v0.2.0 — ModelWrapper Architecture and CLI Support
This release introduces the ModelWrapper
abstraction system and shared model state across all assumption checks.
✨ Highlights
- Added
BaseModelWrapper
andLinearModelWrapper
- Only fit one model per dataset, shared across all assumptions
- New CLI flag:
--model-type
(currently supports "linear") - Rich console output now includes model metadata
- Refactored
linearity.py
to use wrapper-based architecture - Assumption checks can now declare supported model types via
@register_assumption(model_types=...)
🛠 Internal Improvements
- Dispatcher now returns both results and model
- Parameter orders standardized across assumption checks
- Improved crash safety and default fallbacks