Sep 10, 2026: Using dbt artifacts for Slim CI and defer to production in dbt Projects on Snowflake (General availability)

Slim CI, defer to production, failed-execution recovery, and concurrent execution are now generally available for dbt Projects on Snowflake. Reuse dbt artifacts from recent production executions to accelerate development, streamline CI/CD workflows, and recover from failed production runs without maintaining a separate artifact store.

With this release, you can:

  • Run Slim CI: Import dbt artifacts from the latest successful production execution, then use --state and --select state:modified+ to validate only changed resources and their downstream dependencies. This reduces CI execution time and warehouse use.
  • Defer to production: Use defer to production in Snowflake Workspaces or with Snowflake CLI to resolve unbuilt upstream references to existing production relations instead of rebuilding unchanged models in an isolated development or CI target.
  • Build complete CI/CD workflows on Snowflake: Use Snowflake CLI and Snowflake system functions to separate deployment from execution, create an isolated database for each pull request, and retrieve recent or query-specific dbt artifacts. Record the Git branch and commit for each deployment so that a dbt project object remains traceable to its source.
  • Run one project concurrently: Execute independent slices of the same deployed dbt project object at different cadences without maintaining duplicate objects. Use DEFAULT_WRITEBACK or the per-execution WRITEBACK setting to prevent concurrent executions from writing to shared live target and log paths. Alternatively, give each concurrent execution a distinct --target-path and --log-path.
  • Recover efficiently from failed executions: Reuse artifacts from the latest failed run with selectors such as result:error+, so you can rerun errored resources and their downstream dependencies without repeating work that already succeeded.
  • Choose when and how to compile: Skip automatic compilation during deployment with AUTO_COMPILE = FALSE or Snowflake CLI --no-auto-compile to take advantage of Slim CI and simplify governance with a deploy-only role.

These workflows require dbt project objects to use a single mutable live version. Until the 2026_06 behavior change bundle is generally enabled, enable the bundle in your account to opt in. When enabled, new and recreated objects use the live version, and you can migrate existing versioned objects with SYSTEM$MIGRATE_DBT_PROJECT. Once the bundle is generally enabled, Snowflake will automatically migrate all remaining versioned objects.

For more information, see: