rushiwatpalwad23/AI-Chatbot-MCP

AI Tool Calling Application

Ollama + Qwen3 + MCP Architecture

A pure Python AI application that uses Ollama with Qwen3 model and Model Context Protocol (MCP) for tool calling.

Features

• 🤖 Ollama + Qwen3: Local AI model (no OpenAI needed)
• 🔧 MCP Tools: Calculator and Weather tools
• 🏠 100% Local: No external API dependencies
• 🚀 Fast Setup: Automated installation script

Quick Start

1. Setup (Run Once)

```bash python setup.py ```

2. Run Application (3 Terminals)

Terminal 1 - Ollama: ```bash ollama serve ```

Terminal 2 - MCP Server: ```bash python mcp_server/server.py ```

Terminal 3 - Main App: ```bash python app.py ```

Test Queries

• What is 25 + 17? (Calculator tool)
• Tell me the temperature in Pune (Weather tool)
• Hello! (Direct response, no tool)

Project Structure

``` ├── app.py # Main application ├── mcp_client/ # MCP client package │ ├── init.py │ └── client.py ├── mcp_server/ # MCP server package │ ├── init.py │ ├── server.py # FastAPI server │ └── tools.py # Tool implementations ├── config/ │ ├── config.json # Configuration │ └── modelfile # Ollama model config ├── setup.py # Setup script ├── requirements.txt # Dependencies └── README.md ```

How It Works

1. User Query → Main app (app.py)
2. Tool Decision → Ollama + Qwen3 decides if tool needed
3. Tool Execution → MCP Client → MCP Server → Tool
4. Final Answer → Qwen3 generates response using tool result

Requirements

• Python 3.8+
• Ollama installed and running
• Internet connection (for initial model download)

Troubleshooting

• Ollama not found: Install from https://ollama.ai
• Model not found: Run python setup.py again
• Connection refused: Make sure all 3 terminals are running