college-chatbot

"AI Chatbot for Smarter Campus Communication!!"

Comment

🚀 Inspiration College students often waste time searching for simple information like event dates, syllabus PDFs, attendance portals, or FAQs. We wanted to build a smart, always-available chatbot that eliminates this friction and provides instant, contextual answers using cutting-edge AI.

🤖 What it does College ChatBot is an AI-powered assistant for the Keshav Memorial Institute of Technology. It: •Answers queries about syllabus, timetables, and departments •convert voice input and speech-to-text support •Provides real-time updates about college events via Supabase •Redirects users to portals like attendance, exam info, and more •Redirects to the clubs of college social media pages •Redirects to the location of the college •Uses RAG (Retrieval-Augmented Generation) to give accurate, context-based responses •Integrates as a floating chatbot on the official website for easy access

🛠️ How we built it We combined several powerful technologies: •Frontend: Built with Vite + TypeScript, with a responsive floating chat UI •Backend: A Flask API with CORS support handles queries, embeddings, and retrieval •AI Engine: Uses Google Gemini API for conversational response generation •Embeddings: We used sentence-transformers (MiniLM-L6-v2) to embed queries and context •Vector Search: Implemented FAISS to find the most relevant info from JSON data •Database: Supabase is used to fetch live event and announcements data •Scraping: Used Selenium to extract syllabus and portal data during initial setup

🧗 Challenges we ran into •CORS and security policies between frontend and Flask backend •Ensuring the chat context remains relevant without overloading the Gemini API •Designing a chatbot that feels intuitive and helpful rather than robotic •Setting up Supabase auth and realtime syncing for dynamic updates

🏆 Accomplishments that we're proud of •Successfully integrated Gemini LLM with RAG using FAISS and sentence transformers •Built a working end-to-end chatbot that answers natural language questions accurately •Created a responsive and user-friendly floating UI that integrates seamlessly with the college website •Enabled live event updates and smart redirect functionality

📚 What we learned •How to implement Retrieval-Augmented Generation in real projects •Working with Google’s Gemini API and prompt engineering techniques •How to combine static and dynamic data using JSON and Supabase •Optimizing embedding similarity search for speed and accuracy

🔮 What's next for KMIT College ChatBot 🌐 Support for Telugu and Hindi queries (multilingual support) 🤝 Integration with college ERP or LMS for personalized responses 📱 Launch as a mobile app widget for Android/iOS •🏫 Scalable for Other Colleges: Transform the chatbot into a plug-and-play solution that can be easily integrated into other college websites. By updating the knowledge base and branding, any institution can quickly deploy their own AI assistant using the same architecture.

Built With

Share this project:

Updates