How ghodeswarshreya/vjti_chatbot works
VJTI AI Assistant is a document-grounded question-answering product. In the main workflow, a user asks through the React chat, which posts to the FastAPI endpoint; the backend retrieves PDF-derived chunks from FAISS, sends context to Groq for answer generation, and returns an answer with source citations for display. Separately, ingest.py builds the searchable index from the PDF collection, and ask.py offers a terminal-based querying path. The sampled files show these core relationships; other README-listed capabilities are not expanded beyond documented behavior.
Main entry points: backend.py Chat API · ask.py Terminal Chat · App.jsx Web Chat · faiss_index FAISS Index
Components 10
Knowledge Preparation
- VJTI PDFs
data - PDF Ingestion
ingest.py - FAISS Index
faiss_index
Answer Service
- Chat API
backend.py - Chunk Retrieval
backend.py - Answer Generation
backend.py
User Interfaces
- Web Chat
frontend/src/App.jsx - Terminal Chat
ask.py
Other components
- Questioner
- Groq Llama
Connections 12
- Questioner → Web Chat: asks
frontend/src/App.jsx - Web Chat → Chat API: posts question
frontend/src/App.jsx - Chat API → Chunk Retrieval: handles query
backend.py - Chunk Retrieval → FAISS Index: searches chunks
backend.py - Chat API → Answer Generation: builds context
backend.py - Answer Generation → Groq Llama: requests answer
backend.py - Chat API → Web Chat: returns answer and sources
frontend/src/App.jsx - VJTI PDFs → PDF Ingestion: loads documents
README.md - PDF Ingestion → FAISS Index: builds and saves
ingest.py - Questioner → Terminal Chat: enters query
ask.py - Terminal Chat → FAISS Index: searches chunks
ask.py - Terminal Chat → Groq Llama: requests answer
ask.py