microsoft/semantic-kernel · error · VectorStoreModelException

Vector field '{options.vector_property_name}' not found in t

Error message

Vector field '{options.vector_property_name}' not found in the data model definition.

What it means

Thrown by _inner_search (in_memory.py:672-674) as VectorStoreModelException when options.vector_property_name does not match any vector field in the data model definition (definition.try_get_vector_field returns None). This happens before any distance computation, during search setup.

Source

Thrown at python/semantic_kernel/connectors/in_memory.py:673

    async def collection_exists(self, **kwargs: Any) -> bool:
        return True

    @override
    async def _inner_search(
        self,
        search_type: SearchType,
        options: VectorSearchOptions,
        values: Any | None = None,
        vector: Sequence[float | int] | None = None,
        **kwargs: Any,
    ) -> KernelSearchResults[VectorSearchResult[TModel]]:
        """Inner search method."""
        if not vector:
            vector = await self._generate_vector_from_values(values, options)
        return_records: dict[TKey, float] = {}
        field = self.definition.try_get_vector_field(options.vector_property_name)
        if not field:
            raise VectorStoreModelException(
                f"Vector field '{options.vector_property_name}' not found in the data model definition."
            )
        if field.distance_function not in DISTANCE_FUNCTION_MAP:
            raise VectorSearchExecutionException(
                f"Distance function '{field.distance_function}' is not supported. "
                f"Supported functions are: {list(DISTANCE_FUNCTION_MAP.keys())}"
            )
        distance_func = DISTANCE_FUNCTION_MAP[field.distance_function]  # type: ignore[assignment]

        for key, record in self._get_filtered_records(options).items():
            if vector and field is not None:
                return_records[key] = self._calculate_vector_similarity(
                    vector,
                    record[field.storage_name or field.name],
                    distance_func,
                    invert_score=field.distance_function == DistanceFunction.COSINE_SIMILARITY,
                )
        if field.distance_function == DistanceFunction.DEFAULT:

View on GitHub (pinned to c028a0c7dc)

Solutions

Set vector_property_name to the exact name of a declared vector field in the model definition.
Ensure the model actually declares at least one vector field with a matching name.
If the model has multiple vector fields, always specify vector_property_name explicitly.

Example fix

# before
opts = VectorSearchOptions(vector_property_name='embeding')  # typo / not declared
results = await collection.search(search_type=SearchType.VECTOR, options=opts, values='q')
# after
opts = VectorSearchOptions(vector_property_name='embedding')  # matches declared field

Defensive patterns

Strategy: validation

Validate before calling

def resolve_vector_field(definition, name: str | None):
    field = definition.try_get_vector_field(name)
    if field is None:
        raise ValueError(
            f"no vector field named {name!r}; available: {definition.vector_field_names()}"
        )
    return field

Type guard

def has_vector_field(definition, name: str | None) -> bool:
    return definition.try_get_vector_field(name) is not None

Try / catch

from semantic_kernel.exceptions.vector_store_exceptions import VectorStoreModelException
try:
    results = await collection.search(search_type=SearchType.VECTOR, options=opts, values='q')
except VectorStoreModelException as e:
    # fix vector_property_name and retry
    ...

Prevention

Keep search vector_property_name in sync with the declared vector field name.
After renaming a vector field, update all search call sites.
For multi-vector models, always set vector_property_name explicitly.

When it happens

Trigger: Calling vector search with VectorSearchOptions(vector_property_name='foo') where 'foo' is not a declared vector field; omitting vector_property_name when the model has zero or multiple vector fields; a name/storage_name mismatch.

Common situations: Renaming a vector field without updating search options; mis-typing the property name; model with multiple vector fields where the default resolution is ambiguous.

Related errors

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/6dd0f8a54235e7ed. Report an issue: GitHub.