Schemas
The schema registry stores JSON Schema documents (draft 2020-12) under a subject. Registering the same subject again creates version N+1. After create, the definition is immutable; PUT only updates name and description.
There is no discovery from sample data. You write the schema.
Create
Schemas → New:
- Subject — family name, for example
nmea.vdmorais.decoded - Name / description — display only
- Definition — full JSON Schema object
The API compiles the definition on create. Invalid JSON Schema is rejected.
Validate tab
On an existing schema, paste a sample JSON object to POST /api/v1/schemas/{id}/validate. This does not change the stored definition.
How pipelines use schemas
Each classifier step references a schema id (not “latest”). If you register a new version, edit the pipeline to point at the new id.
Classifier failures become stage errors. That record does not continue; the rest of the run continues.
Example subjects
| Subject | Typical place in a chain |
|---|---|
nmea.vdm | After the NMEA reader |
ais.decoded | After the AIS transformer |
copernicus.catalog.item | Copernicus catalog item |