Enrich Portolan metadata with AI
Match a geospatial file or table to its public catalogue record, then use the correct producer, description and licence when preparing the collection metadata.
Enrich Portolan metadata with AI when a downloadable map table arrives without a reliable description or licence. The enrich-metadata skill helps an agent look for the source record in a supported open-data portal. It matches a file URL or recognisable table reference instead of treating similar dataset titles as proof.
A single record might cover both a property table and a separate history of assessments. Copying the bundle description into one collection could make promises about fields that do not exist there. The same care applies to attribution, since the organisation that uploaded the record may differ from the producer.
What it does
- Find a resource using its exact file location or table name.
- Separate a specific table’s description from facts about a larger bundle.
- Distinguish a publisher or portal maintainer from the organisation that produced the data.
- Check licence identifiers against the documented licence version.
- Present ambiguous matches for a human to select before updating a collection.
Install and use
- Provide the resource identifier, collection metadata file and authorised portal access or a local JKAN checkout. The workflow uses its Python reader and the Portolan CLI for validation.
- Ask, “Find the record for this table and show where its producer and licence came from before updating the manifest.”
- Review the proposed fields against the source and validate the resulting file.
FAQ
Source
Made by Portolan. Original skill: enrich-metadata, published on GitHub under the Apache-2.0 licence.
View on GitHub