My First RAG - PDF Question Answering System

A simple Retrieval-Augmented Generation (RAG) pipeline that lets you ask questions about PDF documents and get AI-generated answers.

Built as a learning project to understand how RAG works — from PDF extraction to vector search to LLM-powered answers.

Features

• Extract text from PDF documents
• Clean and chunk text using a sliding window approach
• Generate embeddings using SentenceTransformers
• Build a FAISS vector index for fast similarity search
• Query your documents and get answers using Ollama (Mistral model)
• Bonus: FastAPI experiments for REST API learning

Tech Stack

Technology	Purpose
Python	Core language
pypdf	PDF text extraction
SentenceTransformers	Text embeddings (all-MiniLM-L6-v2)
FAISS	Vector similarity search
NumPy	Array operations
Ollama (Mistral)	Local LLM for generating answers
FastAPI	REST API experiments

Folder Structure

my1strag/
??? data/                       # Source PDF documents
?   ??? HR_Policy_Document.pdf
?   ??? IT_Security_Policy.pdf
?   ??? Project_Guidelines.pdf
??? fastapi_experiments/        # FastAPI learning experiments
?   ??? order_api.py            # Simple order management API
?   ??? crud_api.py             # SQL Server CRUD API
??? chunk_pdf.py                # Step 1: Ingest PDFs ? chunks ? FAISS index
??? query_rag.py                # Step 2: Ask questions ? get AI answers
??? pdf_reader_demo.py          # Standalone PDF reading experiment
??? requirements.txt
??? .gitignore
??? README.md

Installation

Prerequisites

• Python 3.9+
• Ollama installed and running
• Mistral model pulled: ollama pull mistral

Setup

# Clone the repository
git clone https://github.com/YOUR_USERNAME/my1strag.git
cd my1strag

# Create a virtual environment (recommended)
python -m venv venv
source venv/bin/activate        # Linux/Mac
venv\Scripts\activate           # Windows

# Install dependencies
pip install -r requirements.txt

How to Run

Step 1: Ingest a PDF (build the vector index)

Place your PDF file in the project folder (or update the path in chunk_pdf.py), then run:

python chunk_pdf.py

This will:

• Extract text from the PDF
• Clean and chunk the text
• Generate embeddings
• Save the FAISS index (vector_indices.faiss) and chunks (chunks.pkl)

Step 2: Ask questions

Make sure Ollama is running, then:

python query_rag.py

Type your question when prompted — the system will find the most relevant chunks and generate an answer using Mistral.

Future Improvements

• Process multiple PDFs from the data/ folder automatically
• Store vectors in a proper vector database (Pinecone, ChromaDB, etc.)
• Build a FastAPI endpoint to expose the RAG pipeline as an API
• Add a simple web UI for asking questions
• Support more file formats (Word, TXT, etc.)

License

This project is for learning purposes. Feel free to use and modify it.