⚙️ Cómo Autohospedar

Para clonar y ejecutar este proyecto manualmente:

git clone https://huggingface.co/spaces/UntilDot/Flask
cd Flask
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python app.py

O si deseas usar Docker (usuarios avanzados):

docker build -t moa-chat .
docker run -d -p 7860:7860 --env-file .env moa-chat

🔑 Llaves API

Debes definir las llaves API en la configuración del espacio o en tu archivo .env:

OPENROUTER_API_KEY=tu-clave-openrouter
GROK_API_KEY=tu-clave-grok
TOGETHER_API_KEY=tu-clave-together
GROQ_API_KEY=tu-clave-groq

🛠️ Agregar Modelos

Para agregar modelos, edita llm/model_config.json así:

{
  "providers": {
    "openrouter": {
      "url": "https://openrouter.ai/api/v1/chat/completions",
      "key_env": "OPENROUTER_API_KEY"
    }
  },
  "models": {
    "deepseek/deepseek-chat-v3-0324:free": "openrouter"
  }
}

Cada modelo debe apuntar a un proveedor que define el endpoint y la clave secreta necesaria.

📖 Licencia

Este proyecto está licenciado bajo Apache 2.0.