Catalogue

The registry of everything shareable on the exchange: datasets (databanks), AI models, and applications. If a resource is not in the catalogue, it does not exist to the platform.

What a catalogue item records

• The resource's provider and owning organisation.
• Descriptive metadata as linked-data (JSON-LD) — semantically described and machine-interpretable: formats, units, tags, spatial extent, licence.
• Which access types it supports (api, file, sub) — the Policy Service validates every grant against this list.
• The item type: DATABANK, AIMODEL, or APPS.

Discovery

Consumers search the catalogue without any grant — metadata is public even when data is not:

• Attribute search — by key-value pairs such as name and tags.
• Geo-spatial search — by circle, bounding box, polygon, or line string.
• Text search — free-text relevance search.
• Complex search — combinations of the above.

The search index is Elasticsearch; results return the metadata plus the access modalities, so a consumer knows how they could consume a resource before requesting access.

Management

Providers (and their delegates) create, update, and delete their own items. Writes are identity-checked — only the owner or an org admin of the owning organisation may modify an item — and every item is schema-validated for its type before it is accepted. Providers also get a my assets view of everything they own.

Why other components depend on it

Whenever another component needs an authoritative resource fact — who owns this dataset? which organisation does it belong to? which access types does it support? — it asks the catalogue. The Policy Service validates every grant request against it; the File Exchange confirms a file belongs to a real item before issuing a download.

Interactions

Talks to / heard by	Why
Policy Service	Item details for grant validation (gRPC)
Elasticsearch	Metadata index for discovery
API Gateway	Public read path; identity-checked write path