Is it possible to call the embedding model via API, so that it can be easily deployed on inexpensive servers?