AI-Powered Knowledge Assistant (RAG System with Database Integration)

An AI-powered knowledge assistant that answers questions from a user's own documents (PDF, DOCX, TXT) using Retrieval-Augmented Generation (RAG), with full database integration.

Key features:

- Documents are automatically chunked, embedded, and stored in a vector database (ChromaDB) for semantic search

- A relational database (SQLite) stores document metadata and a full audit log of every question asked and answer given

- Answers are generated using an LLM and are strictly grounded in the retrieved document content, with inline citations for every claim

- The system is designed to honestly say "I don't know" when a question falls outside the uploaded documents, rather than guessing — verified through direct testing

- Built with FastAPI (backend), a custom chat interface (frontend), and a modular design that supports multiple LLM providers (Groq, Gemini, Claude)

This project demonstrates end-to-end RAG architecture: ingestion pipeline, dual-database design (vector + relational), grounded generation, and trust/confidence scoring — the full stack needed for a real, usable knowledge assistant rather than just a demo chatbot.