No user-submitted related servers found.
Scored across 9 tools
Most tools have a crisp, distinct role (train, predict per model, evaluate, select, explain, export). The only overlap is predict_both versus predict_logistic_regression/predict_knn, but its description clearly frames it as a convenience wrapper that also reports agreement, so misselection risk is low.
Names are uniformly snake_case and mostly verb-first (train_models, predict_knn, select_best_model, get_sample_input, export_parquet_audit). The lone deviation is the noun-first confusion_matrix_metrics, and shap_explain_selected stacks verbs, but both remain readable and predictable.
Nine tools map cleanly onto a single ML workflow (train → sample input → predict → evaluate → select → explain → export), with no padding. This is squarely in the well-scoped 3-15 range.
The surface covers the full lifecycle: training, per-model and combined prediction, metrics, model selection, SHAP explanation, and audit export. Minor gaps exist around inspecting or listing saved model artifacts and resetting/retraining with custom configuration, but agents can work around these via train_models.