Doctors API

BaseDoctor

The abstract base class that all Doctors must implement.

class BaseDoctor(ABC):
    name: str
    dimension: str
    priority: int

    @abstractmethod
    def examine(self, context: EvaluationContext) -> Diagnosis:
        pass

Built-in Doctors

• OverfittingDoctor: Evaluates train/test memorization gap.
• LeakageDoctor: Evaluates proxy feature leakage.
• PredictionDoctor: Evaluates basic f1/accuracy/r2 sanity checks.
• DataDoctor: Evaluates row/column integrity, missing values, and class balance.
• FeatureDoctor: Evaluates dimensionality and variance.
• CalibrationDoctor: Evaluates probability reliability (Expected Calibration Error).
• ProductionDoctor: Evaluates serialized model size and inference latency.
• GeneralizationDoctor: Evaluates cross-validation fold stability.