I built the Document‑Intelligence‑System, a Retrieval‑Augmented Generation (RAG) based document intelligence system for PDF question answering with semantic retrieval, grounded answers, citations, confidence scoring, and guardrails.
The repository has 1 star on GitHub.
Development started in 2026-06 and the last update was in 2026-07.
Project topics include confidence-scoring, document-ingestion, document-intelligence, faiss, fastapi, guardrails, huggingface, llm, rag, semantic-search, sentence-transformers, vector-search.
I chose Google FLAN‑T5 Base as the language model for fully local, offline inference, and FAISS as the vector database for fast semantic search on moderate collections.
The pipeline uses fixed‑size overlapping chunks, runs on CPU for ease of deployment, and provides a lightweight HTML/CSS/JavaScript frontend.
Example responses show confidence scores such as High (0.81).
For more of my work see my GitHub profile.