Traces and observability on the AI Agents track. A trace is the full story of a run: prompts, tool calls, results, guardrail decisions, and the final answer. Without a trace you cannot tell a model mistake from a tool bug.
This lesson assumes you already worked through Scoring an agent.
The idea in practice
Log tool name, arguments, latency, and a hash of the result. Redact secrets and payment numbers before storage. Keep traces for long enough to debug a customer complaint.
A concrete check
goal = {
'track': 'AI Agents',
'lesson': 'Traces and observability',
}
checks = [
'input available at decision time',
'score matches the real decision',
'failure case written down',
]
print(goal['lesson'])
for item in checks:
print('-', item)
Run the sketch locally if you have Python. The printout is a reminder of the checks, not a trained model. Replace the strings with the real inputs from your own example before you treat it as a design.
What usually goes wrong
Logging only the final answer makes every failure look like 'the AI was wrong'. When this happens, stop adding parameters or tools. Fix the check, the data, or the permission, then run the same example again.
What to write down
- The input you are allowed to use at decision time.
- The output and the score or pass rule.
- One failure you will test on purpose.
- What you will not claim the system can do.
Practice
Sketch a trace schema with six fields. Mark which two must be redacted.
Self-check
- Say Traces and observability in one sentence that mentions an input and an output.
- Name the failure mode in this lesson and the check that would catch it.
Done when: you can explain this lesson without the page open, and you have a written failure case.