Last 7 days
1
Features: 0
Changes: 0
Fixes: 1
Deprecations: 0
Dagster
Data orchestration platform with the managed Dagster+ cloud service.
Latest Dagster changelog updates, official release notes, breaking changes, security patches, pricing changes, and developer reactions in one product feed.
Follow this Dagster release-notes page to spot useful features, risky migrations, noisy announcements, and source links before they hit your backlog.
Where Dagster updates are checked
Changes.Watch links back to official changelog and release-note sources so summaries stay easy to verify.
Related changelog channels
Use channels to follow groups of tools around a stack, workflow, or topic.
Dagster updates by type
Rolling windows show how many product updates landed in the last 7, 30, 90, and 365 days, grouped by existing changelog semantics.
1
4
13
51
Latest Dagster changelog updates
Added an expand/collapse all toggle to the automation condition evaluation table in the UI
Asset health now reports a warning status (instead of degraded) when all failed partitions are pending automatic retry, with updated popover and alert messages.
Asset health now reports WARNING for pending retries and UI gains search improvements, copy key action, and SLURM tag icon
Added Declarative Automation preview to trigger jobs via automation conditions, with evaluation history visible on the Automation tab.
Added SnowflakeDbtProjectComponent (Preview) for native dbt orchestration on Snowflake
Fixed 401 error from serverless I/O manager when materializing assets in organizations with multiple full deployments
Asset launch dialog now auto‑expands the warnings section when alerts are present.
Improved runs feed performance by fetching a bounded preview and loading full lists on demand; UI pool name overflow fixed.
UI: Asset catalog now uses a virtualized list to prevent freezes on large workspaces; run detail header shows asset check count and the "View asset lineage" action correctly scopes to the run's assets.
Fixed stale materialization bug for wiped assets and partition wipes
Hierarchical asset groups now support '/' separators, wildcard selection, and a new `is:` filter for asset types.
Allow special characters in team owners and add kind tags/icons for Microsoft Fabric and OneLake
Added owners parameter for partitioned job schedules, Python 3.12‑3.14 support, and Fivetran column‑metadata fetching.
Round‑robin dispatch for schedule, sensor, and asset daemon ticks to prevent single code locations from blocking launches; added CLI "dg api run launch" command and configurable dbt Cloud job pool to work around concurrency limits.
Added new kind tags for Alteryx, Boomi, and SAP; increased SQLite busy timeout to 30 s; UI updates include removing the “Scheduled” tab and expanding the sensor tick timeline.
Added new features including typing.Mapping/Sequence support, GraphQL client path prefix, storage kind field, asset job owner validation, schedule/sensor selector matching, Azure endpoint suffix, dbt translation parameter, and a preview ...
Added an owners parameter to asset jobs, introduced new ClickHouse libraries, and added workload‑identity‑federation authentication options for PostgreSQL.
UI enhancements: Cmd/Ctrl click opens runs in a new tab, Azure DevOps kind tag icon added, and branch deployments now expose run‑queue settings in the configuration dialog.
Added preview PipesCompositeMessageReader, new asset selection attributes (sensor:, schedule:, job:, automation type:), and defaulted state‑backed integration components to local filesystem storage.
Introduced AI‑assisted development with Dagster Skills and added virtual asset support (is_virtual flag) for modeling views without materialization.
Added ability to configure agent queue and image via pyproject.toml and introduced new `dg api` commands for jobs, asset checks, and partition status
Added partitions attribute filtering, Spark Declarative Pipeline preview, and UI date range picker for backfills
Execution context classes now expose a MultiPartitionKey for multi‑partition runs and new Braze/Runpod kind tags were added.
Added a Report Execution dialog for manual asset check results and expanded configurability with DB pool options and new dg+ config view command.
Added event‑type filter to Asset Recent Updates timeline (persisted preference) and removed the transitive psycopg2‑binary dependency, requiring explicit addition if needed.
Fixed bug causing the Dagster UI to fail loading due to inclusion of the built webapp in the dagster webserver.
Pool naming now accepts any non‑whitespace character and ECS RunLauncher retries on throttling errors.
Added `partitions_def` support to @asset checks, enabling checks on specific asset partitions.
Added PolytomicComponent to represent Polytomic bulk sync schemas as external assets in Dagster.
Added validation for Dagster Plus deploy start, new AWS resource‑mapping components, UI ability to unpin empty asset groups, and raised max Kubernetes version to 35.x.x.
Deprecate the `dagster cloud ci check` command; use `dg plus deploy start` which now validates configuration during deployment initialization.
Fixed UI bug where the Catalog tab sometimes displayed no assets after upgrading to 1.12.9
Added Python 3.14 support for core and libraries, plus Python 3.13 support for dg plus deploy; introduced a region flag for dg plus login and a Helm chart concurrency setting.
Added Python 3.13 and 3.14 support to dg plus deploy commands.
Optimized partition key calculation and added datetime/timedelta contexts, FreshnessPolicy support, and increased asset annotation limits to 10 kinds.
