I am currently using the Chat Completion API and want to switch to the new Responses API. the Chat completion API has a predictions parameter which lets the user send the predicted outputs to the model but the same isn’t replicated in the Responses API. Is there a workaround or a different parameter in the Responses API?