Dagster icon

Dagster

Official · maintained by Marmot
marmotdata/dagster

Discover jobs, ops and software-defined assets from Dagster

Pipelines Assets Lineage Run History Experimental

The Dagster plugin discovers jobs, the ops inside them, and software-defined assets from a Dagster webserver.

It reads everything through the GraphQL endpoint at {host}/graphql, so it works against both open source Dagster and Dagster+. For Dagster+, set token and the plugin sends it as the Dagster-Cloud-Api-Token header.

Assets

  • Pipeline per job, named after the job. Schedules, sensors and a summary of recent runs are attached as metadata.
  • Task per op, named <job>/<op handle>.
  • Dataset per software-defined asset, named after its asset key joined with /. A TableSchema metadata entry becomes the asset's columns.

Dagster's implicit asset jobs (the ones named __ASSET_JOB) are skipped.

Lineage

Pipelines contain their Tasks, Tasks depend on the Tasks feeding their inputs, Datasets feed the Datasets built from them, and a Pipeline produces every Dataset it materialises. When an asset names a warehouse table and its compute kind is Snowflake, BigQuery, Postgres or DuckDB, the Dataset also produces that native table.