I’m building a workflow where an OpenAI response can trigger an external action, such as creating a support ticket. If my client times out after sending the request, I may not know whether the tool call ran or whether the response was simply lost.
How do you handle retries in this situation without creating duplicate actions? I’m especially interested in practical approaches to idempotency and checking whether an action already completed.
Can you answer this question?
Write Answer0 Answers