Removed all Dagster project CLI commands; use `create dagster` instead (breaking change)
Updated NextJS to version 15.5.7 for the Dagster webserver and dg docs serve command.
Added new CI scaffolding, slash support in concurrency pool names, pod wait timeout config, and a Zendesk kind tag icon.
Improved AutomationCondition cursoring, expanded template variables, and added active venv support to dg commands.
Dropped Python 3.9 support, raising the minimum Python version to 3.10 (breaking change).
Added numerous UI and CLI enhancements: visible dg commands, Dremio icon, asset freshness legacy indicator, cost metrics on catalog, custom date range in Observe UI, and new automation conditions for asset freshness checks.
Redesigned UI with a collapsible sidebar; Components framework and dg CLI reach GA, and new integration components (Sigma, Looker, Tableau) plus a StateBackedComponent for persistent state management.
Add configurable proxy GRPC server heartbeat TTL via DAGSTER_GRPC_PROXY_HEARTBEAT_TTL_SECONDS env var.
Added full All sequence support in AssetKey.with_prefix and notebook task support in PipesDatabricksClient.
Added UnionTypes support in ConfigurableResources plus output metadata and freshness policy fields; Updated Airbyte, Fivetran, PowerBI, Sling, Dlt, and DBT components with overridable asset spec/execution methods and removed default io‑m...
Dropped support for deltalake <1.0.0 in dagster deltalake and dagster deltalake polars (breaking change).
UI enhancements: wildcard search in selection inputs, Cost insights, display inherited org roles, and fixed per‑asset/job insight graphs.
Split Anthropic, MCP, and Claude SDK dependencies into an optional "ai" extra, allowing dagster‑dg CLI installation without them.
Added inline component scaffold command, new `require_upstream_step_success` executor config, and default GraphQL log limit of 1000.
Resolved subclasses now support dict[str, T] fields and Dagster Plus Serverless pex builds accept pyproject.toml packages
Added A param exclusions for time‑window partitions, protobuf==6.x support, and new dg scaffold help and dg check toml commands.
dg now aggregates multiple configuration errors, supports run config for asset backfills, and adds opt‑in validation for requirements.env; UI root URL now shows Overview/Timeline for locations with schedules/automations but no jobs.
Added explicit git platform selection for link code references.
Static functions can now accept a ComponentLoadContext and an MCP server exposes dg CLI capabilities to MCP clients.
Schedules now accept a subset of asset checks in RunRequest; added a docs‑integrations CLI and job filtering selection syntax.
Introduced AssetExecutionContext.load for dynamic asset value loading and added ownerReferences option to the K8s executor for proper garbage collection
Added Python 3.13 support and new Tableau asset subsetting/materialization features.
Definition validation now raises errors for mismatched partition mappings, adding stricter checks (breaking change).
Stabilized the Components framework and dg CLI (now Release Candidate), enabling YAML or lightweight Python definitions, custom component wrapping, and extensive tooling.
Added Tableau assets decorator with refresh/poll support, a new Fivetran snapshot command, UI job tag display, and private package management in the integrations marketplace.
Added @asset hooks argument, multi‑asset pool support, and DagsterDbtTranslator get_asset_check_spec customization; introduced AirbyteCloudWorkspaceComponent and backfill daemon thread‑pool configuration.
- Configurable DbIOManager table and updated backfill daemon defaults (threads=true, workers=4) with Helm support.
GA BackfillPolicy and new component framework enhancements: per‑project dg CLI, renamed component files (defs.yaml), @component instance, and new dagster create CLI commands.
Fixed UI error displayed while viewing run logs
- UI enhancements: automation condition auto‑expand & navigation, faceted asset node facet, and experimental integrations marketplace tab (feature‑flagged).
Added config for backfill threadpool, asset‑failure re‑execution, UI asset lineage navigation, GA for FivetranWorkspace, and new scaffolding features
Added refined Tableau integration for API 3.25+, new UI kinds tag for TreasureData and Supabase icon, and introduced PipesK8sComponent and a dagster.job scaffolder.
Warn instead of error for unselected asset checks and ignore missing asset keys in automation conditions.
UI enhancements: asset graph now displays health info, new feature flag to customize asset facets, removal of By‑partition grouping view, and fixes for asset event filtering and loading indicator bugs.
Added UI run tagging, log level counts, AirflowFilter API, and rollback of errored code locations.
Added a new blocking parameter for freshness checks, raised the default gRPC request/response limit to 100 MB, introduced Definitions.map for asset spec transformations and stricter partition‑mapping validation.
Fix UI datepicker filter selection for users with custom Dagster timezone settings.
Added UI capabilities to specify extra tags on re‑execution, copy run tags as YAML, view execution plans for queued runs, and performance improvements for partition pages along with several UI bug fixes.
Sensor test result changes now also apply dynamic partition requests and a backfill policy is added for dlt assets; UI improvements include command‑click run view, extensive graph zoom, dbt Cloud kind tag, and better backfill handling.
