Skip to content
Kitbay
Skill

paddleocr-doc-parsing

Apache-2.0Updated on Sep 16, 202692,573GitHub stars

The PaddleOCR project's skill for turning PDFs and document images into structured Markdown or JSON.

What the skill teaches

The PaddleOCR project's skill for turning PDFs and document images into structured Markdown or JSON. The agent calls the paddleocr command-line tool's api subcommand, installed with uv, on a file path or URL, optionally choosing a model such as PP-StructureV3, page ranges and whether to straighten warped or rotated pages. Results keep tables, formulas as LaTeX, figures, seals, charts, headers and footers, multi-column layout and reading order, returned per page with Markdown text and images. It needs a PADDLEOCR_ACCESS_TOKEN, and the skill's error handling for rate limits and quotas indicates the processing happens on PaddleOCR's hosted service rather than on the user's machine.

Capabilities

  • Convert PDFs and document images to Markdown or JSON
  • Keep tables, LaTeX formulas, charts and reading order
  • Process selected page ranges
  • Correct warped or rotated pages before parsing