I've noticed that more frequently, the OpenAI Responses API calls using the Code Interpreter tool are hanging or failing. In the past week, it seems like about 1 in 4 of these calls fails to return any response. I'm able to replicate this issue using Postman, and it's happening both in my development and production environments under the same tenant. I set the container type to 'auto', but I'm not sure what's causing these failures. Has anyone else been experiencing this problem, or does anyone have insights on what might be going wrong?
3 Answers
Have you checked what version of the Foundry Agent Service you’re using? There’s a difference between the old version using Assistants API and the new one with the Responses API. Also, have you submitted a support ticket yet? It might be worth doing that.
I've been having the same issue. When using the /responses API with auto container creation, it sometimes never returns a result. Creating containers directly works sometimes, but often fails or takes too long. We’re using the standard endpoint, and this wasn’t a problem a couple of weeks ago.
I ended up creating my own code interpreter containers and it seems to fix the "auto" issue that you’re facing.
I’m directly calling the Responses API after provisioning the OpenAI models. I usually use auto, but lately, it doesn’t create containers reliably. If I manually create containers, sometimes they work, but it takes too long. I thought it was my app, but Postman shows the same issue. I’ll definitely submit a ticket tomorrow!