Added AutomationCondition.executed with tags(), AssetCheckEvaluation yielding from ops, kinds argument for AssetOut, stream option for SlingResource.replicate, auth support in DagsterGraphQLClient, and UI improvements like slash support ...
Added async `yield` support for ConfigurableResource and an event‑loop argument to context builders for direct execution.
Added several UI enhancements, including updated asset overview metadata, run log partition details, backfills navigation, and pool event linking
Added several UI improvements: links from run logs to pool configs, queued run info, new low‑contrast themes, new asset kind icons, viewer‑role alerts visibility, and preserved asset filters.
Added pool support for dbt, dlt, sling and new resources (AWS RDS, MySQL, Azure Blob); introduced Terraform modules for AWS ECS deployment and a Dagster Plus setting for direct snapshot uploads.
Fixed run queue deadlock for pool-assigned ops without limits
Added new AutomationCondition.version_changed, PowerAutomate AdaptiveCard support, pool support for dbt, dlt, sling, and switched default run coordinator to queued (requires daemon).
Fixed exception raised when runs using global op concurrency attempted to claim a concurrency slot.
Introduce a top‑level `pool` argument for assets/ops, UI support for concurrency pools, and required schema migration with run‑queue config moved under concurrency settings
Fixed dagster dev crash on Windows platforms
Added new AutomationCondition methods (.replace, .allow, .ignore) and helper get_partition_key for partitioned assets; introduced task definition prefix for EcsRunLauncher, PipesEMRContainersClient for AWS EMR on EKS, timestamp metadata ...
Added a new `load_definitions` function to load assets, checks, schedules, sensors, and jobs from a module into a single Definitions object.
Fixed load assets error when spec and AssetsDefinition share the same key, now only errors if include specs=True.
Introduced dagster.load_definitions_from_module to import all assets, checks, schedules, sensors, and jobs in one call, and enhanced backfill support with mixed selections, partition definitions, thread‑pool processing, and title validat...
Updated cronitor pin to support version 5.0.1 (enables DayOfWeek = 7) and added UI enhancements such as Google Drive icons, renamed run lineage sidebar, and clickable sensors/schedules.
Added a unified automatic retry daemon with run‑retry tags, backfill cancellation handling, and exposed DagsterRun execution info and SensorReturnTypesUnion for typing
Enabled global op concurrency on default SQLite (requires migration) and added bulk asset spec modifications via map_asset_specs, replace, and merge methods; introduced Pipes S3 include_stdio param, experimental external log stream, Powe...
Added run‑id field to run‑tags index for better DB performance and migrated Dagster instance; introduced many new kind‑tag icons and experimental dagster‑airlift feature.
Added new AssetOut.from_spec constructor, UI tag icons and tooltips, and support for propagating tags to ECS tasks, custom dbt code versions, and arbitrary Pipes metadata.
Added project scaffolding templates and multiple UI enhancements (persistent catalog filters, partition display, case‑insensitive owner/group/tag filtering). Introduced new helpers and config options for Tableau, Looker, Fivetran, Snowfl...
Declarative Automation is now stable with built‑in and custom conditions for asset checks and observable assets
Added performance enhancements for multi‑asset code location loading, DBT asset selection, Sigma parallel API calls, and Looker fetching/rendering
Added new AutomationCondition methods (eager, missing, on, all_deps) and AssetSelection improvements (materializable, broader checks) plus a code version parameter for the @graph asset decorator.
Added experimental automation conditions (eager and asset matches) and multi‑kind support for assets
Add run‑tags parameter to JobDefinition/@job for separating definition‑level and run‑level tags; introduce UI backfills view and new navigation flag with revert option
Added AssetSpec.with_io_manager_key method and updated SourceAsset deprecation warning; introduced max runtime seconds config and enhanced ScheduleDefinition and SensorDefinition with tag support and custom execution functions.
Added a partition‑range option to the asset‑materialize CLI and new .without() and on missing() methods for AutomationCondition, plus support for loading multiple Python modules as separate code locations; UI now shows a button for code‑...
AssetSpec now validates group names early and assets stored via DbIOManager automatically attach relation identifier metadata
Fixed Dagster+ Serverless partitioned asset materialization failures caused by missing 'base_path' attribute.
Added new capabilities such as a PipesECSClient for ECS tasks, increased subprocess termination timeouts for dbt and sdf, and UI option to clear IndexedDB caches; consolidated log manager APIs and expanded caching/asset selection support.
Fixed exception when viewing run logs with watchdog 5.0 release
Consolidates implicit asset jobs for assets with differing PartitionsDefinitions into a single job per code location, reducing load time.
Improved UI performance for Automation history and added ability to delete dynamic partitions;
Added checkpointing for sensor daemon failures, a run‑tags argument for build contexts, and support for nested partially configured resources outside definitions.
Added AssetSpec support (including partitions) and Definitions.merge for composing large projects; sensors/schedules can now target assets directly; AutomationCondition replaces AutoMaterializePolicy; Dagster Pipes and dbt integrations a...
Introduced experimental PipesGlueClient for AWS Glue pipes in dagster-aws.
