👐 Create a vector search index

To retrieve documents from the knowledge base using vector search, you must configure a vector search index on the knowledge base collection.

To do this, open the Database Deployments page in the Atlas UI and select Create Index in the lower right corner under Atlas Search.

Click the Create Search Index button.

Click JSON Editor under Atlas Vector Search to create your index

Select the mongodb_agents_lab database and the knowledge collection, change the index name to vector_index, and add the following index definition in the JSON editor:

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1024,
      "similarity": "cosine"
    }
  ]
}

info

The number of dimensions in the index definition is 1024 since the Arxiv dataset we used to create the knowledge base uses Mixedbread AI's open-source mxbai-embed-large-v1 model for embeddings.