Would you ever consider allowing VetiverAPI without VetiverModel?
In an MLOps best practices world, you should version each experiment and then choose one of the models to deploy, and VetiverModel is more easily versioned/reasoned about than the API as a whole, especially if people have custom metadata for the model they want to track.
It might be useful to add some documentation around this question somewhere.
From a vetiver user:
In an MLOps best practices world, you should version each experiment and then choose one of the models to deploy, and
VetiverModel
is more easily versioned/reasoned about than the API as a whole, especially if people have custom metadata for the model they want to track.It might be useful to add some documentation around this question somewhere.