Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
- Added installation instructions for required Python dependencies in the usage guide. - Included optional installation step for document optimization and PDF splitting support. - No changes to API, features, or workflow.
What This Skill Does
Extracts structured content from PDFs and images via a PaddleOCR API endpoint. Handles text, tables, formulas, charts, and multi-column layouts in a single call. The agent runs a Python script against the API and presents results based on what the user asks for.
Unlike basic OCR tools, it preserves full document structure including tables, formulas, reading order, and layout across complex multi-column and mixed-content documents.
When to use it
- Pulling all tables from an invoice PDF
- Extracting LaTeX formulas from a research paper
- Parsing multi-column magazine or newspaper layouts
- Getting structured data out of scanned financial reports
- Reading figures and charts from technical documents
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: Extract all the tables from this financial report PDF
- 1Runs python scripts/vl_caller.py --file-path "financial_report.pdf" --pretty
- 2Reads the saved JSON result from the temp directory path printed to stderr
- 3Locates table data in result[n].prunedResult and result[n].markdown fields
- 4Displays all tables in full without truncating or summarizing
All tables from the document rendered in structured markdown, with the full raw JSON saved to a temp file for reference
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.