Feature Type Catalog

Defines a reusable aggregation pattern for summarizing historical behavior over a time window.

Reads a value that was valid for an entity at a reference instant. Versioned, as-of semantics.

Pick the most recent record's value in a recency window.

Compute per-row arithmetic, conditional flags, or CASE labels from existing features.

Normalize one feature by another for scale-invariant signals.

Days/hours since the most recent occurrence of an event.

Mean gap between consecutive events in a window.

Rank a feature within a peer cohort.

Aggregate across child entities under a parent.

Graph degree / neighborhood size for an entity.

Escape hatch for features that don't fit standard Feature Types.