answer

The text of the model's final answer, or null while the conversation does not end with one: before the first run, when the run paused with tool calls waiting, or when the agent used its token budget before it had an answer.

The text is incomplete when answerTruncated is true.