Return Generation Options

[parsehex]

Need to return the exact generation parameters passed to the model so that client can be accurate.

At least for LLMs and Img AI, maybe for TTS but doesn't seem necessary for STT (maybe if we ever add another client).

• Update Pydantic models
• Return correct results from clients, managers and routes
• Anything else?

[parsehex]

I don't think it's necessary to return the info object as a string, needing to parse the JSON.

[parsehex]

Didn't test thoroughly but it seems like this is in, for LLMs and Img. Do need to change Img param "info" -> "params" for consistency.