Evaluating and Controlling Agents

Know the Vocabulary

This lesson uses the terms below many times. Read them once now, and come back here when a later section needs one.

• Agent: A system in which a model chooses some of its next steps, such as which tool to call. Ordinary code still decides which actions are allowed.
• Runtime: The ordinary code that hosts the model. It checks each requested tool call, runs it, and enforces every limit.
• Trajectory: The full path of one task. It holds every reasoning step, tool call, result, and recovery, in order.