Layout-preserving extraction for PDFs, Office docs, and images. Fast local parsing via LiteParse, heavy OCR via Chandra.