This repository implements a Retrieval-Augmented Generation (RAG) service that allows users to:

• Index documents (e.g., PDFs, DOCX, Markdown, TXT).
• Query the indexed documents using FAISS for similarity search.
• Automatically watch a directory for new documents and index them.

The service supports both command-line and HTTP API interactions using Flask.

• Add documents in various formats (PDF, DOCX, MD, TXT).
• Query documents using similarity search and get a response from an LLM.
• Watch a directory for new documents.

pip install -r requirements.txt

You can run the service in two modes:

• CLI Mode: Watch a directory for new documents and index them.
• API Mode: Start an HTTP API to interact with the service.

python main.py --model_name "intfloat/e5-large-v2" --watch_directory "/path/to/documents" --mode cli

python main.py --mode server

POST /add_document
Content-Type: application/json

{
    "document": "This is the content of the document to be indexed."
}

POST /query
Content-Type: application/json

{
    "query": "What is the content of the document?"
}

• You can modify the default models for embeddings and the RAG query handler.
• Supported document formats include .pdf, .md, and .txt.

This project is licensed under the MIT License.