Design observability for tool choice, tool arguments, and execution outcomes.

Instruction: Describe how you would observe the full tool-calling workflow.

Context: Assesses whether the candidate can design a practical architecture and explain the main tradeoffs. Describe how you would observe the full tool-calling workflow.

Official answer available

Preview the opening of the answer, then unlock the full walkthrough.

I would instrument the workflow as three separate decisions: why the system chose a tool, what arguments it proposed, and what happened when the tool executed. Those three stages fail differently and need to be visible separately.

For tool choice, I want candidate tools, routing context, and policy...

Related Questions