Added per‑step Kubernetes config option for the Celery‑K8s executor.
Added slash‑delimited asset keys in ShellCommandBlueprint and introduced an mlflow run ID attribute with automatic retry on fetch failures
Stabilized dbt integration: DbtProject is no longer experimental, added --use-dbt-project flag, moved manifest preparation to DbtProject.prepare, and replaced the old CLI command with a new prepare + package flow (breaking change).
Fix UI behavior issues on jobs page introduced in 1.7.11
UI and performance improvements: faster loading of large asset graphs and backfills, code location page filter, scrollable backfill page, and fixed partition filtering issues.
Introduced a dbt asset freshness API and expanded tagging support (dagstermill, Sling, dlt) plus a backfill‑daemon log view and read‑only flag for Dagster Cloud CI deployments.
UI and asset metadata enhancements: added storage‑kind tag, dbt asset retry policy, new Runs tab on backfill details, external asset materialization status, and catalog search filters by type
Added several new features: custom titles/descriptions for GQL backfills, get_all_asset_specs, terminate runs in GraphQL client, partitioning & custom metadata for dlt assets, and row-count metadata for Polars IO managers.
UI enhancement: clicking nodes in the asset lineage tab and external asset links now opens them in a new tab.
Added detailed backfill daemon logging, performance improvements for data version handling, and standardized table row count metadata across integrations.
Added UI enhancements: Asset Checks tab now plots numeric metadata, Asset Events support infinite scrolling, tags can be attached to materializations/observations, and various UI fixes (scrollbar alignment, filter URLs, asset‑graph query...
Added start/end offset support to TimeWindowPartitionMapping for mismatched upstream/downstream partitions, and owners on @graph assets; Dagster Plus branch deployments now accept a base deployment for change tracking.
Added several UI enhancements: automatic system‑dark mode sync, new default Geist fonts, experimental navigation toggle, asset health overview page, feedback sharing, and virtualized team member list for scalability.
Performance boost for large asset graph loading; new unit‑testable @asset check functions, UPathIOManager.load_partitions, and expanded Polars support (multi‑partition, non‑string columns).
Added experimental `dagster dbt project prepare` CLI and new `DbtProject` class for managing dbt manifests, replacing `DbtArtifacts` (breaking change)
Asset definitions now support tags, owners, and optional I/O managers; the Asset Details UI is revamped with an Overview tab and a unified “Unsynced” status view.
Resolved bugs affecting column lineage metadata generation in Dagster DBT.
Fixed crash during asset job construction when an asset depended on a subset of a parent multi‑asset's keys.
Added description field to AssetCheckResult with UI display, extended AssetSelection.checks to accept AssetCheckKeys, and introduced TimestampMetadataValue type.
Fixed an error in Dagster UI and `dagster dev` when loading jobs created with op or asset selections.
Improved scheduler latency and added configurability for max concurrent runs in Dagster Cloud.
Added Dagster OpenAI library and new UI enhancements like tooltip dates for logs, persistent search cache, and alphabetically sorted asset graph groups.
Fixed SlingConnectionResource error when connecting to a database (dagster embedded elt).
Added a configurable run retry setting for op/asset failures and op concurrency‑aware run dequeuing support.
Added Python 3.12 support, new dagster-polars integration, and expanded dbt, DuckDB, and observable asset capabilities.
Deterministic lexicographical ordering for partitioned run submissions; added experimental @observable source assets returning ObserveResult, new AutoMaterializeRule.skip, and run‑priority handling for global op/asset concurrency slots.
Added new core features: static partitioned job schedules, enhanced PipesK8sClient (auto‑detect namespace, multi‑container context injection, init container handling), Snowflake Pandas I/O manager using write pandas, sensor UI filter, La...
Added pendulum 3.0 support, downstream partition mapping, and new asset check fields; introduced Kubernetes Helm chart enhancements for initContainers and RBAC, plus an experimental UI for asset auto‑materialize history
Removed the warning for unloadable sensors and schedules in the UI and added a button to reset their status to the default code definition
Added experimental option to hide user code errors in the UI via DAGSTER_REDACT_USER_CODE_ERRORS environment variable.
Revamped asset lineage UI with left‑to‑right layout, searchable sidebar, collapsible groups, dark mode, and UI‑based materialization reporting
Viewing logs for sensors and schedule ticks is now generally available, enabled by default and visible even on failures, with the feature flag removed.
Added new SensorEvaluationContext properties (last start time, first‑tick flag) and widened @sensor asset selection to accept AssetDefinitions, strings, or AssetKeys.
Fixed EnvVars handling in Sling source/target configs for Dagster Embedded ELT.
Added UI enhancements: asset graph now shows active filters, can filter by compute kind, supports group node collapse/expand, includes experimental dark mode and removed tree view.
Added a new MetadataValue.job type, improved backfill UI to show asset statuses after partition changes, and introduced schema precedence for BigQuery, DuckDB, and Snowflake I/O managers; also added shell log exclusion and static DBT tar...
