Question 110
A conversation agent's prompts grow with every turn until context-window errors occur. Tool and search usage remain steady. Which cause should be prioritized?
Explanation
Correct Answer
Unbounded memory in each request consumes context tokens even when the latest question is short.
Incorrect Answers
Search capacity does not enlarge a deployed model's context window.
Credentials are not the meaningful source of the excessive historical prompt volume.
Parallelism does not reduce the tokens inside one conversation history.