from transformers import PretrainedConfig
from typing import List, Optional


class KNNConfig(PretrainedConfig):
    """
    Minimal Transformers-style config for a scikit-learn KNN model.

    This stores only metadata needed to describe the model on the Hub.
    
    For ensemble models (7T-21T, Synthetic), is_ensemble=True and 
    ensemble_members lists the sub-model filenames.
    """

    model_type = "knn"

    def __init__(
        self,
        n_neighbors: int = 3,
        metric: str = "euclidean",
        feature_names: Optional[List[str]] = None,
        is_ensemble: bool = False,
        ensemble_members: Optional[List[str]] = None,
        data_source: Optional[str] = None,
        training_version: Optional[str] = None,
        **kwargs,
    ):
        self.n_neighbors = n_neighbors
        self.metric = metric
        self.feature_names = feature_names or []
        self.is_ensemble = is_ensemble
        self.ensemble_members = ensemble_members or []
        self.data_source = data_source
        self.training_version = training_version
        super().__init__(**kwargs)