UI and asset graph enhancements: collapsing asset groups, keyboard navigation, right‑click actions, job search by run ID, launching runs from the job actions menu, and improved Users page performance.
Fixed error when directly invoking a run status sensor with resources.
Added partition‑key properties to OpExecutionContext and AssetExecutionContext, introduced a faster "tight tree" asset graph layout, top‑level partition filter in Runs table, and support for dbt core 1.7 and per‑asset DB IO manager schem...
Added new configuration options for PipesK8sClient, expanded DBT asset capabilities (partition mapping, self‑dependencies, scaffold profile), UI backfill preview, and experimental global op concurrency slot release.
Added Pydantic 2 model support for Config/Resources, a new report‑asset‑observation REST endpoint, and performance improvements for partition mapping; UI now offers quick "Add to filter" for "Launched by" tags.
Added new REST API for report asset checks, S3 context support, configurable dbt executable, DuckDB connection config, UI performance improvements, and faster asset graph layout algorithms.
Added new alert policies on assets, a configurable live‑data poll rate, materialize‑changed assets dialog, AMP Timeline page, and a report asset‑materialization REST endpoint;
Fixed asset backfill error when using multi‑run backfill policies.
Automatic inference of dependencies between time‑partitioned and multi‑partitioned assets and viewport‑only asset graph fetching for smoother navigation of large graphs.
Added new execution‑dependency APIs (AssetDep, MaterializeResult, AssetSpec) and experimental asset checks, plus customizable AutoMaterializePolicies.
DbtCliResource now enforces a minimum dbt core version, respects DBT_TARGET_PATH, and stores artifacts in unique subdirectories.
Added a loading spinner for search inputs on Overview pages while code locations load
Added support for passing @multi asset definitions to the new `deps` parameter and introduced an optional 64‑bit storage‑id migration (may require downtime).
Added a tooltip to asset runs UI and fixed navigation when re‑executing a run from its page.
Added ability to call OpExecutionContext.add output metadata multiple times per output and introduced experimental features: dbt tests as asset checks, @graph asset checks, and freeing global op concurrency slots via a deployment setting.
Added context asset key property, log level flag for dagster commands, Airbyte poll interval, Databricks OAuth and init script support, and performance improvements to the auto‑materialize daemon.
Added graceful shutdown for Dagster code servers and new ‑‑op-selection flag for `dagster execute job` CLI; UI reload shortcut (Alt+R) on OSS.
- Fixed static file loading issue on Windows in Dagster webserver
Fixed missing UI icons issue.
Added a new @partitioned config decorator and renamed ConfigurablePickledObject* IO managers to simplified names (S3PickleIOManager, ADLS2PickleIOManager, GCSPickleIOManager), deprecating the old classes.
Added experimental option to respect materialization data versions for auto‑materialization
Added support for multi-output ops/assets returning None and allowed SensorResult skip reasons as strings; introduced per‑step K8s config, extra run‑pod label, UI tag‑visibility toggle, and dbt scaffold now creates multiple files with ne...
Added config parameter to @graph asset, dynamic partitions store argument, and partitionsByAssets backfill support; extended dbt core support to 1.6, added profiles dir to DbtCliResource, and added restart policy for k8s jobs.
- Added several UI enhancements for auto‑materialized assets, including tags, history view updates, DAG links, and auto‑observation indicators.
Added support for using existing dbt `profiles.yml` when scaffolding a Dagster DBT project and changed the multiprocessing executor sentinel default to `None`.
Fix bug in dagster‑dbt that prevented correct materialization of dbt asset subselections.
Fixed bug in Dagster dbt integration that prevented efficient loading of dbt projects from a manifest
Added UI enhancements: Auto‑materialize history tab, Assets Overview dashboard, default config auto‑populate in the launchpad, and renamed Dagit to Dagster UI.
Added stable DynamicPartitionsDefinition and SensorResult, new DagsterInstance methods for partition status and code versions, build asset context, expanded AssetSelection APIs, Pandas 2.x support, and numerous UI and dbt CLI enhancements.
Fixes a bug in the Dagster project example that prevented proper downloading of examples
Added optional name argument, direct asset key via @asset, AssetKey.with_prefix, performance boost for AutoMaterializePolicy with many partitions, and experimental global concurrency limits
Lazy auto-materialize policies now skip missing assets that aren't needed for downstream freshness, and UI now shows policy skip conditions.
Added new configuration options and APIs, including Pythonic Config classes for @configured, AssetExecutionContext, enhanced TimeWindowPartitionMapping, and dbt asset freshness/auto‑materialize via +meta.
Fixed import error that caused the Dagster Cloud agent to fail during startup in the previous 1.3.8 release.
Added multipartitioned asset dependencies, asset job hooks argument, UI revamp with better filtering, and Helm chart instance config support.
Added multiple features: .env files can now set Dagster system variables and env keys, the GraphQL Python client accepts RunConfig objects, a UI button to cancel asset backfills, Polars tag display on the asset graph, and a setting to au...
Added experimental dagster code server start command, new sensors.num_submit_workers config, and expanded k8s job executor to launch each step in its own pod; UI now supports filtering schedules/sensors by running state and includes an e...
