mem0ai/mem0 · error · Exception

Update failed for document {vector_id}: {doc}

Error message

Update failed for document {vector_id}: {doc}

What it means

Raised by AzureAISearch.update after merge_or_upload_documents returns a non-success status for vector_id. update() rebuilds the document from vector/payload and merges it into the existing index doc; failures typically stem from schema violations (vector length ≠ index dimension, payload JSON not valid for the field), a missing pre-existing document when merge semantics require it, or throttling/auth. The guard only triggers on dict-like responses lacking status_code, so a raise means a genuine rejection captured in doc.

Source

Thrown at mem0/vector_stores/azure_ai_search.py:318

        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.
        """
        document = {"id": vector_id}
        if vector is not None:
            document["vector"] = vector
        if payload is not None:
            json_payload = json.dumps(payload)
            document["payload"] = json_payload
            for field in ["user_id", "run_id", "agent_id"]:
                document[field] = payload.get(field)
        response = self.search_client.merge_or_upload_documents(documents=[document])
        for doc in response:
            if not hasattr(doc, "status_code") and doc.get("status_code") != 200:
                raise Exception(f"Update failed for document {vector_id}: {doc}")
        return response

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

        Args:
            vector_id (str): ID of the vector to retrieve.

        Returns:
            OutputData: Retrieved vector.
        """
        try:
            result = self.search_client.get_document(key=vector_id)
        except ResourceNotFoundError:
            return None
        payload = json.loads(extract_json(result["payload"]))
        return OutputData(id=result["id"], score=None, payload=payload)

View on GitHub (pinned to 001c235229)

Solutions

  1. Inspect the doc dict status: 400 → schema/dimension mismatch, 429 → throttle, 403 → auth
  2. If vector dims changed, recreate the index with the new embedding_model_dims and re-add data
  3. Serialize payloads to plain JSON types before update; strip non-serializable fields
  4. Add backoff/retry and batch pacing for bulk updates

Example fix

# before
memory.update(memory_id, data)  # Update failed for document ...

# after: after switching embedders, rebuild the index
cfg['vector_store']['config']['collection_name'] = 'mem0_idx_v2'
cfg['vector_store']['config']['embedding_model_dims'] = 1536
memory = Memory.from_config(cfg)  # create_col builds correct schema
Defensive patterns

Strategy: retry

Validate before calling

# guard dimension drift before updates
emb_dims = 1536
if index_vector_dim != emb_dims:
    raise ConfigError('recreate index before updating with the new embedder')

Try / catch

import time
for attempt in range(5):
    try:
        memory.update(memory_id, data)
        break
    except Exception as e:
        if 'Update failed for document' in str(e) and attempt < 4:
            time.sleep(2 ** attempt)
            continue
        raise

Prevention

When it happens

Trigger: Calling memory.update(memory_id, data) after changing embedding models so the new vector dimension mismatches the index field; payload containing non-JSON-serializable objects that json.dumps renders incompatibly; concurrent updates on the same id exceeding etag/ordering constraints; throttled bulk update loops.

Common situations: Embedding-model migration without index recreation; updating memories whose payload grew beyond field limits; background jobs updating many memories in tight loops hitting QPS caps.

Related errors

AI-assisted analysis of mem0ai/mem0@001c235229 (2026-08-15). Data as JSON: /api/errors/e08282a9d5f58f94. Report an issue: GitHub.