Cannot create vector index in Azure CosmosDB for MongoDB vCore

Hi @ Peter Pölzl

Thanks for reaching out! The error you're seeing when creating a vector index is likely due to one of these issues:

Cluster Type – Vector indexes are only supported on vCore clusters, not on RU-based (Request Unit) clusters. Please make sure your free-tier deployment in France Central is using vCore.

Field Name – The field used for vector search must be named contentVector, not embedding. Please update your index creation command to use that field name.

Vector Size – The dimension 1536 is supported (must be ≤ 2000), so you're good there

Here’s the corrected version of your command:

db.runCommand({
  createIndexes: "documents",
  indexes: [
    {
      name: "vectorSearchIndex",
      key: { "contentVector": "cosmosSearch" },
      cosmosSearchOptions: {
        kind: "vector-ivf",
        numLists: 5,
        similarity: "COS",
        dimensions: 1536
      }
    }
  ]
});

Hope this helps. Do let us know if you any further queries.

I'm sure that I'm using Azure Cosmos DB for MongoDB (vCore).
I've renamed my embedding field "embedding" to "contentVector".

I ran the command you provided, but I still get the same internal error:

MongoServerError[InternalError]: [ActivityId=609ad61d-44d1-057b-e4b6-bad705000000] Index Creation failed

Hi @ Peter Pölzl

Thank you for confirming that you're using a vCore cluster and have updated the field name to contentVector.

Since you're still encountering the internal error, here are a few things worth checking:

Free Tier Limitations: While you're on a vCore cluster, the free tier may not fully support vector index creation due to resource constraints. Users have reported that upgrading to an M30 or higher SKU often resolves this.

Cluster Resources: Creating a vector index on high-dimensional vectors like yours (1536) can be memory-intensive. If the cluster doesn't have sufficient memory or CPU, the operation may fail.

Timing of Index Creation: If the collection already contains a large number of documents, try creating the index before inserting data or test on an empty collection to isolate the issue.

Index Parameters: You could try reducing numLists or switching the indexing algorithm to vector-hnsw if supported, which might yield better results on limited resources.

Please refer the below thread that might help you:

https://learn.microsoft.com/en-us/answers/questions/1643839/error-when-creating-vector-index-for-mongodb-vcore

Hi @ Peter Pölzl

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.

Hi @ Peter Pölzl

We haven’t heard from you on the last response and was just checking back to see if you have a resolution yet. In case if you have any resolution please do share that same with the community as it can be helpful to others. Otherwise, will respond with more details and we will try to help.