Agentic_search

File Search Assistant with LLM Integration

This project combines semantic search capabilities with Large Language Models (LLM) to provide intelligent file search functionality in Linux systems.

Features

• Natural language file search queries
• Semantic search using BERT embeddings
• File system integration with MCP server
• Gemini LLM integration for query understanding
• Automatic file extension inference

Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (optional, for faster processing)
• Linux operating system

Installation

1. Clone the repository:

git clone <repository-url>
cd LLm_To_agent/data_dir

2. Install dependencies:

pip install -r requirements.txt

3. Configure environment variables:
   • Create a .env file in the project root
   • Add your Gemini API key:

   GEMINI_API_KEY="your-api-key-here"
   

Project Structure

data_dir/
├── main.py              # Main application entry point
├── tools/
│   └── file_finder.py   # File search implementation
├── utils/
│   └── doc_Search.py    # Document search utilities
├── llm/
│   └── gemini_client.py # LLM integration
├── .env                 # Environment variables
└── OSData_store.pth     # Embedded data storage

Usage

1. Start the application:

python main.py

2. Enter your search query when prompted:

What file are you looking for?
> "find a python script that handles file operations"

3. The system will:
   • Extract the relevant filename using LLM
   • Search the file system using semantic search
   • Display matching file locations

Features in Detail

• LLM Integration: Uses Google's Gemini for natural language understanding
• Semantic Search: Employs BERT embeddings for context-aware file matching
• MCP Server: Handles file system operations efficiently
• Extensible Architecture: Easy to add new search capabilities

Troubleshooting

If you encounter the error "Embeddings file not found", run:

cd ~ && find / -type f 2>/dev/null >> Desktop/LLm_To_agent

License

[Your chosen license]

Contributing

[Your contribution guidelines]