dbt entities are typed join keys that link semantic models

Entities are business concepts—customers, transactions, campaigns—and in Semantic Layer models they serve as join keys. In the semantic graph (v1.12+), name and type are required; name may match a key column or alias via expr, and must not collide with a simple metric or dimension in the same model. Types: primary (one record per row, all records), unique (one per row, nulls allowed, may be a subset), foreign (links to another table’s row), and natural (real-world identifying attributes). Entities may also be used as dimensions for aggregation grain.

Modeling teams need explicit join-key types so MetricFlow joins semantic models without ambiguous keys or duplicated names.

Typed entities are the shared identity vocabulary across models: the same named keys carry meaning from warehouse tables into governed metric graphs.