datahub-yaml-source

A standalone DataHub ingestion source plugin that reads a directory tree of declarative YAML "metadata as code" files and emits the DataHub entities they describe (platforms, tags, glossary, domains, containers, datasets with schema/lineage, data products, pipelines, pipeline run history, and data quality assertions).

See docs/sources/yaml/yaml.md for a narrative introduction, docs/sources/yaml/reference.md for a generated field-by-field reference of every kind, and docs/sources/yaml/yaml_recipe.yml for an example recipe. See _PLANNING.md for the architecture decisions behind this connector.

A JSON Schema for the document format (autocomplete/validation in VS Code, IntelliJ, ...) is also generated from the Pydantic models -- see the "Editor autocomplete and validation" section in yaml.md.

Both reference.md and the JSON Schema are generated from src/datahub_yaml_source/models.py; regenerate them after changing a model:

uv run python scripts/generate_json_schema.py
uv run python scripts/generate_markdown_docs.py

Installation

pip install -e .

This registers the yaml source type with acryl-datahub via a datahub.ingestion.source.plugins entry point. Verify it's picked up with:

datahub check plugins

Usage

datahub ingest -c docs/sources/yaml/yaml_recipe.yml

Development

Managed with uv:

uv sync --group dev --extra git --extra s3
uv run pytest tests/unit          # unit tests
uv run pytest tests/integration   # integration test against a curated fixture,
                                   # golden-file checked

To regenerate the integration golden file after an intentional output change:

uv run pytest tests/integration/yaml_source/test_yaml_source_golden.py --update-golden-files