Recipes

Small, single-task examples for common operations. Coming soon. Planned recipes:

• Extract every table in a workbook to CSV.
• Pull the first page of a PDF to plain text.
• Stream a directory of mixed documents to JSONL.
• Render PDF pages 1–5 to PNG at 300 DPI.
• Extract images from a DOCX preserving the asset deduplication.
• Find every footnote in a paper and resolve it to its target.
• Walk a corpus, log the warning kinds, fail on any StreamLengthMismatch.

Until they land, the Overview and the Guide cover the underlying calls.