RAG Book Assistant

A simple Retrieval-Augmented Generation (RAG) project that lets you upload a PDF, create a vector database, and ask questions using the document content.

Features

• Upload a PDF through a Streamlit interface
• Split the document into chunks and store embeddings in Chroma
• Retrieve relevant context with MMR search
• Ask questions and get AI-powered answers based on the uploaded document

Project Structure

• app.py - Main Streamlit application for uploading PDFs and asking questions
• main.py - Command-line RAG demo using the existing Chroma database
• create_database.py - Script to generate the vector database from a PDF
• documents/ - Sample document files and loaders
• retrievers/ - Example retriever implementations
• vector store/ - Additional Chroma-related experiments

Tech Stack

• Python 3.12
• Streamlit
• LangChain
• Chroma DB
• Mistral AI and Google Gemini embeddings
• PyPDF / LangChain PDF loader

Prerequisites

1. Python 3.10+
2. A virtual environment
3. API keys for the AI providers used in your setup

Setup

1. Create and activate a virtual environment bash python -m venv .venv .\.venv\Scripts\Activate.ps1
2. Install dependencies bash pip install -r requirements.txt
3. Create a .env file using .env.example and add your API keys.

Run the App

Start the Streamlit interface:

bash python -m streamlit run app.py

Generate a Vector Database

If you want to create or refresh the Chroma database from a PDF:

bash python create_database.py

Notes

• The app uses the chroma_db/ directory to persist embeddings.
• Make sure your environment variables are available before running the app.
• The current setup is intended for local development and experimentation.