GoldenCheck plugs into the two places data-quality regressions usually slip through: warehouse models and pull requests.
dbt
Add the package to packages.yml:
packages:
- git: "https://github.com/benseverndev-oss/goldenmatch.git"
subdirectory: "packages/dbt/goldencheck"
revision: main
dbt deps. The package offers a SQL-native test macro and a Python runner.
Use the test macro in a schema file:
models:
- name: orders
tests:
- goldencheck_not_empty
python scripts/run_goldencheck.py orders --fail-on error
python scripts/run_goldencheck.py patients --fail-on warning --domain healthcare
profiles.yml, queries the model via dbt show, writes a temp CSV, runs the full profiler, and exits non-zero on findings at or above --fail-on.
Requires dbt-core 1.7+, goldencheck 0.5.0+, and Python 3.11+.
GitHub Actions
Scan CSV files in CI and post a findings summary on the pull request:
- uses: benseverndev-oss/goldencheck-action@v1
with:
files: "data/*.csv"
fail-on: error
config: goldencheck.yml
llm-boost: true
llm-provider: openai
env:
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
| Input | Required | Default | Values |
|---|
files | Yes | — | Glob pattern (for example data/*.csv). |
fail-on | No | error | error, warning. |
config | No | — | Path to goldencheck.yml. |
llm-boost | No | false | true, false. |
llm-provider | No | anthropic | anthropic, openai. |
python-version | No | 3.12 | Python version string. |
version | No | latest | GoldenCheck version. |
Outputs
| Output | Type | Description |
|---|
errors | number | Total error count. |
warnings | number | Total warning count. |
health-grade | string | Worst health grade (A–F). |