Tool errors and retries on the AI Agents track. Tools fail. Timeouts, 404s, and validation errors are different. Retry timeouts. Do not retry a 404 as if the server blinked. Tell the model the status code and a short body.
This lesson assumes you already worked through Traces and observability.
The idea in practice
Use idempotency keys on anything that writes. One retry must not create two refunds. Cap retries at two or three.
A concrete check
goal = {
'track': 'AI Agents',
'lesson': 'Tool errors and retries',
}
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
Blind retries on a payment call double-charge the customer. 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
Classify five fake errors into retry, revise arguments, or stop and ask a person.
Self-check
- Say Tool errors and retries 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.