RAG PDF and Document Retrieval Framework

This repository provides a comprehensive framework for extracting and retrieving information from PDFs and text documents. It combines OCR capabilities with state-of-the-art language models to process text or image-based PDFs, create semantic embeddings, and retrieve information efficiently using a variety of advanced techniques.

---

Features

1. PDF Text and Image Extraction

• Utilizes MiniCPM-Llama3-V-2_5, an advanced model from Hugging Face, for OCR-based text extraction from image-based PDFs.
• Provides simple text extraction for text-based PDFs using the PyPDF2 library.

Model Reference: MiniCPM-Llama3-V-2_5 on Hugging Face

2. RAG (Retrieve, Augment, Generate) Workflow

• Built on LangChain, a robust framework for building information retrieval systems.
• Employs Ollama as a chat and embedding model for advanced natural language interaction and retrieval.

LangChain Reference: LangChain Documentation
Ollama installation guide: Ollama Installation and Usage Guide

Ollama available models: Ollama Installation and Usage Guide

3. Multi-Vector and Semantic Retrieval

• Supports semantic chunking of text and multi-vector retrieval for enhanced accuracy.
• Leverages Chroma and FAISS for vector storage and similarity search.

4. Query Answering

• main.py Allows users to pass a question (query) along with the pdf document path to retrieve answers based on context.

---

Installation

1. Clone the repository:

   git clone https://github.com/balbakri1/SimpleRAG.git
   cd rag-retrieval