does typesense support creating remote embeddings with OpenAI’s models using a

dimensions

setting lower than the default for the model? my understanding from the documentation and OpenAI’s announcement for the

text-embedding-3

models is that you can use both the large and small ones with dimensions other than the default 3072 and 1536, respectively. however, when i try setting the

num_dim

to my desired value (

1024

for

text-embedding-3-large

) in my collection, i get a failure response with an error for

Vector size mismatch

{
  "created_at": 1712762528,
  "default_sorting_field": "",
  "enable_nested_fields": false,
  "fields": [
    {
      "facet": false,
      "index": true,
      "infix": false,
      "locale": "",
      "name": "embedding_source",
      "optional": true,
      "sort": false,
      "stem": false,
      "type": "string"
    },
    {
      "embed": {
        "from": [
          "embedding_source"
        ],
        "model_config": {
          "api_key": "sk-tb**********************************************",
          "model_name": "openai/text-embedding-3-large"
        }
      },
      "facet": false,
      "hnsw_params": {
        "M": 16,
        "ef_construction": 200
      },
      "index": true,
      "infix": false,
      "locale": "",
      "name": "embedding",
      "num_dim": 1024,
      "optional": false,
      "sort": false,
      "stem": false,
      "type": "float[]",
      "vec_dist": "cosine"
    }
  ],
  "name": "xxx",
  "num_documents": 0,
  "symbols_to_index": [
    "."
  ],
  "token_separators": []
}

{
    "code": 400,
    "document": "{...}",
    "error": "Vector size mismatch.",
    "success": false
  },

the documents still get created in the collection, but their unsearchable.