Support the new json_mode

[dasheck0]

Is your feature request related to a problem? Please describe.

I want consistent json outputs of my model especially when providing a json schema. However, most models include an explanation to their responses. Prompt engineering is not enough here.

Describe the solution you'd like

OpenAI supports now a json mode that at least guarantees that the output is json. See: https://platform.openai.com/docs/guides/text-generation/json-mode

Describe alternatives you've considered

I looked into manually calculating logits. However, seems to be too complicated for me.

Additional context

[mudler]

This is already supported in master, been introduced with #1254 . Closing