A smart AI chatbot for the Keshav Memorial Institute of Technology (KMIT), built to help students, faculty, and visitors instantly access important information, events, and resources via a floating assistant on the website.
- 💬 Conversational AI powered by Gemini LLM (Google Generative AI)
- 📚 RAG (Retrieval-Augmented Generation) using FAISS + Sentence Transformers
- 🗃️ College Info + Event Data loaded via JSON and Supabase
- 🔗 Quick Redirects – Like Attendance Portal, Exam Info, Syllabus, etc.
- 🧑🎓 Smart Responses to FAQs, context-based queries, and navigation needs
- 🖼️ Floating Chatbot UI integrated into the college website (Vite + TypeScript)
- 🔐 CORS enabled Flask backend to serve AI and database responses
| Layer | Tech Used |
|---|---|
| 🧠 AI Model | Gemini API (Google Generative AI) |
| 🔍 Embeddings | Sentence-Transformers (MiniLM-L6-v2) |
| 📥 Vector DB | FAISS (Facebook AI Similarity Search) |
| 🌐 Backend | Flask + Flask-CORS + Supabase SDK |
| 📊 Data | JSON file + Supabase DB |
| 🧪 Scraping | Selenium (ChromeDriver) |
| 🖥️ Frontend | Vite + TypeScript |
Here’s how the chatbot looks when embedded on the KMIT website:
- User submits a question through the floating chatbot.
- Query is normalized and embedded via Sentence Transformers.
- FAISS finds the most relevant context from JSON data.
- Context + query are sent to Gemini API for final response generation.
- Supabase is queried for dynamic content like event updates.
- Response is shown in the chatbot UI.