Catalogue metadata engine
Every image should be findable
by the one who seeks it.
Barbara reads digitised images and proposes structured catalogue metadata — title, description, subject terms, and only what the image visibly supports. A human reviews and approves; approved records export to a CSV for the collection system.
1 · Describes each image→
2 · Human reviews & approves→
3 · Exports clean CSV
Sample output · generated by Grok vision, reviewed for the demo
About this showcase. These three records were generated by the live Barbara engine
(Grok vision through the owner's account) and are shown here as sample output. Every record
uses an evidence-first rule — where an image does not visibly support a date or place,
Barbara leaves it blank rather than guessing. If a detail is uncertain, it is flagged for the
reviewer. This is a static preview of the running demo; the interactive review-and-export
workflow runs on the owner's machine.
Barbara · demo build · powered by Grok vision · human review before any record ships