📚 Multi-Mode RAG Bot

Made by Aelaf Eskindir

🔗 Live Deployment: Click Here to Try the App 🚀 🔗 Github Repository: Click Here to get the source code 🚀

---

✨ Overview

The Multi-Mode RAG Bot is a Streamlit-powered AI assistant that allows you to:

• 📖 Chat (Q&A) with your uploaded PDF documents using RAG (Retrieval-Augmented Generation).
• 📝 Summarize entire documents with a map-reduce summarization chain.

It integrates LangChain, FAISS, HuggingFace embeddings, and Groq LLMs to deliver a smooth and interactive experience.

---

⚙️ How It Works

🔑 API Authorization

• The app uses your GROQ API Key, securely stored in secrets.toml.
• Keys are never exposed in the codebase.

📂 Document Processing

1. Upload a PDF file in the sidebar.
2. The app will:
   • Save it temporarily
   • Extract the text using PyPDFLoader
   • Split it into manageable chunks with RecursiveCharacterTextSplitter

🧠 Knowledge Base Creation

• Each document chunk is embedded using HuggingFace MiniLM embeddings.
• FAISS builds a vector database for fast retrieval.

💬 Q&A Mode

• Queries are matched against the document using FAISS retriever.
• A cross-encoder reranker refines the best results.
• The selected context is passed into Groq’s LLaMA-3 model for precise answers.

📝 Summarization Mode

• A map-reduce summarization chain condenses the document into an easy-to-read summary.

---

🖥️ User Interface

Sidebar

1. 📤 Upload your PDF file
2. ✅ Get confirmation once it’s loaded

Main Panel

• Choose between Q&A (Chat) or Summarization mode.

Q&A (Chat)

• Ask questions in the chat box
• Get contextual answers based only on the uploaded PDF

Summarization

• Click "Generate Summary" to get a concise overview of the document

---

Tech Stack

• Streamlit – Web app framework

• LangChain – AI orchestration

• Groq – LLaMA-3 inference

• FAISS – Vector search

• HuggingFace Transformers – Embeddings & reranker

---

📌 Example Use Cases

• 📚 Research papers → Summarize and Q&A

• 🏛️ Legal documents → Extract key points quickly

• 🧑‍🎓 Study material → Summarize chapters or test knowledge

• 📊 Reports → Fast insights