Compatibility

Warehouses

The decision server is warehouse-independent — the dbt-state client reads each relation's last_modified and ships the SQL; the server just fingerprints and checks freshness.

warehouse	status
postgres, snowflake	supported (tested)
databricks, redshift, bigquery	supported (untested — please open an issue if you hit problems)

Note

postgres needs track_commit_timestamp=on (the client reads freshness from pg_xact_commit_timestamp). Snowflake needs nothing extra.

dbt engines

engine	status	how to enable
dbt Core 1.7–1.11 (dbt-state plugin)	supported	auto-engages once RUN_CACHE_API_URL is set
dbt Core 1.12+ (native, beta)	supported	DBT_ENGINE_MANAGE_STATE=true
dbt Fusion (preview)	supported	DBT_ENGINE_MANAGE_STATE=true or --manage-state

In 1.7–1.11, dbt-state is a separately-installed plugin that hooks the run, so it turns on automatically. From 1.12+ it's bundled into the engine (and Fusion reimplements it natively), so you switch state on explicitly.

All three speak the same com.fivetran.query_cache gRPC protocol, which is why the same self-hosted server works across them.