📘 RAG-PDF-QnA — Retrieval-Augmented Generation for Intelligent PDF Question Answering

This project implements a Retrieval-Augmented Generation (RAG) pipeline that enables users to upload PDF documents and interactively ask context-aware questions.
The system retrieves relevant sections from the document using vector embeddings and generates accurate, human-like answers using LLMs (Large Language Models).

---

🚀 Features

• 📄 Upload any PDF and ask natural-language questions.
• 🧠 Combines retrieval-based search with generative AI for precise answers.
• 🔍 Uses FAISS vector database for semantic search.
• 🧩 Built with LangChain, OpenAI embeddings, and FastAPI for modular, real-time interaction.
• 💬 Supports both CLI and web API usage.

---

🏗️ Tech Stack & Architecture

Layer	Technology	Purpose
Frontend / Interface	Streamlit or FastAPI UI	File upload, question input
Backend Framework	FastAPI	Manages routes, queries, and response handling
Document Processing	PyPDF2 / LangChain	Extracts and chunks text from PDFs
Vector Storage	FAISS	Stores document embeddings for semantic retrieval
Embeddings Model	OpenAI text-embedding-ada-002	Converts document chunks into dense vectors
LLM for Generation	GPT-3.5 / GPT-4 (via OpenAI API)	Generates context-based answers
Pipeline Control	LangChain	Orchestrates retrieval + generation workflow

---

🔄 Workflow Overview

📁 PDF Upload
   ↓
📄 Text Extraction → Chunking → Embedding Generation
   ↓
🧮 Embeddings stored in FAISS Vector DB
   ↓
❓ User Question
   ↓
🔍 Retrieve top relevant chunks (semantic search)
   ↓
💬 Combine retrieved context + user query
   ↓
🤖 Generate final answer using OpenAI LLM
   ↓
📤 Return answer to user