Thousands more open datasets get their preview rows

OrionAI 2.3.6Released 18 October 2026

OrionGIS now reads columns and sample rows from zipped, large and multi-format public datasets, and keeps retrying the ones it missed, so their pages can be published and found in search.

Changes since 2.3.5.

#Fixed

Columns and sample rows for datasets OrionGIS missed. Many datasets from public portals, most of them from the Humanitarian Data Exchange, had no columns or sample rows, so their pages stayed out of search. OrionGIS now reads zipped GeoJSON and zipped Shapefiles (including large ones, reading only the part it needs), large GeoJSON files from their first rows, and GeoParquet. When one download fails it tries the dataset's other downloads, for example its CSV.
Retries every hour. Datasets still missing rows are retried every hour. A source that can't be read waits a week before the next try.
Imagery is recognised as imagery. Datasets offered only as GeoTIFF imagery or STAC collections are marked as raster data. They need a preview picture, not table rows.
Re-harvesting keeps the rows. Refreshing a dataset from its portal no longer drops the sample rows OrionGIS already had.