dagster

Use when building, testing, running, or debugging Dagster -- software-defined assets, asset checks (blocking gates), partitions and partition mappings, backfills, schedules, sensors, declarative automation, FreshnessPolicy, resources and config, IO managers, jobs and ops, executors, Definitions and code locations, the dg CLI and components, dagster-dbt, dagster-duckdb, dagster-dlt, Pipes, unit tests with materialize(), dagster dev, and Dagster+ deployment.

methodology

1. model assets and a time partition first. add entity loops inside the asset. declare lookback with partition mappings.
2. write storage explicitly and return MaterializeResult when the layout is a contract. use an io manager only for in-process handoffs.
3. put every external dependency in a resource. put gates in @asset_check(blocking=True).
4. pick the trigger: partitioned schedule for "yesterday", a sensor with a cursor and run_key for source changes, AutomationCondition for declarative upkeep.
5. pin in_process_executor for single-writer stores. for range backfills give every asset a single-run policy and read partition_time_window.
6. prove it headless: dagster definitions validate or dg check defs, cli materialize, pytest with materialize() and build_sensor_context / build_schedule_context. verify api names against the installed version.
7. record lessons in trained. for pipeline design questions see data-pipeline.

contents

• trainedlearned layer: mental model, gotchas, practiced cases; read first
• dagster-docsofficial docs snapshot (2026-09-25): assets, checks, partitions, automation, resources, io managers, jobs/ops, components, pipes, testing, deployment
• changes-1.10-to-1.13api renames and deprecations by version
• dagster-expertofficial dagster-io agent guide: dg cli, components, integrations index, dagster+ mcp
• dagster-crypto-assetsrunnable assets + blocking check + dbt + schedule + sensor + pytest
• known-bugsdagster issues with workarounds
• sourcesevery source with its version

related gurus: data-pipeline (design, correctness, airflow/prefect comparison), dbt, olap (duckdb), feature-engineering.