Documents
Turn a PDF into a text artifact you can store, index, or feed to a model. Today the surface is a single node: convert.
Formats
Inputs are PDF only; outputs are plain text or markdown. Tables → Compatible files.
| Direction | Formats |
|---|---|
| Send us | pdf |
| Get back | txt, md |
Extraction reads the PDF's embedded text layer. There is no OCR — a scanned page without text comes back empty rather than failing.
A branch with no
document.convert is a passthrough — the source PDF bytes are stored unchanged.Nodes
document.convert — txt or md.
PDF → markdown with runQueue(..., 'document').convert(...) → Recipes (same pattern as other media; fetch the signed URL if you need the text bytes in-process).