OrionAI describes thinly documented datasets
OrionAI 2.4.0Released 21 October 2026
A new hourly step writes "About this data" for published datasets whose publisher said little about them, using only the dataset's own details.
Changes since 2.3.8.
#New
- About this data. Every hour, OrionAI takes up to 120 published datasets whose own description is under 120 characters and writes two or three paragraphs about each: what one row or feature represents, what the main columns hold, the area and years covered, and who publishes it under which licence. It writes only from the dataset's columns, sample rows, coverage and publisher, never adds numbers, places or uses that aren't there, and writes nothing when there isn't enough to say. The text is kept apart from the publisher's description, which is never changed, and a changed page is passed on to search engines within the hour.