📑 RAG Document Assistant

An AI-powered document question-answering system built using LangChain, HuggingFace Embeddings, FAISS, and Streamlit. Ask questions about any PDF document — the system retrieves relevant context and answers using an LLM.

---

Visual Architecture

---

🌐 Live Demo

👉 Try it live here

---

📸 Screenshots

Chat Interface

Document Q&A

---

✨ Features

• 📄 PDF document ingestion and processing
• 🔍 Semantic search using HuggingFace embeddings
• 🧠 MMR retrieval for diverse and relevant results
• 💬 Conversational chat interface
• ⚡ Fast and accurate answers from document context

---

🛠️ Tech Stack

• Python 3.11
• Streamlit
• LangChain
• HuggingFace Embeddings (all-mpnet-base-v2)
• FAISS
• Groq API (Llama 3.3 70B)
• PyPDF

---

📂 Project Structure

rag-document-assistant/
│
├── DocumentLoaders/
│   └── deeplearning.pdf
│
├── assets/
│   ├── demo1.png
│   ├── demo2.png
│   └── rag.png
│
├── .env.example
├── .gitignore
├── app.py
├── runtime.txt
├── requirements.txt
└── README.md

---

⚙️ Installation

1️⃣ Clone the Repository

git clone https://github.com/bangarukondabollapally/rag-document-assistant.git
cd rag-document-assistant

2️⃣ Create Virtual Environment

python -m venv .venv

Activate:

Windows

.venv\Scripts\activate

Mac/Linux

source .venv/bin/activate

3️⃣ Install Dependencies

pip install -r requirements.txt

---

🔑 Environment Variables

Create a .env file:

GROQ_API_KEY=your_groq_api_key

Get your free Groq API key at console.groq.com

---

▶️ Run the Application

streamlit run app.py

Then ask questions like:

What is backpropagation?
Explain dropout in deep learning
What is a neural network?

---

🧠 How It Works

PDF Document
↓
PyPDFLoader → loads pages
↓
RecursiveCharacterTextSplitter → chunks (500 tokens, 100 overlap)
↓
HuggingFace Embeddings → converts chunks to vectors
↓
FAISS → stores vectors in memory
↓
MMR Retriever → fetches diverse relevant chunks
↓
Groq LLM → generates answer from context
↓
Streamlit UI → displays response

---

📦 requirements.txt

streamlit
langchain
langchain-groq
langchain-huggingface
langchain-community
langchain-text-splitters
pypdf
sentence-transformers
faiss-cpu

---

🔮 Future Improvements

• 📁 Multi-document support
• 🌐 Web URL ingestion
• 📊 Source citation with page numbers
• 🔎 Hybrid search (semantic + keyword)
• 🧾 Export Q&A as PDF report
• 🗂️ Document management UI