v4.2.0 — 31 July 2026

Course documents can now make the whole round trip — out of Canvas, through remediation, and back into the course — without anyone moving a file by hand.

Added

• Canvas round trip, end to end. Pull files from a Canvas course, remediate them, and write the corrected version back to the course. Write-back is governed by the Replace original file toggle, which is off by default, so a run leaves the course untouched unless you ask it not to. Available to Admin and Department Admin roles. See Connect a source.
• Hybrid extraction engine. A third option alongside PaddleX and Document Intelligence, which applies each of the two to the parts of a page it reads best rather than committing the whole document to one. Selecting it shows both threshold panels. With three engines to pick from, which one suits which document now matters — see choosing an engine.
• Configurable PaddleX detection thresholds. Tables, formulas and layout elements each get their own control, so a document class that keeps coming out under- or over-tagged can be corrected at the source instead of in the editor. They apply to PaddleX only. See raising and lowering a threshold.
• Advanced filters on the batch list. Narrow a long list of batches instead of paging through it, which is the difference between finding last week's course in seconds and scrolling for a minute.

Changed

• A batch now takes up to 100 files, raised from 50 — a course-sized run more often fits in a single batch.
• Structural auto-tagging improvements. The tagging engine recognizes document structure more accurately, which shows up as higher accessibility scores on the same source files. The score comes from the PDF/UA-1 check, so a better score means fewer structural failures to work through in the Tag & Caption Editor — not that a document is certified conformant.

---

Also in this release: platform changes beyond PDF accessibility. See the full changelog.