Legacy compat (habit.compat)

v2 removed the third-party interop adapters that used to live under habit.compat.monai, habit.compat.nnunet, and habit.compat.sklearn. Those modules are gone and must not be imported. Do not expect Sphinx autodoc pages for them.

Core tabular ML still uses TablePipeline (a scikit-learn Pipeline subclass) and habit.pipeline.sklearn_interop internally; those are not third-party interop surfaces.

habit.compat itself exports nothing (__all__ = []). Remaining submodules are frozen v0.1 orchestration helpers pending further migration or deletion:

  • habit.compat.engines — legacy habitat / ML / preprocessing runners

  • habit.compat.dicom_sort_runner

  • habit.compat.plugin_registries

  • habit.compat.registry_facade

  • habit.compat.legacy_core

These are not part of the v2 public API. Prefer habit.recipes, habit.pipeline, and habit.precision.align_habitat_map().