LoginSign Up
← Back to all sparks
D

Dagster

ANALYTICS
Velocity6.3

Data orchestration platform for building and observing data pipelines.

dagster.io ↗

Dagster is turning declarative automation from an asset feature into the way the whole platform schedules work.

data-orchestrationdeclarative-automationdbtasset-healthcomponents
◆Current state
Dagster ships a core/libraries pair on a near-weekly cadence, and the release notes read like an engineering log: a few genuinely new capabilities per version, a long bugfix tail, and steady community contributions. The current cycle is concentrated in three places — Declarative Automation, the dbt-on-Snowflake integration, and asset health reporting. Serverless and Kubernetes deployment paths get frequent hardening.
◆Where it's heading
Declarative Automation is expanding past its original asset scope: it can now launch entire jobs from a condition, with its own evaluation history tab. In parallel, the component model is becoming the packaging unit for integrations, with SnowflakeDbtProjectComponent moving from preview toward parity with DbtCloudComponent via versioned state storage. Asset health is being made more honest — failures pending an automatic retry now warn rather than report degraded, so alerts stop crying wolf.
◆Prediction
Declarative Automation for jobs is the clearest candidate to graduate from preview, and SnowflakeDbtProjectComponent is following the same preview-to-parity path. Expect the component surface to keep absorbing integrations that were previously bespoke code.

◆Recent moves

  1. 4d ago

    Partition-level retry warnings and defs_state for the Snowflake dbt component

    Extends last week's health-status change down to the partition level: an asset whose failed partitions are all awaiting retry now warns instead of reporting degraded, and the popover says a retry is pending. SnowflakeDbtProjectComponent picks up defs_state, plus three fixes around discriminated-union defaults and dbt run attribution.

    View source ↗
  2. 11d ago

    Retry-pending failures now warn instead of degrading

    Retry-pending failures stop reporting as degraded and now render a 'Failed, retry pending' line in Slack, Teams, and email alerts. The rest is asset graph search work and a notable Kubernetes fix where Hybrid deployments were misidentified as Serverless.

    View source ↗
  3. 19d ago

    Declarative Automation can now launch jobs (preview)

    ⚡ SPARK

    Declarative Automation crosses from assets to jobs: an automation_condition on define_asset_job launches a single run when root-asset conditions become true. This is the pivot point for the automation arc that runs through the rest of these releases.

    View source ↗
  4. 26d ago

    Snowflake dbt component preview and MCP server docs

    SnowflakeDbtProjectComponent arrives in preview for orchestrating dbt projects natively on Snowflake, alongside documentation for the Dagster MCP server. The starting point for the component work that continues through the next several releases.

    View source ↗
  5. 1mo ago

    Serverless I/O manager 401 and 400 errors fixed

    Two serverless I/O manager fixes: a 401 in organizations with multiple full deployments, and a 400 for storage keys containing special characters. Deployment plumbing, no user-facing change.

    View source ↗
  6. 1mo ago

    Install-time protobuf version conflict fixed

    A grpcio-health-checking resolution bug that broke code server imports on fresh installs, plus a tick-details error and a launch dialog tweak. Install-path hygiene.

    View source ↗