Assistants API unable to succesfully Run


I am seeing an issue when I try and create a run on an existing thread. I receive the following error:

"code"=>"server_error", "message"=>"Sorry, something went wrong."

This is not very descriptive.

I can discount anything having changed my side as the logic is the same and I’ve not changed it. I’ve looked at the API docs and don’t see any changes there. I have also recreated this behaviour in the console manually.

Has something changed OpenAI side? This looks a bit like this thread (new-assistant-message-role-assistant-is-stopping-all-my-runs-with-state-failed/735392 - I can’t post the link as I’m new I think) from a few days ago. @nikunj could you comment perhaps?

After looking at this post ( <openai_community_url>/assistants-api-run-fails-with-server-error/628789) I changed my API key to full permissions and now it is working correctly. I think the permissions I supplied initially should have worked given the operations I was trying to run.

A suggestion for the development team OpenAI side would be an improved error message in the scenario where a user makes a call with insufficient permissions and an update to this page outlining this: <platform_url>/docs/guides/error-codes/python-library-error-types

1 Like

This is helpful feedback! we’ll look into improving our error messages here!


Hello, thanks for the quick reply. To add some more detail, the API permissions I had specified on the key that didn’t work, were:

  • Models - Read
  • Assistants - Write
  • Threads - Write

Not entirely sure why this failed to create a new Run on an existing thread, perhaps I have not added the correct permissions. Full permissions did work.