mem0ai/mem0 · error · ValueError

Vector with id {vector_id} not found in collection {self.col

Error message

Vector with id {vector_id} not found in collection {self.collection_name}

What it means

Raised by MilvusVectorStore.update() when only one of vector/payload is supplied, so the store must fetch the existing record to fill in the other, but the Milvus client returns no entity for that vector_id in the configured collection. It is a precondition failure: the update path cannot reconstruct the full record because nothing exists under that ID.

Source

Thrown at mem0/vector_stores/milvus.py:293

        Args:
            vector_id (str): ID of the vector to delete.
        """
        self.client.delete(collection_name=self.collection_name, ids=[vector_id])

    def update(self, vector_id=None, vector=None, payload=None):
        """
        Update a vector and its payload.

        Args:
            vector_id (str): ID of the vector to update.
            vector (List[float], optional): Updated vector.
            payload (Dict, optional): Updated payload.
        """
        if vector is None or payload is None:
            existing = self.client.get(collection_name=self.collection_name, ids=vector_id)
            if not existing:
                raise ValueError(f"Vector with id {vector_id} not found in collection {self.collection_name}")
            if vector is None:
                vector = existing[0].get("vectors")
                if vector is None:
                    raise ValueError(f"Existing record {vector_id} has no vector data")
            if payload is None:
                payload = existing[0].get("metadata")

        schema = {"id": vector_id, "vectors": vector, "metadata": payload}
        if self._has_bm25_schema:
            text = ""
            if payload:
                text = (payload.get("text_lemmatized") or payload.get("data", ""))[:65535]
            schema["text"] = text
        self.client.upsert(collection_name=self.collection_name, data=schema)

    def get(self, vector_id) -> Optional[OutputData]:
        """
        Retrieve a vector by ID.

View on GitHub (pinned to 001c235229)

Solutions

1. Verify the record exists first: call vector_store.get(vector_id) and handle a None result before updating
2. Check that collection_name in the Milvus config matches the collection the original insert targeted
3. If the memory was deleted upstream, drop the stale reference instead of updating it
4. For very fresh writes, retry the update after the Milvus flush/consistency interval

Example fix

// before
store.update(vector_id=memory_id, payload=new_payload)

// after
if store.get(memory_id) is None:
    raise KeyError(f"memory {memory_id} no longer exists; refresh history")
store.update(vector_id=memory_id, payload=new_payload)

existing = milvus_store.get(vector_id)
if existing is None:
    raise KeyError(f"vector {vector_id} not found; cannot partial-update")
milvus_store.update(vector_id=vector_id, vector=vec, payload=payload)

try:
    store.update(vector_id=vid, payload=p)
except ValueError as e:
    if "not found in collection" in str(e):
        # treat as deleted upstream: drop stale reference, re-add instead of update
        store.insert(vectors=[vec], payloads=[p], ids=[vid])
    else:
        raise

When it happens

Trigger: Calling mem0's Milvus vector store update(vector_id=..., vector=None or payload=None) after the memory was deleted, after a collection reset/recreation, or with an ID from a different collection/environment. Also occurs when MilvusGetResult comes back empty because the flush/consistency lag means the record is not yet visible.

Common situations: Stale memory IDs held by an application after Mem0's memory history was cleared; pointing the vector store config at a different collection name or Milvus instance than the one that wrote the data; updating a memory immediately after insertion before consistency kicks in.

Related errors

• Invalid filter key: ${JSON.stringify(key)}
• Filter value for ${JSON.stringify(key)} must be a string, nu
• The 'pymilvus' library is required. Please install it using
• Existing record {vector_id} has no vector data
• AWS Bedrock model ${this.model} returned no embedding for on

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/2b1b652e525dcfec. Report an issue: GitHub.