About SPMVV EDUBOT
An offline AI-powered assistant built to help students and faculty at Sri Padmavati Mahila Visvavidyalayam find information instantly — no internet required.
What is this project?
SPMVV EDUBOT is an offline AI-powered assistant built for Sri Padmavati Mahila Visvavidyalayam. It helps students and faculty quickly find information about college departments, courses, timetables, events, and administrative resources — without needing an internet connection.
The system leverages Retrieval Augmented Generation (RAG) to search over college documents and generate accurate, context-grounded answers using a locally running Mistral LLM via Ollama.
Key Features
Semantic Search
Semantic search over college PDFs and documents using FAISS for deep contextual understanding.
BM25 Keyword Retrieval
Exact-match keyword retrieval with BM25 to handle precise terminology queries.
CrossEncoder Reranking
High-precision reranking of retrieved results using a CrossEncoder model.
Fully Offline LLM
Local LLM generation via Mistral through Ollama — works entirely without internet.
Role-Based Access
Secure role-based access control so students and faculty see only what's relevant to them.
Direct PDF Links
Instant PDF link resolution for question papers and handouts — no digging required.
How it Works
Query Submitted
You type a question about college academics, events, or resources.
Document Matching
The system checks for a matching college document by keyword. If a PDF is found, a direct download link is returned immediately.
Embedding & Retrieval
Otherwise, the query is embedded using nomic-embed-text and the most relevant chunks are retrieved from the FAISS vector store.
Fusion & Reranking
BM25 scores are fused with semantic scores, and the top candidates are re-ranked with a CrossEncoder for high precision.
Answer Generation
The final context is passed to Mistral running locally via Ollama to generate a grounded, accurate answer.