Framework Deployment Options

The Lakeflow Framework can be deployed to a Databricks workspace in three ways. Choose the mode that matches how your team manages dependencies and configuration.

Mode

How it works

When to use

Flat DAB deploy (default)

The framework repository is cloned and deployed with databricks bundle deploy. framework.sourcePath points to the deployed src/ directory on workspace files; the cluster reads Python modules and default config directly from there.

Default for all existing customers. No change to existing pipelines required.

Wheel install

pip install lakeflow-framework installs the package into the cluster Python environment. Default configs and schemas are bundled inside the wheel via importlib.resources.

Teams that manage Python dependencies via PyPI, a UC Volume, or an internal Artifactory feed.

Wheel + local overlay

Wheel installed, plus framework.sourcePath still set so that src/local/config/ sparse overrides are deep-merged on top of the bundled defaults.

Teams that want pip-managed installs but need per-deployment config customization.

Note

For wheel based deployments it is your responsibility to provide an entry-point notebook and a location for local config overrides and libraries/init scripts when moving away from the default flat deploy.

See also