We want users to be able to generate embeddings on the server and not have to bring their own. We'll need to create a new EmbeddingFunction class than can be extended for custom embedding models on the server, custom endpoints, as well as OpenAI, Cohere, Sentence Transformers, MiniLM, etc.