mem0ai/mem0 · error · ValueError

Update vector has dimension {len(vector)}, but the index '{s

Error message

Update vector has dimension {len(vector)}, but the index '{self.collection_name}' expects dimension {self.embedding_model_dims}. Ensure your embedding model's output dimensions match the vector store configuration.

What it means

update() checks a supplied vector's length against the index's embedding_model_dims before modifying the document. The OpenSearch knn_vector field is fixed-dimension, so an update with a different-size vector would be rejected by the server after the lookup; this moves the failure to the caller with a clear message.

Source

Thrown at mem0/vector_stores/opensearch.py:325

        response = self.client.search(index=self.collection_name, body=search_query)
        hits = response.get("hits", {}).get("hits", [])

        if not hits:
            return

        opensearch_id = hits[0]["_id"]

        # Delete using the actual document ID
        self.client.delete(index=self.collection_name, id=opensearch_id)

    def update(self, vector_id: str, vector: Optional[List[float]] = None, payload: Optional[Dict] = None) -> None:
        """Update a vector and its payload using the custom 'id' field."""
        if vector is not None:
            if len(vector) == 0:
                raise ValueError("Cannot update with an empty vector.")
            if len(vector) != self.embedding_model_dims:
                raise ValueError(
                    f"Update vector has dimension {len(vector)}, "
                    f"but the index '{self.collection_name}' expects dimension {self.embedding_model_dims}. "
                    f"Ensure your embedding model's output dimensions match the vector store configuration."
                )

        # First, find the document by custom ID
        search_query = {"query": {"term": {"id": vector_id}}}

        response = self.client.search(index=self.collection_name, body=search_query)
        hits = response.get("hits", {}).get("hits", [])

        if not hits:
            return

        opensearch_id = hits[0]["_id"]  # The actual document ID in OpenSearch

        # Prepare updated fields
        doc = {}

View on GitHub (pinned to 001c235229)

Solutions

Make the update vector come from the same embedding model (and dims) as the index
Fix the dimension configs and recreate the index if the model genuinely changed
Pre-check len(vector) == embedding_model_dims before calling update (see guard below)

Example fix

# before
store.update(vector_id=vid, vector=small_model_embed(text), payload=p)  # 768 vs 1536 index

# after
store.update(vector_id=vid, vector=embed(text), payload=p)  # same model/dims as index

Defensive patterns

Strategy: type-guard

Validate before calling

if vector is not None and len(vector) != store.embedding_model_dims:
    raise ValueError(f"update vector is {len(vector)}-dim; index expects {store.embedding_model_dims}")
store.update(vector_id=vector_id, vector=vector, payload=payload)

Type guard

def update_dims_ok(vector, dims: int) -> bool:
    return vector is None or len(vector) == dims

Try / catch

try:
    store.update(vector_id=vid, vector=vec, payload=p)
except ValueError as e:
    if "expects dimension" in str(e):
        vec = embed(text)  # re-embed with the index-matching model
        store.update(vector_id=vid, vector=vec, payload=p)
    else:
        raise

Prevention

Always re-embed update text with the same model that built the index
Single source of truth for dimension config across embedding and vector store
Rebuild the index when the embedding model changes; do not mix old and new vectors

When it happens

Trigger: update(vector_id=X, vector=<768-dim>) on an index created with embedding_model_dims=1536, or any embedding model/config change made after index creation without re-indexing.

Common situations: Same drift causes as insert: provider switch, dims typo in config, multi-tenant setups where one index is shared across models with different dimensions.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/c5a698cb515ecfa0. Report an issue: GitHub.