AI Restaurant Order Automation Assistant

This is an AI-powered conversational bot for ordering food from Zomato via Telegram. It uses Google's Gemini Flash model for natural language understanding and connects to Zomato via the Model Context Protocol (MCP).

Prerequisites

• Python 3.10+
• Telegram Bot Token
• Gemini API Key
• Zomato MCP Server installed/configured

Setup

1. Clone the repository (if you haven't already):

   git clone https://github.com/Gagan2004/Zomato_MCP_Bot.git
   cd Zomato_MCP_Bot

2. Create and activate a virtual environment:

   python -m venv venv
   # Windows
   venv\Scripts\activate
   # Linux/Mac
   source venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Configure Environment: Rename .env.example to .env (or create .env) and fill in the values:

   ZOMATO_MCP_COMMAND=npx.cmd          # Command to run Zomato MCP (Windows: npx.cmd, Mac/Linux: npx)
   ZOMATO_MCP_ARGS=-y mcp-remote https://mcp-server.zomato.com/mcp
   
   LLM_PROVIDER=openai 
   # if LLM_PROVIDER is not set to openai , ageny.py will look for GEMINI_API_KEY
   
   OPENAI_API_KEY=<your_openai_api_key>
   
   TELEGRAM_TOKEN=your_telegram_bot_token
   
   GEMINI_API_KEY=your_gemini_api_key
   

   Note: This configuration connects to the remote Zomato MCP server.

Running the Bot

1. Start the bot:

   python main.py

2. Interact: Open your Telegram bot and send /start. You can then ask:

   • "Show pizza places near [Location]"
   • "Show me the menu for [Restaurant Name]"
   • "Order 1 Margherita Pizza"
   • "Track my order"

Troubleshooting

• MCP Connection Failed: If the bot says "MCP Session not active", check your ZOMATO_MCP_COMMAND. You can verify execution by running python verify_mcp.py.

• Gemini Error: Ensure your API key is valid and has access to gemini-2.5-flash.

Project Structure

• main.py: Entry point, Telegram bot handler.
• agent.py: LangChain agent logic with tool calling and multi-LLM support.
• tools.py: Zomato MCP client wrapper with pagination support.
• config.py: Configuration loader.

Features

• Multi-LLM Support: Switch between Google Gemini and OpenAI (GPT-4o) using the LLM_PROVIDER env variable.
• Advanced Search: Fetch all restaurants with automated pagination and smart filtering.
• Robust Error Handling: Automatic retry for API limits and intelligent message splitting for long responses.

Configuration

Add these to your .env file for advanced features:

# Optional: Use OpenAI instead of Gemini
LLM_PROVIDER=openai
OPENAI_API_KEY=sk-...

# API Key Rotation (Optional)
GEMINI_API_KEY_2=...
GEMINI_API_KEY_3=...