Knowledge Catalog can bring established metrics into one graph

Google Cloud announced general availability on October 29 for expanded semantic-model support in Knowledge Catalog. The service can ingest definitions from Looker LookML, BigQuery Graph, dbt MetricFlow, Snowflake Semantic Views, Databricks Metric Views and SAP Core Schema Notation. It connects those definitions with lineage, quality signals, schemas and glossary entries instead of requiring teams to recreate their business logic in a Google-specific modeling language.

The practical aim is consistency. An AI agent can retrieve the same metric formula, filters and join relationships that analysts already use, while overlapping definitions remain separately identified by source. That does not automatically settle which definition is correct; data stewards still need to resolve conflicts and certify the underlying data.

The GA label does not cover every associated interface

Google marks all six listed semantic-model integrations generally available. Knowledge Studio, the visual workspace for exploring the context graph and packaging reusable context sets, is still a preview. Teams should separate that preview interface from the production status of the ingestion connectors when deciding which workflow to depend on.

Google says internal benchmarks found nearly four times higher answer accuracy, as much as 33% lower latency and up to 75% fewer tokens when agents were grounded through Knowledge Catalog. Those are issuer-reported results, not independent tests, and the public article does not provide the full benchmark design or a guarantee for customer workloads.

Billing and location choices remain part of deployment planning

Knowledge Catalog uses pay-as-you-go Dataplex processing and metadata-storage SKUs. Gemini-powered functions can be billed through Gemini in BigQuery or Gemini Code Assist, while some connector and scan jobs can generate separate BigQuery, Dataflow, Cloud Scheduler, storage or Spark charges. The announcement does not establish one flat price for the new support.

Google documents regional, US and EU multi-region, and global placement options. A resource location cannot be changed after creation, and the global option does not keep metadata in one jurisdiction. The article gives a date but no release time. CGH has not tested the connectors, audited Google's benchmarks or confirmed compatibility with every source-system edition.