Method _resolve_vector_length

sdk/python/feast/doc_embedder.py:208–237

Determine the vector length to use for the generated FeatureView. Priority: 1. Explicitly provided vector_length 2. Auto-detected from embedder via get_embedding_dim("text") 3. Default of 384 (matching all-MiniLM-L6-v2) Args:

(
        self, explicit_length: Optional[int], modality: str
    )

            self.apply_repo()

    def _resolve_vector_length(
        self, explicit_length: Optional[int], modality: str
    ) -> int:
        """
        Determine the vector length to use for the generated FeatureView.

        Priority:
            1. Explicitly provided vector_length
            2. Auto-detected from embedder via get_embedding_dim("text")
            3. Default of 384 (matching all-MiniLM-L6-v2)

        Args:
            explicit_length: User-provided vector length, or None.

        Returns:
            The resolved vector length as an integer.
        """
        _DEFAULT_VECTOR_LENGTH = 384

        if explicit_length is not None:
            return explicit_length

        try:
            dim = self.embedder.get_embedding_dim(modality)
            if dim is not None:
                return dim
        except Exception:
            pass

        return _DEFAULT_VECTOR_LENGTH

    def save_to_online_store(self, df: pd.DataFrame, feature_view_name: str) -> None:
        """