Dagster
Dagster shipped major tooling and integration expansions over the past 90 days. Version 1.13.0 introduced AI-assisted development through the new skills repository and expanded dg api commands, partitioned asset checks that align data quality logic with how data gets produced, and virtual assets (preview) for modeling database views that reflect upstream changes automatically. Meanwhile, state-backed components—used by dbt, Fivetran, Airbyte, Tableau, and Looker integrations—now persist local state by default, reducing code location loading friction. Across the 1.12.x line leading up to 1.13.0, the team added Components API support throughout the platform (dbt Cloud, Azure, Databricks, Polytomic), expanded partition handling in asset checks and asset selection syntax, integrated Databricks and Fivetran observability into the UI, and shifted dg from being a Cloud-only CLI to a general-purpose project tool with deployment validation built in.
Dagster focused on expanding asset selection and observability tooling. Partition filtering arrived in asset selection syntax and the UI gained date range pickers for backfills, while execution contexts now expose multi_partition_key for multi-partition runs. Component support extended to Azure resources and dbt Cloud projects, and the Fivetran integration added a polling sensor for externally-triggered syncs. A breaking change removed the psycopg2-binary transitive dependency from dagster-postgres.