Pdf

Converts PDFs to markdown, HTML, text, JSON, or DOCX by first classifying the document type, then choosing the right extraction route instead of running one default pipeline. It assumes you're using opendataloader-pdf as the primary engine and walks through heuristics for medical reports, slide decks, scanned documents, and table-heavy files. The core idea is to validate output on representative sections and retry with better settings before delivering, since what works for one PDF often fails on another. Most useful when you're dealing with structured PDFs where fidelity actually matters and you don't want to manually tune flags every time.