Parses complex documents into structured Markdown or JSON, preserving tables, LaTeX formulas, charts, headers, footers, and multi-column layouts in reading order. Built on PaddleOCR's PP-StructureV3, it handles the messy documents that trip up simple OCR: invoices with nested tables, academic papers with equations, financial reports with charts, scanned multi-column PDFs. You point it at a URL or local file, it returns a JSON envelope with full text plus per-page structured data including bounding boxes and confidence scores. Processing time scales with complexity, from seconds for single images to minutes for dense 50-page PDFs. Best for documents where structure matters. Skip it if you just need plain text fast.
npx skills add https://github.com/aidenwu0209/paddleocr-skills --skill paddleocr-doc-parsing