The extractor version this run used.
Field paths whose re-extracted value differs from the stored one.
True only when the re-extraction is byte-identical to the stored record.
OptionalreasonShort reason when ok is false.
The extractor version the record was produced with.
OptionalunreadableSet when the stored record or the archived bytes it was built from could not be read (a corrupt file, a missing or altered blob): the corpus is damaged, which is a different finding from a record that does not reproduce.
What the check found, so an upgrade does not read as a broken corpus:
reproduced — byte-identical;version-only — identical apart from extraction.extractor_version: the content
reproduces, the record was stamped by another build (ka reextract restamps it);differs — the content differs (contentDifferences), whatever the version;unreadable — the record or its archived bytes could not be read;unchecked — nothing to compare: no such record, or an OCR record without its model.
differenceswithout the version stamp: what the content disagrees on.