Parsed output from the judge LLM. Converted to JudgeVerdict by the judge implementation.
Confidence in the verdict (0.0 - 1.0)
Whether the goal has been accomplished
Feedback for the controller if continuing
Whether the judge has determined the goal is impossible to accomplish
Raw text response from the LLM
Reasoning behind the verdict
Parsed output from the judge LLM. Converted to JudgeVerdict by the judge implementation.