Introduced max_materializations_per_minute on AutoMaterializationPolicy and added end-date support for daily, hourly, weekly, and monthly partitions; also added op‑name option for loading dbt assets and Helm SSL support.
Run monitoring now auto‑cancels stuck CANCELING runs and TimeWindowPartitionMapping is time‑aware, returning only current partitions.
Added source-key-prefix support for loading assets, partition-aware IO managers, and lifecycle hooks for Pythonic resources; exposed new resources (GCS, BigQuery, Snowflake) following the Pythonic system.
Added performance improvements (yielding time‑partitioned run requests, global search) and UI batch actions for sensors/schedules; introduced replica configuration for Hybrid ECS/K8s agents and zero‑downtime ECS agent updates.
Added performance improvements for evaluating time‑partitioned run requests in sensors and schedules, and for UI loading of the asset catalog and launchpad when many partitioned assets are present.
Introduced AutoMaterializePolicy to replace asset reconciliation sensors and made Pythonic config/resources stable, with resources now auto‑bound to jobs (breaking change).
Added Pythonic resource support for assets and introduced new DuckDB, BigQuery, and Snowflake IO managers following the resource system; renamed Resource annotation to ResourceParam for upcoming 1.3 feature.
Added a selection argument to materialize and materialize_to_memory, and made MultiPartitionsDefinition stable (no longer experimental).
Added experimental source asset conversion, Multi‑partition mapping, and performance boosts for asset reconciliation and large backfills.
Define asset jobs now auto‑infer partition definitions and allow observable source assets to be scheduled, plus HookContext gains an instance property for accessing the current DagsterInstance
Added opt‑in BindResourcesToJobs, new CLI asset list/materialize commands, support for MultiPartitionsDefinition with time dimension, SpecificPartitionsPartitionMapping, source asset loading, I/O manager enhancements (BigQuery timeout, s...
Fixed daemon heartbeat failures in Postgres storage caused by missing instance migration after upgrading to 1.2.0.
Added a dagster dev command, experimental Pydantic config/resources, new integrations (Weights & Biases, Snowflake + PySpark, BigQuery, Airflow migration focus, Databricks op factories) and UI/asset enhancements like freshness policies, ...
Add backfill support for asset selections with mapped partitions and display of asset staleness causes in Dagit
Introduced @graph asset and @graph multi‑asset decorators that auto‑infer dependencies between single‑dimensional and multipartitioned assets, plus a Test button for dry‑run sensors and schedules.
Added timezone support to FreshnessPolicy, new asset utilities (from_graph with freshness by output, to_source_asset, assets as inputs to ops), and @asset sensor SkipReason; introduced dagit host/port arguments.
Added FreshnessPolicy support for time‑window partitions and made dbt materializations emit events during execution; run monitoring now includes ECS task logs for crashed runs.
Dagster Airflow library upgraded to 1.x.x stable, adding schedule/job creation from Airflow DAGs and converting task retries to op RetryPolicies.
Added support for Executor instances in Definitions and introduced a request_assets param for @multi_asset sensors, plus performance improvements for large asset sets and new Docker and Snowflake‑PySpark ops;
Enable materializing entire large asset graphs in Dagit without pre‑selecting a subset
Fixed the nux section of dagster.yaml configuration.
Fix call stack size exceeded error in Dagit when viewing a root asset materialization
Added a `dagster dev` command for running Dagit and the daemon together, new M1/M2 install guide, and repository‑definitions args to backport to the Definitions API.
Added list support for asset selections in define_asset_job and a stale‑assets‑only flag for RunRequest, enabling more precise scheduling and sensor behavior.
Asset backfills now respect partition mappings, add tag‑based concurrency limits, and materialize/result APIs support tags and asset‑value fetching.
Definitions API is now the preferred, non‑experimental entry point and the Dagit Workspace UI is replaced by a Deployment/Code Locations view.
Added an experimental Definitions entrypoint for tools/UI and a new Slack sensor for freshness policy status changes (feature).
Fixed Partitions tab load failure for asset jobs in Dagit
Added many new features: optional job cancellation, Snowflake role config, AWS Systems Manager resource, dagstermill retry policy and notebook yield/save options, dbt Cloud partition/subset/group/AssetKey support, and an ephemeral Airflo...
Fixed duplicate run submissions in asset reconciliation sensor when assets have differing partitioning from parents
Fixed assets incorrectly marked as “Stale” after upgrading to 1.1.1 in Dagit
Added multi‑dimensional partitioning, asset versioning, freshness policies, and new asset reconciliation and multi‑asset sensor APIs for richer asset management
Added asset selection parameter for sensors and raise_on_error option for materialize ops; introduced multi‑dimensional asset partitions with optional schema migration.
New Overview and Workspace pages with virtualized tables, pagination, and improved navigation are now the default UI in Dagit.
Added UI improvements in Dagit: run timeline now displays all future schedule ticks, asset graph auto‑refreshes on materializations, and shift‑click Materialize opens the asset launchpad for partitioned assets; a feature flag for new ass...
