GitHub Actions composite action for running dbt in CI/CD pipelines. Currently supports Snowflake and is structured to add more adapters incrementally.

- uses: ta93abe/dbt-jobs@v1
  with:
    type: ci
    adapter: snowflake
    command: |
      dbt build --select state:modified+
    deferral: main
    post-pr-comment: true

Runs commands with set +e to capture all failures, optionally posts a PR comment with build results, then propagates the failure.

Runs commands with set -e (fail-fast) and uploads manifest.json + run_results.json as artifacts for Slim CI deferral.

Same as merge, plus optional source freshness checks.

Each line in the command input gets flags injected automatically (if not already present):

• --profiles-dir / --project-dir -- always
• --target / --threads -- if configured
• --defer --state ./prod-manifest -- if deferral is set, manifest was downloaded, and command contains state:

dbt-core version is resolved in this order:

1. Explicit input -- dbt-version: "1.8.0"
2. pyproject.toml -- PEP 621 dependencies or Poetry [tool.poetry.dependencies]
3. requirements.txt -- dbt-core>=1.7.0
4. setup.cfg -- dbt-core>=1.7.0 in install_requires
5. setup.py -- dbt-core>=1.7.0 in install_requires
6. Pipfile -- dbt-core = "==1.7.0"
7. Latest -- if no version is found

Adapter setup is centralized in actions/setup/action.yml. When adding a new adapter, extend the adapter normalization/validation case and map it to the correct package name.

Set credentials as environment variables at the workflow level. Point profile-dir to a profiles.yml in your repository that references them via env_var().

env:
  SNOWFLAKE_ACCOUNT: ${{ secrets.SNOWFLAKE_ACCOUNT }}
  SNOWFLAKE_USER: ${{ secrets.SNOWFLAKE_USER }}
  SNOWFLAKE_PRIVATE_KEY: ${{ secrets.SNOWFLAKE_PRIVATE_KEY }}

See the examples/ directory for complete workflow configurations:

• CI (Pull Request) -- Slim CI with PR comment
• Merge -- Post-merge build with artifact upload
• Deploy -- Scheduled production deploy

MIT