Small PDFs used by the integration tests (PRD §8). Generated synthetically where possible so they can be committed and stay tiny:
| Fixture | Purpose |
|---|---|
born-digital.pdf |
A real text layer — text extraction + TextBased classification. |
scanned.pdf |
A single full-page image, no text — Scanned classification + OCR. |
mixed.pdf |
Substantial text and image — Mixed classification. |
forms.pdf |
An AcroForm — form rendering + fill_form. |
encrypted.pdf |
Encrypted with a known password — password handling. |
multi-heading.pdf |
Several heading levels — chunk heading_path tests. |
Fixtures are generated by cargo run -p xtask / the helpers in each crate's
tests/, not hand-authored binaries, so regenerating them is deterministic.