Asset reconciliation sensors now accept tags; graph inputs auto‑infer DagsterType when inner inputs share a type; Airbyte asset loading caches project data and generates schema metadata at definition time.
Added metadata dict to AssetMaterialization, get_latest_materialization_event, InMemoryIOManager to public API, TableSchema.from_dict constructor, and new CLI command `dagster run migrate repository`.
Added experimental multi‑asset sensor with AssetSelection, made AssetValueLoader public, and added partition‑key support to repository load methods; introduced new AWS ECS launcher (Fargate/EC2), Airbyte asset loader, Airflow Dagster ope...
Added RepositoryDefinition.load_asset_value and partition‑key retrieval methods for notebooks and op contexts.
Run status sensors can now monitor all runs across the Dagster instance (monitor_all_repositories=True) and the run key on RunRequest/partition is optional.
Experimental multi‑asset sensor now tracks asset partitions via latest materialization records; added experimental Airbyte helpers for auto‑generating assets, log forwarding control, and normalization tables.
Added arbitrary and multiple cron expression support to TimeWindowPartitionsDefinition for flexible scheduling.
Improved Dagit run timeline UI (persistent time window, clearer “Now” label, and separate upcoming ticks) and added asset job launch via “Materialize All” with optional Shift‑click config; launchpad tab removed for asset jobs
Added nbconvert as an extra for Dagit, multi‑asset sensor (experimental), run‑status sensors for external repos, and asset selection support in RunRequests.
Added conditional asset materialization with `output required=False`, sensor partition request support, and runtime type validation for `AssetIn`.
Added retry control to Failure, a cron schedule method to TimeWindowPartitionsDefinition, and an optional key argument to GCSFileManager.
Added workspace update notifications and auto-refresh in Dagit.
Fix warning logs when Dagster libraries incorrectly reported version mismatches
Comprehensive docs site overhaul and new production/deployment guides.
Added support for non-dict asset config schemas and updated OpDefinition and dagstermill op constructors to accept `ins`/`outs` for easier direct construction.
Added job name property to DagsterRun, partition range method to TimeWindowPartitionsDefinition, and asset partition definition helpers to OpExecutionContext; also build output context now accepts a partition key.
Dagit UI improvements: full exception chain display, global asset graph view, simplified CSP, and faster partition/backfill pages.
Added new UI features: global asset lineage view with depth control, collapsed DAG edges, asset launchpad for configurable resources, and default routing to asset groups.
Reverted the sensor threadpool changes introduced in 0.15.3 to restore daemon stability
Added custom input manager support for AssetIn/In, nested list support in repository definitions, data retention and sensor daemon thread‑pool settings, and partition key arguments for materialize/materialize_to_memory and resource_defs ...
Fixed asset dependency resolution when two assets in the same group shared the same name.
Added placeholder events for unknown types and introduced a default executor definition on repositories; AssetsDefinition.from_graph now accepts a group name and a new AssetsDefinition.from_op utility for asset creation.
Software‑defined assets are now stable, support grouping, configuration, and can be defined from graphs; new APIs for asset jobs, resource attachment, and unified storage config were added.
- Added flexible environment variable support to ECS and Kubernetes run launchers, including name=value syntax and inclusion of image pull secrets when deploying runs.
Added a job metadata field and expanded Dagit launchpad to show detailed config state (errors, scaffolding)
Added Helm chart support to override Dagit deployment name and pinned protobuf to v3 for compatibility with protobuf v4 (breaking change).
Added partition support to AssetsDefinition.from_graph, enabled variable kwargs for @asset functions, introduced ECS task health reporting, and expanded dbt resource configuration (docs URL and custom cloud host)
Sensors/schedules can now return a list of RunRequest objects and OpExecutionContext now includes run, job, and partition window fields, with added partitions support in AssetsDefinition.from_graph.
Config schema view in Dagit now shows default values on hover, and Dagster libraries include py.typed files for static type checking, potentially surfacing new type errors.
Added new features: K8sRunLauncher now supports resource requests/limits via helm chart and can inherit namespace/service account; @asset decorator accepts op tags; EMR PySpark step launcher gains dynamic orchestration, retry policies, a...
Fixed Launchpad in Dagit from opening in an empty state
Fixed scheduler timeout when schedules from partition sets launch multiple runs per tick
Added an `includeConfigInLaunchedRuns` flag to the Helm chart for automatic inclusion of configmaps, secrets, and volumes in user code deployments.
Added multiple asset and UI enhancements such as a new GraphQL assetMaterializations field, AssetGroup.materialize method, async asset graph rendering, and new Dagit controls (copy config, open in Launchpad, bulk selection, persistent wh...
MySQL storage is now production-ready, run IDs can be passed to execute_in_process, and InputContext supports adding input metadata visible in LOADED INPUT events and AssetObservations.
Added telemetry toggle, new Azure ADLS IO manager, and custom partition start times with enhanced PartitionedConfig (run tags and get‑run‑config support).
