Ollama Chat With MCP

🚀 Ollama Chat with MCP Integration

A powerful demonstration of integrating local large language models (LLMs) with real-time web search capabilities through the Model Context Protocol (MCP).

🚀 Quick Start

This project showcases how to extend the capabilities of local language models using tools. It combines the power of a locally run LLM provided by Ollama with real-time web search functionality via the MCP server.

Prerequisites

• Python 3.11+
• Installed and running Ollama
• A Serper.dev API key (free tier available)
• Internet connection for searches

Installation

1. Clone the repository:

   git clone https://github.com/redbuilding/ollama-chat-with-mcp.git
   cd ollama-chat-with-mcp
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Create a .env file in the project root and add your Serper.dev API key:

   SERPER_API_KEY=your_serper_api_key_here
   

4. Ensure Ollama is installed and the preset model is available (default is qwen2.5:14b):

   ollama pull qwen2.5:14b
   

Usage

Launch the web interface:

python chat_frontend.py

This will start the Gradio web interface, accessible at http://localhost:7860 by default.

Use the terminal client:

python chat_client.py

Search commands

In both interfaces, you can use special commands to trigger web searches:

• Search and summarize: #search for "Financial market outlook for April 2025"
• Search and answer questions: #search for "This week's reality TV shows" and what's been happening recently?

Other commands

• Clear conversation history: clear
• Display a list of available commands: help

✨ Features

• 🔎 Web-enhanced chat: Get real-time web search results during conversations.
• 🧠 Local model execution: Run models on your own hardware using Ollama.
• 🔌 MCP integration: Demonstrate the practical implementation of MCP.
• 🌐 Dual interfaces: Choose between a terminal CLI or a web GUI.
• 📊 Structured search results: Present web search data in a clear format for easy understanding.
• 🔄 Conversation memory: Maintain context throughout the chat session.

📚 Documentation

Overview

The project consists of three main components:

• MCP Web Search Server: Provides web search functionality using the Serper.dev API.
• Terminal client: A command-line interface (CLI) for chat and search interactions.
• Web frontend: A user-friendly web interface based on Gradio.

Through this architecture, the application demonstrates how MCP enables local models to access external tools and data sources, significantly enhancing their capabilities.

How it works

1. The MCP Web Search Server receives search requests from the client.
2. It uses the Serper.dev API to perform actual web searches.
3. The results are formatted and returned to the client for users to view and analyze.
4. The client further processes the results via Ollama to generate relevant responses.

File Structure

• MCP Web Search Server: Responsible for receiving and processing search requests.
• Terminal client: Provides a command-line interface to interact with MCP.
• Web frontend: A user-friendly interface based on Gradio for intuitive operation.

Customization

You can customize multiple aspects according to specific needs:

• Modify search scope and parameters.
• Replace web crawling tools.
• Customize output formats and styles.

🤝 Contribution

We welcome contributions from the community! Please refer to Contributing.md to learn how to get involved.

📄 License

The project is released under LICENSE. Please refer to the relevant file for detailed license information.