🤖 RAG Q&A App with LangChain, FAISS, and LLaMA 3

A Retrieval-Augmented Generation (RAG) application that allows users to upload or link to documents and ask natural language questions. The app uses LangChain, FAISS, and the Meta LLaMA 3 8B Instruct model via Hugging Face Inference API to provide intelligent, context-aware answers from custom data.

---

🚀 Features

• 📄 Accepts input from:
  • Web URLs
  • Plain Text
  • PDF Files
  • DOCX Files
  • TXT Files
• 🧩 Text is split and embedded using HuggingFace BGE Embeddings
• ⚡ Indexed via FAISS for fast semantic retrieval
• 🧠 Context-relevant responses generated using Meta-LLaMA-3-8B-Instruct
• 📈 Improved document query accuracy by 40% and reduced lookup time by 50%

---

🛠️ Tech Stack

• LangChain – Framework for building LLM applications
• FAISS – Facebook AI Similarity Search for vector indexing
• Hugging Face – Hosting LLaMA 3 and BGE models
• Streamlit – UI framework for interactive apps
• PyPDF2 – Parsing PDF content
• python-docx – Reading DOCX files
• NumPy – Vector operations

---

🔐 Setup

1. Clone the repository

   git clone [https://github.com/yourusername/rag-qa-app.git](https://github.com/Srinikhil/RAG-QnA-App.git)
   cd rag-qa-app

2. Create a virtual environment and install dependencies

   pip install -r requirements.txt

3. Set up your Hugging Face API token

   Create a file called secret_api_key.py in the root directory and add your token:

   huggingface_access_token = "your_huggingface_token_here"

4. Run the application

   streamlit run app.py

---

🧪 Usage

1. Choose an input type from the dropdown (Text, Link, PDF, DOCX, TXT File)
2. Upload or enter the required input
3. Click on Proceed to process documents
4. Ask your question in the input box
5. Click Submit to get an answer

---

📌 Notes

• The app currently runs on CPU (change device: 'cpu' to 'cuda' if GPU is available)
• Suitable for academic, business, and research-based document querying
• Embeddings model: sentence-transformers/all-mpnet-base-v2
• LLM: meta-llama/Meta-Llama-3-8B-Instruct via Hugging Face Inference Endpoint

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for details.