[tomsideguide]

A mixed PDF silently dropped its scanned pages from the output, and a fully scanned one failed as unsupported with nothing to branch on.

Conversion now fails with ConvertError::NeedsOcr naming the pages: needsOcr with pages and pageCount on the error in Node and wasm, NeedsOcrError with pages and page_count in Python, exit code 3 on the CLI.

Detection alone over-reports short or image-heavy text pages (the existing text.pdf fixture is one), so the page list is confirmed through per-page extraction. Watch the second pass on documents detection flags.

---

Summary by cubic

Reports PDF pages that need OCR as a distinct NeedsOcr error instead of silently dropping them from mixed documents or failing fully scanned ones as unsupported.

Error contract

• Node and wasm reject with code: 'needsOcr' carrying pages and pageCount; Python raises NeedsOcrError with pages and page_count.
• The CLI exits 3 for these documents.
• Messages name the specific pages, like "page 2 of 2 needs OCR" or "all 2 pages need OCR".

Detection

• Page lists are confirmed with a per-page extraction pass because detection over-reports short or image-heavy text pages.
• Flagged documents now take a second pass before failing.

^{Written for commit b5ce6eb. Summary will update on new commits.}

[cubic-dev-ai]

All reported issues were addressed across 26 files

<sub>Reply with feedback, questions, or to request a fix.

Fix all with cubic | Re-trigger cubic</sub>