Search Engine with RAG and MCP

A powerful search engine that combines LangChain, Model Context Protocol (MCP), Retrieval-Augmented Generation (RAG), and Ollama to create an agentic AI system capable of searching the web, retrieving information, and providing relevant answers.

Features

• Web search capabilities using the Exa API

• Web content retrieval using FireCrawl

• RAG (Retrieval-Augmented Generation) for more relevant information extraction

• MCP (Model Context Protocol) server for standardized tool invocation

• Support for both local LLMs via Ollama and cloud-based LLMs via OpenAI

• Flexible architecture supporting direct search, agent-based search, or server mode

• Comprehensive error handling and graceful fallbacks

• Python 3.13+ with type hints

• Asynchronous processing for efficient web operations

Related MCP server: Xiaozhi MCP Aggregate

Architecture

This project integrates several key components:

1. Search Module: Uses Exa API to search the web and FireCrawl to retrieve content

2. RAG Module: Embeds documents, chunks them, and stores them in a FAISS vector store

3. MCP Server: Provides a standardized protocol for tool invocation

4. Agent: LangChain-based agent that uses the search and RAG capabilities

Project Structure

search-engine-with-rag-and-mcp/
├── LICENSE              # MIT License
├── README.md            # Project documentation
├── data/                # Data directories
├── docs/                # Documentation
│   └── env_template.md  # Environment variables documentation
├── logs/                # Log files directory (auto-created)
├── src/                 # Main package (source code)
│   ├── __init__.py      
│   ├── core/            # Core functionality
│   │   ├── __init__.py
│   │   ├── main.py      # Main entry point
│   │   ├── search.py    # Web search module
│   │   ├── rag.py       # RAG implementation
│   │   ├── agent.py     # LangChain agent
│   │   └── mcp_server.py # MCP server implementation
│   └── utils/           # Utility modules
│       ├── __init__.py
│       ├── env.py       # Environment variable loading
│       └── logger.py    # Logging configuration
├── pyproject.toml       # Poetry configuration
├── requirements.txt     # Project dependencies
└── tests/               # Test directory

Getting Started

Prerequisites

• Python 3.13+

• Poetry (optional, for development)

• API keys for Exa and FireCrawl

• (Optional) Ollama installed locally

• (Optional) OpenAI API key

Installation

1. Clone the repository

git clone https://github.com/yourusername/search-engine-with-rag-and-mcp.git
cd search-engine-with-rag-and-mcp

2. Install dependencies

# Using pip
pip install -r requirements.txt

# Or using poetry
poetry install

3. Create a .env file (use docs/env_template.md as a reference)

Usage

The application has three main modes of operation:

1. Direct Search Mode (Default)

# Using pip
python -m src.core.main "your search query"

# Or using poetry
poetry run python -m src.core.main "your search query"

2. Agent Mode

python -m src.core.main --agent "your search query"

3. MCP Server Mode

python -m src.core.main --server

You can also specify custom host and port:

python -m src.core.main --server --host 0.0.0.0 --port 8080

Using Ollama (Optional)

To use Ollama for local embeddings and LLM capabilities:

1. Install Ollama: https://ollama.ai/

2. Pull a model:

ollama pull mistral:latest

3. Set the appropriate environment variables in your .env file:

OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=mistral:latest

Development

This project follows these best practices:

• Code formatting: Black and isort for consistent code style

• Type checking: mypy for static type checking

• Linting: flake8 for code quality

• Testing: pytest for unit and integration tests

• Environment Management: python-dotenv for managing environment variables

• Logging: Structured logging to both console and file

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgements

• LangChain for the agent framework

• Model Context Protocol for the standardized tool invocation

• Ollama for local LLM capabilities

• Exa for web search capabilities

• FireCrawl for web content retrieval