microsoft/semantic-kernel · error · MemoryConnectorInitializationError

Vector dimensions must be a positive number.

Error message

Vector dimensions must be a positive number.

What it means

Constructor guard in `AzureCosmosDBMemoryStore.__init__`: if `vector_dimensions <= 0` it raises `MemoryConnectorInitializationError`. Cosmos DB vector indexes require a strictly positive dimension, so the store refuses to initialize with zero or negative dimensions.

Source

Thrown at python/semantic_kernel/connectors/memory_stores/azure_cosmosdb/azure_cosmos_db_memory_store.py:66

    ef_construction = None
    ef_search = None

    def __init__(
        self,
        cosmosStore: AzureCosmosDBStoreApi,
        database_name: str,
        index_name: str,
        vector_dimensions: int,
        num_lists: int = 100,
        similarity: CosmosDBSimilarityType = CosmosDBSimilarityType.COS,
        kind: CosmosDBVectorSearchType = CosmosDBVectorSearchType.VECTOR_HNSW,
        m: int = 16,
        ef_construction: int = 64,
        ef_search: int = 40,
    ):
        """Initializes a new instance of the AzureCosmosDBMemoryStore class."""
        if vector_dimensions <= 0:
            raise MemoryConnectorInitializationError("Vector dimensions must be a positive number.")
        if database_name is None:
            raise MemoryConnectorInitializationError("Database Name cannot be empty.")
        if index_name is None:
            raise MemoryConnectorInitializationError("Index Name cannot be empty.")

        self.cosmos_store = cosmosStore
        self.index_name = index_name
        self.num_lists = num_lists
        self.similarity = similarity
        self.kind = kind
        self.m = m
        self.ef_construction = ef_construction
        self.ef_search = ef_search

    @staticmethod
    async def create(
        database_name: str,
        collection_name: str,

View on GitHub (pinned to c028a0c7dc)

Solutions

1. Pass a positive `vector_dimensions` matching your embedding model's output size (e.g. 1536).
2. Validate the dimension comes from a populated config field before constructing the store.
3. If the dimension is sourced dynamically, assert it is a positive int before use.

Example fix

// before
store = AzureCosmosDBMemoryStore(cosmos_store, "db", "idx", vector_dimensions=0)

// after
store = AzureCosmosDBMemoryStore(cosmos_store, "db", "idx", vector_dimensions=1536)

def valid_vector_dim(d) -> bool:
    return isinstance(d, int) and d > 0

# assert valid_vector_dim(vector_dimensions) before constructing the store

def is_positive_int(d) -> bool:
    return isinstance(d, int) and d > 0

from semantic_kernel.exceptions import MemoryConnectorInitializationError
try:
    store = AzureCosmosDBMemoryStore(cs, "db", "idx", vector_dimensions=dim)
except MemoryConnectorInitializationError as e:
    if "positive number" in str(e):
        raise SystemExit("vector_dimensions must be > 0") from e
    raise

When it happens

Trigger: Instantiating `AzureCosmosDBMemoryStore(...)` with `vector_dimensions=0` or a negative number, or with an uninitialized/None-derived dimension value that evaluates to <= 0.

Common situations: Passing an unset config value (e.g. `int(None)` producing 0, or a defaulted 0); a misconfigured embedding model dimension; copy-paste leaving the default unset; reading dimension from config that wasn't populated.

Related errors

• Dimensionality of {self._embedding_dim} exceeds the maximum
• Dimensionality of {dimension_num} exceeds the maximum allowe
• Database Name cannot be empty.
• Index Name cannot be empty.
• API type {cosmos_api} is not supported.

AI-assisted analysis of microsoft/semantic-kernel@c028a0c7dc (2026-08-13). Data as JSON: /api/errors/401bf9478c4402c7. Report an issue: GitHub.