Model Registry and Capabilities

The dashboard and CLI expose only registered recipes with validated task and parameter metadata.

ID	Family	Tasks	Optional dependency
linear_regression	Linear baseline	Regression	None
logistic_regression	Linear baseline	Classification	None
elastic_net	Regularized linear	Regression	None
random_forest	Ensemble	All	None
extra_trees	Ensemble	All	None
hist_gradient_boosting	Boosting	All	None
svm	Kernel	All	None
sklearn_mlp	Neural network	All	None
xgboost	Boosting	All	pspso[xgboost]
lightgbm	Boosting	All	pspso[lightgbm]
pytorch_mlp	Neural network	All	pspso[torch]

Use list_estimators(task=...) to discover canonical IDs and get_estimator_info(name) to inspect capabilities, dependency status, fixed defaults, typed search spaces, and accepted parameters.

Probability-dependent metrics are allowed only for recipes that provide scores or probabilities. Native feature importance is used for tree models, coefficients for linear models, and permutation importance as an explicit fallback.