portolan-pipeline
A mixed reference catalog exercising every major case in the Portolan v0.1 specification with real, openly licensed data pulled from its original upstream sources, vector polygons and points, raster, tabular, mirror provenance, nested catalogs and flat collections.
Every Collection here is a mirror that cites its true upstream source with a real checksum, and every README and AGENTS.md is written to the standard in the spec's documentation best practices, so the catalog doubles as the worked example of what good catalog documentation looks like. The Netherlands provinces Collection also mirrors its upstream ISO 19115 record as a metadata-role asset, and every code block in these docs is executed by the repository's checks before it publishes.
Agents, each Collection carries its own AGENTS.md with join keys, quirks, and tested queries, and the catalog-level AGENTS.md maps the joins between Collections.
Licenses, CC-BY-4.0, CC0-1.0, PDDL-1.0. Provenance, 8 mirror Collections.
Upstream sources.
| 5,000 points |
| A fixed 5,000-record extract of San Francisco's 388,550-record master address database, kept small on purpose for spec demonstration. |
| Sample Raster COG | 3-band raster | A 1 MB Landsat-derived reference COG for learning and testing cloud-optimized raster access, not for analysis. |
| Eurostat Electricity Prices for Household Consumers | 65,466 rows | 65,412 semi-annual household electricity prices for 41 European countries from 2007 on, the official record of the 2022 energy crisis. |