📚 RAG Document Q&A System

Author: Sandeep Kumar

---

What It Does

Upload any PDF → system chunks it → embeds with OpenAI → stores in FAISS vector store → you ask questions → retrieves relevant chunks → GPT-4o-mini generates grounded answers with source citations.

---

Key Results

Component	Choice	Why
Embeddings	text-embedding-3-small	Best cost/quality ratio
Vector Store	FAISS	Fast local similarity search
LLM	GPT-4o-mini	Fast, cheap, accurate
Chunking	RecursiveCharacterTextSplitter	Preserves sentence boundaries
Framework	LangChain	Industry standard RAG tooling

---

Project Structure

project4-rag/
├── app.py           ← Streamlit UI (upload, chat, sources)
├── rag_engine.py    ← Core RAG pipeline (chunk, embed, retrieve, generate)
├── .env             ← API keys (never commit this)
├── .gitignore
├── requirements.txt
└── README.md

---

Quickstart

1. Install dependencies

pip install -r requirements.txt

2. Add your OpenAI API key

Open .env and replace the placeholder:

OPENAI_API_KEY=sk-your-actual-key-here

3. Run the app

streamlit run app.py

4. Use it

• Upload any PDF in the sidebar
• Click Build Vector Store
• Ask questions in the chat box

---

How RAG Works

PDF
 └─► PyPDFLoader (extract text per page)
      └─► RecursiveCharacterTextSplitter (chunk into ~500 char pieces)
           └─► OpenAI text-embedding-3-small (embed each chunk → vector)
                └─► FAISS (store all vectors locally)
                     └─► User asks question
                          └─► Embed question → similarity search → top-4 chunks
                               └─► GPT-4o-mini (answer grounded in chunks)
                                    └─► Display answer + source citations