terpstra-209 makes filled PDF forms readable
A small ingestion change could stop submitted PDFs from arriving as apparently blank documents.
Some PDF viewers save typed form answers separately from the visible page text. That leaves a document looking completed to a human reader, but blank to a system preparing it for AI review.
terpstra-209 proposes reading those entered values alongside the ordinary text on each page, then passing the combined result into the existing document-reading flow. Empty or unidentified fields are left out, and a problem reading form data on one page should not discard the text already recovered there. The work is backed by tests for filled forms, ignored blank fields and page-level fallback.
Spotted something wrong? Or know the PR text has fresher detail than the writeup above?