I’m building a small service that sends requests to the OpenAI API, and occasionally my client times out before it receives a response. At that point, I can’t tell whether the server never got the request or finished it and the reply was simply lost.

What’s the safest way to handle a retry here? I’m especially interested in avoiding duplicate API work and figuring out whether the original request completed.