MCP

🚀 Jeneen's MCP Agent

This project is a multi-functional AI agent built with FastMCP, offering an Arabic legal chatbot, a Google search tool, and a VAE model for generating handwritten digit images.

🚀 Quick Start

Environment Setup

Create and activate a virtual environment:

python -m venv .venv
source .venv/bin/activate    # On Windows: .venv\Scripts\activate

Install Dependencies

pip install -r requirements.txt

Train or Use the Pre-trained Model

• Optional: Train the VAE model using VAE.ipynb.
• Or use the pre-trained model in output/vae_epoch_50.pth.

Run the MCP Agent

python main.py

✨ Features

• ✅ An Arabic legal chatbot that answers common legal questions.
• 🔍 A Google search tool.
• 🧬 A Variational Autoencoder (VAE) model that generates handwritten digit images.

📦 Installation

The installation steps are included in the Quick Start section above.

💻 Usage Examples

1. legal_chat(query: str) → str

This is an Arabic-language chatbot that responds to legal questions such as annual leave, divorce, custody, employment rights, and rental agreements. Example:

{ "tool": "legal_chat", "input": "ما هي حقوقي في حال الطلاق؟" }

2. search_google(query: str) → str

This tool opens a Google search in the default browser. Example:

{ "tool": "search_google", "input": "قانون العمل الأردني" }

3. vae_generate(n_images: int) → str

This function generates handwritten digit images using a trained VAE model and returns a base64-encoded PNG image. Example:

{ "tool": "vae_generate", "input": { "n_images": 8 } }

📚 Documentation

📁 Project Structure

├── main.py # Main script to run the MCP agent
├── chatbot.py # Arabic legal chatbot logic
├── vae_model.py # VAE model definitions (Encoder, Decoder, VAE)
├── output/ # Model checkpoints and generated images
├── data/ # MNIST dataset (auto-downloaded)
├── VAE.ipynb # Jupyter notebook for training the VAE model
└── README.md # This documentation file