Multiple Outputs from Responses API

[jfedorowicz]

'm building an internally facing app for my media company and am using the Responses API. One of the steps of this build is to iterate of created PowerPoint slides and generate a script to go along with it. This works fine, but some of my requests generate multiple Assistant Output responses and I don’t know why. My request contains a previous_response_id and is just sending a json blob with system instructions on how to handle.

Any idea as to what would cause this? It is a simple request and it only happens about 50% of the time. Is this normal?

[aprendendo.next]

I just stumbled into one myself using o3, in a prompt that was taking unusually too long.

After manually canceling I noticed later in the logs that it was completed and there were several outputs.

The only thing I found was this recommendation:

How to prevent the API returning multiple outputs?

Hello! Thanks for surfacing this! At the moment, there isn’t a server-side fix we can apply for this behavior. The best workaround is prompt-tuning: add an instruction such as “Please return one complete answer and then stop.” Users who tighten their prompt in this way generally avoid the duplicate-message behavior. We’ll keep an eye on any future model updates and share news if that changes.

While this isn’t a definitive solution, let’s hope it can help mitigate the issue.