🧠 RAG Wikipedia QA (Offline)

A fully offline Retrieval-Augmented Generation (RAG) system that lets you query any Wikipedia topic using a local LLM and vector store — with no internet or API keys required.

🚀 Features

✅ Search any Wikipedia topic and save the content locally
✅ Embed saved documents into a local FAISS vector store
✅ Ask questions using Retrieval-Augmented Generation (RAG)
✅ Runs fully offline — no API keys, no cloud dependency
✅ Lightweight model (sshleifer/tiny-gpt2) — runs even on CPU
✅ CLI interface for easy interaction
✅ LangChain deprecation warnings cleaned

🗂️ Folder Structure

RAG-Wikipedia-QA/
│
├── docs/ # Saved Wikipedia text files
├── embeddings/ # FAISS vector DBs saved here
├── rag_wikipedia.py # Main script
├── wiki.png # Image (Step 1 & 2)
├── QA.png # Image (Step 3)
└── README.md # This file

🔧 How It Works (Step-by-Step)

✅ Step 1 – Fetch Wikipedia Content

• User is prompted to enter a topic name (e.g., India, Python programming language)
• The script fetches the article summary and saves it as a .txt file under /docs

✅ Step 2 – Embed Text with FAISS

• Loads the saved .txt file
• Splits the text into chunks using LangChain’s CharacterTextSplitter
• Embeds the chunks into vectors using Hugging Face embeddings
• Stores them in a FAISS vector database (.faiss and .pkl) inside /embeddings

📸 Screenshot:

✅ Step 3 – Ask a Question (RAG)

• Prompts user to enter the same filename
• Loads the vector store and retrieves relevant chunks based on the question
• Feeds context + question into a local GPT2 model
• Generates and returns an answer offline

📸 Screenshot:

📦 Requirements

Install dependencies using:

pip install -r requirements.txt


👤 Author
Asif Noushad Sharafudeen
🔗 LinkedIn
🔗 GitHub