界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
开发与工程 · SKILL.md

asset-definitions

asset-definitions

【开发与工程】Model, validate, test, and troubleshoot Dagster assets, jobs, Definitions, partitions, asset checks, and local materialization workflows.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: asset-definitions description: "Model, validate, test, and troubleshoot Dagster assets, jobs, Definitions, partitions, asset checks, and local materialization workflows." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Asset Definitions

Use this sub-skill when the user is working with Dagster software-defined assets, multi-assets, asset checks, jobs, ops/graphs, partitions/backfills, Definitions, asset selection, metadata, or local definition tests.

Route Here

  • Defining @asset, @multi_asset, @asset_check, @op, @graph, or @job code.
  • Assembling Definitions with assets, jobs, asset checks, and module-loaded definitions.
  • Creating asset jobs with define_asset_job, AssetSelection, groups, tags, owners, kinds, code versions, or metadata.
  • Debugging asset dependencies, missing upstream assets, selection/subsetting, partition keys, backfill policy, or local materialization failures.
  • Testing definitions with materialize, materialize_to_memory, Definitions.resolve_job_def, and small in-memory smoke checks.

Route Elsewhere

  • Resource construction, config schemas, ConfigurableResource, IO managers, secrets, and environment variables: use ../configuration-resources/SKILL.md.
  • Schedules, sensors, automation conditions, freshness sensors, or daemon-driven automation: use ../automation-schedules-sensors/SKILL.md.
  • dagster, dg, project scaffolding, webserver, and CLI run commands: use ../cli-local-development/SKILL.md.
  • Deployment, code locations, run launchers, executors, and production operations: use ../deployment-operations/SKILL.md.

Start Here

  1. Read references/workflows.md for common asset definition, selection, partition, and testing workflows.
  2. Read references/api-reference.md for the public APIs and patterns most often needed in agent-generated Dagster code.
  3. Read references/troubleshooting.md when imports, Definitions validation, selection, materialization, or partitioned backfills fail.
  4. Use scripts/validate_defs_smoke.py --help or run it against a user module/file to perform a safe local Definitions smoke check.

Fast Triage

  • Prefer import dagster as dg in examples so public APIs are clearly namespaced.
  • Use function parameters or deps=[...] for asset dependencies; use AssetIn only when remapping input names, keys, partitions, or metadata.
  • Use Definitions(assets=[...], jobs=[define_asset_job(...)], asset_checks=[...]) as the unit loaded by tools and tests.
  • Use materialize_to_memory([...]) for small pure-Python tests; avoid it when assets require persistent IO managers or external services.
  • For unresolved asset jobs from define_asset_job, use Definitions.resolve_job_def(name) rather than assuming get_job_def returns a concrete job without resolution.
我的收藏