NITA-CHATBOT

RAG-Based Chatbot for NIT Agartala

A Retrieval-Augmented Generation (RAG) based chatbot designed for the
National Institute of Technology Agartala (NITA).
The system provides accurate, grounded responses to institutional queries using verified data sourced from the official NITA website.

---

Features

• Retrieval-Augmented Generation pipeline using Google Gemini (gemini-2.0-flash)
• Semantic embeddings with HuggingFace all-MiniLM-L6-v2
• AstraDB vector database for efficient document retrieval
• Recursive web crawler + asynchronous data extraction
• Streamlit-based chat interface with NITA-themed design
• Strict grounding to institutional documents (hallucination-safe)

---

Architecture Overview

User Query → Query Rewriting (Gemini) → Vector Retrieval (AstraDB) → Context Assembly → Response Generation → Streamlit UI

---

Tech Stack

• Language: Python
• Frameworks: Streamlit, aiohttp, BeautifulSoup
• Embeddings: HuggingFace all-MiniLM-L6-v2
• Vector Store: AstraDB
• LLM: Google Gemini 2.0 Flash
• Environment Management: python-dotenv

---

Installation

pip install -r requirements.txt