We should probably also update the embedding endpoint helper functions to add an optional model parameter (currently it's not one of the parameters because there's only been one model), as well as update the embedding request to support dimensions. But adding these models is a good start.

True! I only added what I needed, and the new models were enough for me, as the model can already be stated in the constructor of the EmbeddingRequest. I'm on holiday for the next week, but if you like I can have a look and enhance the helper functions and also the dimensions.
I also thought about changing the default model to text-embedding-3-small, as the ada-model will probably become deprecated in a while. But this would be a problem with backwards compatibility for many users I guess.

Alright, I made a new commit which includes

• the 2 new models as in the initial commit
• the model as optional parameter in all embedding endpoints
• enhancement of the embedding request object with dimensions and user property
• a new unit test which tests the dimensions and using one of the new models

I've already found the option to initialize a custom model, was easy to find when looking at the code base. As I don't want to use customized libraries, I did it that way in my codebase which uses your nuget package. But of course I'm happy to update soon to the new version once available. ;)

One more question regarding running the tests. I've let the embeddings tests run locally, to make sure I didn't break anything. To make it work, I manually inserted my OpenAI API key in the constructors. And of course I forgot to take it out again before committing... I'm glad OpenAI and Github both notified me right away about the issue.
Is there a better way to let the tests run locally? Or do you do it the same way and just always make sure to remove the key before publishing?