The content that should be matched when generating a model response. If generated tokens would match this content, the entire model response can be returned much more quickly.
The type of the predicted content you want to provide. This type is currently
always content.
Static predicted output content, such as the content of a text file that is being regenerated.