Consistency - used to control the degree of randomness or creativity in the generated text. A high consistency value will lead to more surprising and diverse responses, while a low consistency value will lead to more predictable and conservative responses. Another way to think of it is that a high temperature value allows more exploration of the model's knowledge. In contrast, a low temperature value encourages exploitation of what the model already knows.
Maximum length - controls the maximum length of the reply you can get.
