Vertex AI Feature Group
Does ZopNight manage Vertex AI Feature Group?
A Vertex AI feature group is free metadata that defines features over BigQuery tables for the current-generation Feature Store. Cost lives in the BigQuery storage underneath and in any feature online store serving those features. ZopNight inventories feature groups via Cloud Asset Inventory so they appear alongside the online stores and BigQuery datasets in the same project.
Rules that fire on Vertex AI Feature Group
No active rule family targets Vertex AI Feature Group today. Rules that used to are retired, and retired rules publish no pages and fire no findings. Scheduling and permissions coverage are unaffected.
A feature group organizes feature definitions over BigQuery data for the newer Vertex AI Feature Store. The group itself is metadata; serving and storage bill through associated resources.
Definitions over BigQuery, priced accordingly
A feature group is the current-generation Feature Store’s registration layer: it declares which columns of a BigQuery table or view are features and which entity keys them. Nothing in that declaration meters. The data never leaves BigQuery, where it bills as ordinary BigQuery storage, and low-latency serving only starts costing money once a feature online store syncs those values into provisioned serving nodes.
Feature groups in the pipeline map
ZopNight inventories feature groups via Cloud Asset Inventory, so they appear alongside the online stores and BigQuery datasets in the same project. The group is the join point between a warehouse asset and the models consuming it, but ZopNight does not trace a group back to its BigQuery source table or forward to its online stores; that link is read in the console. Without it, feature storage reads as anonymous warehouse growth.
Drift between declared and served features
The hygiene patterns worth catching are all indirect. Groups declare features no model has read in months while the source tables keep accruing warehouse storage. Groups point at tables refreshed by scheduled queries whose downstream consumers vanished, leaving compute running to maintain data nobody fetches. And parallel teams define overlapping groups over the same table, doubling the maintenance surface. None of these bill on the group itself; each marks upstream storage and refresh compute doing unclaimed work.
Feature groups in the console
Google Cloud console → Vertex AI → Feature Store, in its feature groups view, lists each group with its BigQuery source and declared features. Following a group through to its online stores shows whether the declared features are actually being served anywhere. A group row can be deleted at any time without touching the underlying BigQuery data it described.