[ML] Add support for dimensions in google vertex ai request

[jonathan-buttner]

This PR adds support for passing the dimensions field through to the request to send to google vertex ai. We had included the field in the service settings but didn't send it in the request.

The dimensions field maps to outputDimensionality https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/text-embeddings-api

Testing

PUT _inference/text_embedding/test { "service": "googlevertexai", "service_settings": { "service_account_json": "<service_account>", "model_id": "gemini-embedding-001", "location": "us-central1", "project_id": "elastic-ml", "dimensions": 10 } } 

Generate embeddings

POST _inference/text_embedding/test { "input": ["The food was delicious"] } 

Response

{ "text_embedding": [ { "embedding": [ 0.0038749145, -0.0017926422, 0.00781796, -0.06123169, -0.0026226782, -0.012737362, 0.016699702, -0.016461374, -0.007187545, 0.016607152 ] } ] } 

[elasticsearchmachine]

Hi @jonathan-buttner, I've created a changelog YAML for you.

[jonathan-buttner]

This change will allow an empty object for parameters to be sent. I verified that sending

"parameters": {} 

succeeds.

[elasticsearchmachine]

Pinging @elastic/ml-core (Team:ML)

[davidkyle]

LGTM