Added a GCS IO manager for assets, enhanced k8s worker labeling, and extended the Databricks step launcher to support retry policies.
Fixed sensor toggle issue in Dagit UI.
Added support for non‑standard @hourly/@daily cron strings and expanded asset metadata capabilities (typed metadata on SourceAssets, InputContext config, AssetGroup loading).
Executor processes now surface Dagster system exceptions; SecretsManager resource added; Jobs gain a run_request_for_partition method; PipelineRunsFilter renamed to RunsFilter (breaking).
Added support for invoking new Run status sensors in unit tests and introduced a build run sensor status context
Sensor tick timeline in Dagit now displays cursor values in tooltips when available.
Introduced mature software‑defined assets with partitioning, cross‑repo graph view, ad‑hoc materialization, dbt model mapping, and new AssetObservation events
Added various performance improvements for asset graph views in Dagit.
Added subgraph op selection syntax, new Map ConfigType, build_resources API, and expanded GraphQL and asset metadata capabilities (including list‑based namespace and new Table metadata types).
Dagit UI enhancements and performance boosts: Runs page speed, asset entry page loading, job name truncation, execution timezone display, quoted graph filter defaults, and richer asset graph rendering.
Add Airbyte integration under dagster-airbyte and allow ops with a config schema to omit the context argument
Added callable PartitionedConfig objects, support for negative end offsets, and flexible runConfigData input (JSON string or object) for the launchRun GraphQL mutation.
Added streaming of dbt and Databricks logs, support for dynamically mapped Databricks steps, and tighter scheduler polling; introduced k8s job labels
Added structured Python logging with timezone timestamps for dagit and daemon processes, and startup/shutdown messages for custom gRPC servers.
Performance and UI enhancements in Dagit – faster Run page, shareable URLs for sensor/schedule ticks, spinners and dot icons for in‑progress asset jobs, and fixed asset catalog links.
Added top‑level job inputs, partitioned job support, and op retry policies; extended build hook context and schedule decorator with RunRequest.
Memoization now compatible with multiprocess executor.
Added dbt Cloud integration with a pre‑built op/resource, richer error messages, UI links, and automatic asset materializations; @asset decorator now accepts a Dagster type and build_assets_job can specify an executor.
Improved load speed for the Runs page in Dagit.
EcsRunLauncher now tags ECS tasks only when the AWS account has long ARN format enabled
Added a new Dagster Fivetran integration with a pre-built sync op and flexible resource.
k8s job executor is now production‑ready, replacing the experimental version and recommended for isolated workloads; celery executor remains optional
Fixed a bug causing retries to fire twice in the Kubernetes job executor.
Added a dagstermill op factory, top‑level `get_dagster_logger`, dbt LS support, ins/outs on OpDefinition, and a new Docker executor with daemon run‑worker monitoring.
Add ls command support to all dbt resources (cli, rpc, and rpc sync).
Introduced new core job/op/graph APIs replacing pipelines/solids, deprecating legacy APIs and providing a migration guide.
Add support for configuring GCSComputeLogManager credentials via a string or environment variable.
Community contributions: updated click version and corrected a typo.
Updated VersionStrategy API to accept SolidVersionContext and ResourceVersionContext, breaking existing experimental usage
Added Microsoft Teams integration with connection resource and messaging support (feature).
Added TLS support to Helm ingress and fixed Dagit empty workspace configuration; updated Dagit fonts.
Added configurable image pull policy for KubernetesRunLauncher, GitHub Enterprise support, and hostname config for the GitHub resource
Added support for specifying service accounts via Kubernetes tags and now show an error dialog with stack trace when a repository location fails to reload in Dagit.
Added RunStatusSensorContext for accessing Dagster Instance in run status sensors and improved Dagstermill input handling, enabling proper reruns of notebooks; Dagit CLI now supports targeting bare graph files and markdown tables render ...
- Moved repository locations list to the Workspace page with a warning icon for errors; logging calls now fully follow the Python logging API, breaking non‑standard keyword arguments in context.log.
Added a synchronous RPC dbt resource and changed secret merging and default local file manager path
Fixed numerous Dagit UI and CLI bugs, including timeline tick display, backfill progress stats, sensor/schedule listings, execution of similarly named pipelines, and large DAG rendering loops, as well as config handling errors for jobs a...
Updated Kubernetes integration: default to NoOpComputeLogManager, added env‑var support to K8sRunLauncher, and introduced customizable TTL/backoff limits for jobs.
Added a strict Content Security Policy for Dagit and narrowed ECS role permissions for improved security.
Improved Asset catalog load times in Dagit for fully migrated Dagster instances.
Fix implementation issues in @pipeline failure sensor that prevented it from working.
Added first‑class pipeline‑failure and asset sensors, solid‑level retry policies, and direct‑invocation test APIs for solids, resources, hooks, loggers, sensors, and schedules.
Added a new user‑settings page in Dagit (feature flags, timezone) reachable via the gear icon.
Added shutdown repository location API and experimental EcsRunLauncher; schedules can now be directly invoked for testing