general purpose ChatCompletion

[Phodaie]

I think it would be super useful to have a general purpose high-level ChatCompletion class. The class can be instantiated with an ordered list of supported models. The create method tries the models in order and stops if a model returns a satisfactory json otherwise it tries the next model in the list. This way the user can specify models in order of cost and latency. For example, a smaller model is used first and if it fails a larger one is used.

from marvin import ChatCompletion

ChatCompletion(models=['llama2-7' , 'cloud-2' , 'gpt-3.5-turbo' , 'gpt-4']).create{
    messages = [{
       'role' : 'user',
       'content' : 'Can I have small brew, black'
   }],
   response_model = Coffee
).to_model()

[aaazzam]

Woahh, that would definitely save on boilerplate