Project Overview
DocuFlow is a retrieval-augmented generation system built to process complex documents with precision. It converts PDFs, Word files, and images into structured Markdown before ingestion, so layout, tables, and headers survive into the chunks the model actually sees. From there it runs hierarchical, header-aware chunking, embeds with a local BGE model, and indexes into a persistent ChromaDB store.
The part I care most about is the validation layer: every generated answer passes a Gatekeeper (relevance), an Auditor (hallucination), and a Strategist (logical consistency). That's what separates a demo you can't trust from a system you can actually ship.