dbt

Official Dagu action for running dbt Core commands with uv.

This action keeps dbt and adapter packages out of the Dagu core binary and worker image. The action pins uv with Dagu tools; each invocation installs the requested dbt Python requirements into an isolated uv run environment.

Usage

type: graph

steps:
  - id: dbt_build
    action: dbt@v1
    with:
      projectDir: /data/jaffle-shop
      profilesDir: /data/dbt-profiles
      requirements:
        - dbt-duckdb==1.10.1
      command: build
      target: prod
      select:
        - tag:nightly
      targetPath: ${DAG_RUN_ARTIFACTS_DIR}/dbt-target

  - id: print_summary
    depends: [dbt_build]
    run: |
      echo "dbt exit code: ${dbt_build.outputs.exitCode}"
      echo "dbt run results: ${dbt_build.outputs.runResultsPath}"

projectDir must point to the directory containing dbt_project.yml on the worker that runs the action. Remote actions run in their own action workspace, so pass an absolute path or a path that is meaningful on that worker.

Commands

The command input defaults to build.

Supported commands:

Command	Notes
build	Runs models, tests, snapshots, seeds, and supported functions in DAG order.
run	Runs selected models.
test	Runs tests.
seed	Loads seed CSV files.
snapshot	Runs snapshots.
compile	Compiles project SQL without running models.
deps	Installs dbt package dependencies.
debug	Checks project and connection configuration.
parse	Parses the project and writes artifacts.
list / ls	Lists project resources.
show	Previews rows for a selected resource.
run-operation	Invokes a macro. Pass the macro name in args.
source freshness	Checks source freshness.
docs generate	Generates dbt docs artifacts.
clean	Deletes dbt-generated artifacts.

Examples:

steps:
  - id: deps
    action: dbt@v1
    with:
      projectDir: /data/jaffle-shop
      requirements:
        - dbt-duckdb==1.10.1
      command: deps

  - id: changed_models
    action: dbt@v1
    with:
      projectDir: /data/jaffle-shop
      profilesDir: /data/dbt-profiles
      requirements:
        - dbt-duckdb==1.10.1
      command: build
      select:
        - state:modified+
      state: /data/dbt-state
      defer: true

Python Requirements

Use requirements for pip-compatible dbt packages.

The default is:

requirements:
  - dbt-core

Most real projects need an adapter package. Pin exact package versions for production workflows:

requirements:
  - dbt-duckdb==1.10.1

If the adapter already pins a compatible dbt-core, you can provide only the adapter:

requirements:
  - dbt-postgres==1.9.0

For private package indexes or credentials, pass the relevant uv or pip environment variables through env, such as UV_INDEX, UV_DEFAULT_INDEX, or UV_INDEX_URL.

Profiles and Secrets

Use profilesDir when profiles.yml is outside the dbt project:

steps:
  - id: dbt_run
    action: dbt@v1
    with:
      projectDir: /srv/dbt/project
      profilesDir: /srv/dbt/profiles
      requirements:
        - dbt-postgres==1.9.0
      command: run
      target: prod
      env:
        DBT_ENV_SECRET_PASSWORD: ${DBT_ENV_SECRET_PASSWORD}

The action passes env values only to the dbt process. Secrets are still governed by the Dagu worker environment and Dagu secret masking.

Selection and Flags

Structured inputs map to common dbt flags:

steps:
  - id: dbt_build
    action: dbt@v1
    with:
      projectDir: /srv/dbt/project
      profilesDir: /srv/dbt/profiles
      requirements:
        - dbt-bigquery==1.10.0
      command: build
      target: prod
      select:
        - tag:nightly
        - path:models/marts
      exclude:
        - config.materialized:incremental
      vars:
        run_started_by: dagu
      threads: 4
      failFast: true

