A department-level knowledge base built from 200+ PDF drawings combines visual, semantic, and full-text retrieval to return the Top 3 matches with evidence.
Engineers often knew a component's outline, hole pattern, or dimensions but not its drawing number. Reviewing hundreds of files slowed quoting and design work.
Users upload a full drawing or partial PNG crop. VLM, OCR, and LLM extraction combine with image vectors, text vectors, and full-text search, while departments retain isolated knowledge bases.
The result page returns the Top 3 drawings, composite scores, matching details, and side-by-side source PDF comparison so engineers can verify dimensions and tolerances.