More configuration options

[rwetzold]

Would it make sense to make the max tokens configurable? Will the model return different results then? If yes, this would be good to have control over. Are there other options which could be set as well?