dbt semantic models are MetricFlow graph nodes tied to DAG models

Semantic models underpin data definition in MetricFlow, which powers dbt’s Semantic Layer (v1.12+). They are nodes connected by entities; MetricFlow builds the graph from YAML metadata on dbt models. Each semantic model matches one dbt model via a semantic_model block, with an optional display name. Configuration sits in project YAML inside model definitions, or via Apache Ossie documents. Documented components include name, required agg_time_dimension, and entities typed as primary, foreign, unique, or natural.

Analytics engineering teams get a single place to declare how warehouse models participate in metrics, instead of re-encoding joins and grain in every BI tool.

That graph is shared meaning: entities and time dimensions become the join and aggregation vocabulary MetricFlow and downstream tools reuse for governed metric queries.