Use args for command-specific flags that do not have a first-class input:

steps:
  - id: vacuum
    action: dbt@v1
    with:
      projectDir: /srv/dbt/project
      profilesDir: /srv/dbt/profiles
      requirements:
        - dbt-postgres==1.9.0
      command: run-operation
      args:
        - vacuum_table
        - --args
        - '{"relation": "analytics.events"}'

Artifacts

The action captures dbt stdout/stderr into structured outputs and mirrors them to the step log. It also reports known dbt artifact paths when those files exist:

Output	Description
manifestPath	Path to manifest.json under targetPath.
runResultsPath	Path to run_results.json under targetPath.
catalogPath	Path to catalog.json under targetPath.
sourcesPath	Path to sources.json under targetPath.

Set targetPath under ${DAG_RUN_ARTIFACTS_DIR} when you want dbt artifacts to be stored with the DAG run:

steps:
  - id: dbt_docs
    action: dbt@v1
    with:
      projectDir: /srv/dbt/project
      profilesDir: /srv/dbt/profiles
      requirements:
        - dbt-duckdb==1.10.1
      command: docs generate
      targetPath: ${DAG_RUN_ARTIFACTS_DIR}/dbt-target
      logPath: ${DAG_RUN_ARTIFACTS_DIR}/dbt-logs

targetPath defaults to ${DAG_RUN_WORK_DIR}/dbt-target when that runtime variable is available. logPath defaults to ${DAG_RUN_WORK_DIR}/dbt-logs.

Inputs

Field	Required	Default	Description
projectDir	Yes	-	Directory containing dbt_project.yml.
command	No	build	dbt command to run.
requirements	No	["dbt-core"]	Pip requirement specifiers installed with uv run --with.
profilesDir	No	-	Directory containing profiles.yml.
profile	No	-	dbt profile name.
target	No	-	dbt target name.
select	No	-	List of --select arguments.
exclude	No	-	List of --exclude arguments.
selector	No	-	YAML selector name passed with --selector.
vars	No	-	Value passed with --vars. Objects are encoded as compact JSON.
threads	No	-	Thread count passed with --threads.
state	No	-	State directory passed with --state.
targetPath	No	run work dir	Directory passed with --target-path.
logPath	No	run work dir	Directory passed with --log-path.
fullRefresh	No	false	Pass --full-refresh.
failFast	No	false	Pass --fail-fast.
warnError	No	false	Pass --warn-error.
defer	No	false	Pass --defer.
quiet	No	false	Pass --quiet.
args	No	[]	Additional raw dbt CLI arguments appended after structured options.
env	No	-	Extra environment variables exposed to the dbt process.
pythonVersion	No	3.13.9	Python version used by uv for the dbt process.
timeoutSeconds	No	3600	Timeout for the dbt process.

Outputs

Field	Description
ok	true when dbt exits with code 0.
exitCode	dbt process exit code.
command	Full argv used to invoke dbt through uv.
stdout	dbt stdout.
stderr	dbt stderr.
durationMs	Wrapper-measured dbt process duration in milliseconds.
targetPath	Effective dbt target path.
logPath	Effective dbt log path.
manifestPath	Present when manifest.json exists.
runResultsPath	Present when run_results.json exists.
catalogPath	Present when catalog.json exists.
sourcesPath	Present when sources.json exists.
error	Validation or wrapper error object when the wrapper fails before dbt runs.

If dbt exits non-zero, the action step fails after writing the structured output payload.

Local Development

Use source: to call a local checkout:

steps:
  - id: dbt_build
    action: source:file:///path/to/dbt@local
    with:
      projectDir: /tmp/jaffle-shop
      requirements:
        - dbt-duckdb==1.10.1
      command: build

Security

This action is not a sandbox. dbt runs with the same worker permissions, filesystem access, network access, and secrets available to the Dagu worker process. Only run trusted dbt projects and packages.

Package Layout

dagu-action.yaml
workflow.yaml
scripts/
  run_dbt.py
examples/
  basic.yaml