dbt-dagsterizer

dbt-dagsterizer is a Python package for building Dagster automation from dbt metadata.

It is designed to keep Dagster code locations mostly static, while letting developers declare orchestration intent in a small, reviewable YAML file alongside the dbt project.

Two common ways to use it

1. As a CLI tool (bootstrap + maintain dagsterization.yml)

• Install once:

uv tool install dbt-dagsterizer

• Upgrade later:

uv tool upgrade dbt-dagsterizer

This is the recommended workflow for running dbt-dagsterizer project ..., dbt-dagsterizer meta ..., and dbt-dagsterizer macros ... from any repo.

2. As a Python dependency (Dagster runtime imports it)

Dagster code locations typically import dbt_dagsterizer at runtime (for example build_definitions()), so the Dagster project itself must depend on dbt-dagsterizer (for example in its own pyproject.toml). Installing the CLI as a tool does not automatically make it importable inside that project’s runtime environment/container.

Documentation

• Docs index: docs/README.md
• Getting started: docs/getting-started.md
• CLI reference: docs/concepts/cli.md
• Execution model and env propagation: docs/concepts/execution-model.md
• Release notes: RELEASE_NOTES.md
• Template docs (Dagster + dbt + StarRocks): docs/templates/.../README.md

Quick start

CLI

dbt-dagsterizer --help

Initialize orchestration intent and refresh the dbt manifest:

dbt-dagsterizer meta init --parse
dbt-dagsterizer meta validate --prepare

Python

from dbt_dagsterizer.api import build_definitions

defs = build_definitions(dbt_project_dir="./dbt_project")

If the project has no dbt models yet (no models/**/*.sql), build_definitions() still returns a minimal, always-loadable Definitions.

Kubernetes run pod env injection

When using Dagster’s K8sRunLauncher, runs execute in separate Kubernetes Job pods. If your Dagster code-location Deployment uses envFrom (for example a per-code-location ConfigMap/Secret containing DBT/StarRocks credentials), those variables do not automatically propagate into run pods.

To inject a code-location ConfigMap/Secret into run pods created by jobs defined in dbt-dagsterizer, set these environment variables on the code server Deployment:

• LUBAN_RUN_ENV_CONFIGMAP={{app_name}}-config (optional)
• LUBAN_RUN_ENV_SECRET={{app_name}}-secret (optional)

This adds a per-job dagster-k8s/config tag that configures container_config.envFrom for the run pod container. If both variables are unset/empty, no tag is added.

Notes:

• The referenced ConfigMap/Secret must exist in the same Kubernetes namespace where the run pod is launched.
• If you explicitly set a dagster-k8s/config tag on a job, it takes precedence over this env-based injection.
• If you use an executor that launches steps in their own pods (for example k8s_job_executor), job-level configuration may not apply to step pods.

Development

This section is for developing dbt-dagsterizer itself. If you are using it in another repo, start with the CLI install instructions or add it as a dependency in your Dagster code location.

Setup:

uv sync --dev

Run tests:

uv run pytest

Lint:

uv run ruff check .