AssistantsAPI down on 11-16

API Bugs
1

There’s no word on status.openai.com - but all my runs are failing out of the gate this morning? Just getting this:

  "status": "failed",
  "started_at": 1731783181,
  "expires_at": null,
  "cancelled_at": null,
  "failed_at": 1731783181,
  "completed_at": null,
  "required_action": null,
  "last_error": {
    "code": "server_error",
    "message": "Sorry, something went wrong."
  }
8 Likes
2

Hi!
My assistants are running but will keep monitoring this topic.
Let’s hope this gets resolved for you soon!

4 Likes
3

Same here. All runs with my assistant that I use via API from my app are facing the same issue…

2 Likes
4

Have just created another assistant with same instructions and same vector store attached, replicating the assistant settings from the other.

Same issue:

  • thread can be created and run can be started
  • but when polling run status, it shows that the run obviously failed after 5 seconds
{
  "id": "run_wzDTHEsnOoPjNOD23BF4pjd1",
  "object": "thread.run",
  "created_at": 1731829540,
  "assistant_id": "asst_v8fD4u788enoI5365dGQXsde",
  "thread_id": "thread_T5BiI1B6DErXRuYb9CgTAbsw",
  "status": "failed",
  "started_at": 1731829541,
  "expires_at": null,
  "cancelled_at": null,
  "failed_at": 1731829546,
  "completed_at": null,
  "required_action": null,
  "last_error": {
    "code": "server_error",
    "message": "Sorry, something went wrong."
  },
2 Likes
5

Just an update, I was able to get it back up and running using a paired down version of my usual assistant (no vectorstore, no tools), and some re-working of my request (had o1 help me work out some improvements!).

The streaming is “chunky” which could be my implementation though I did see when I first started things up today there was a run and fail.

What was nice is I did see some SSE with an error message with a request ID last night–leads me to think they’re working on it over the weekend which I really appreciate!

NVM UPDATE: still run and fails! Guess I got some more investigating to do…

4 Likes
6

For those keeping track…

“Chunky” responses were indeed me, as I used o1 to help streamline the backend I’m using, it wasn’t sending every event, just packets of 2-3.

Got everything back up and running by duplicating my usual Assistant without tools and then gradually adding them back in to see which one was causing the instant fail.

I finally found it, and I suspect it has something to do with my implementation of “anyOf” in the schema - this tool previously worked for the past two months or so with no modification, and it passes the “save” test.

Going to see if I can rework the tool without making it so overly complicated.

2 Likes
7

It is NOT resolved for me.

My runs still fail - although the same type of messages where successfully processed before the weekend.

Failure from just now:

{
  "id": "run_VDng1Gg9iSIV2oufAPtjwI8O",
  "object": "thread.run",
  "created_at": 1731918005,
  "assistant_id": "asst_O2DO4Mt86W149U4C0o6TNMtU",
  "thread_id": "thread_Q3XcIG4HqZ3NAR5b6s9rKYTS",
  "status": "failed",
  "started_at": 1731918006,
  "expires_at": null,
  "cancelled_at": null,
  "failed_at": 1731918008,
  "completed_at": null,
  "required_action": null,
  "last_error": {
    "code": "server_error",
    "message": "Sorry, something went wrong."
  },
  "model": "gpt-4o-mini",
  "instructions": "You are orchestrating ...",
  "tools": [
    {
      "type": "file_search",
      "file_search": {
        "ranking_options": {
          "ranker": "default_2024_08_21",
          "score_threshold": 0.0
        }
      }
    }
  ],
  "tool_resources": {},
  "metadata": {},
  "temperature": 0.37,
  "top_p": 1.0,
  "max_completion_tokens": null,
  "max_prompt_tokens": null,
  "truncation_strategy": {
    "type": "auto",
    "last_messages": null
  },
  "incomplete_details": null,
  "usage": {
    "prompt_tokens": 1362,
    "completion_tokens": 23,
    "total_tokens": 1385,
    "prompt_token_details": {
      "cached_tokens": 0
    }
  },
  "response_format": {
    "type": "text"
  },
  "tool_choice": "auto",
  "parallel_tool_calls": true
}
2 Likes
8

It seems that the issue the OP is facing is different from yours. It doesn’t appear to be related to the Assistants API in general unless we start receiving many more similar reports.

To help you find the culprit and hopefully a solution, I suggest we move your case to a new topic.

1 Like
9

Un-solution this as I woke up to many more errors and runs failing intermittently.

1 Like
10

I am also seeing errors in my calls to “openai.beta.threads.runs.submit_tool_outputs_stream()”. All the run.status I get back are “failure” with a last_error = “Sorry, something went wrong.”

2 Likes
11

I have the exact same issue.

‘last_error’: {‘code’: ‘server_error’, ‘message’: ‘Sorry, something went wrong.’}

The assistant has been working fine for months, and just starting failing a few hours ago.

2 Likes
12

Same here. Yet the status tracker doesn’t seem to acknowledge any sort of issue.

13

I know, I was really hoping someone would flip the switch on that - maybe they’re all in Singapore?? :man_shrugging:

1 Like
14

A fix has been deployed.
Please keep us posted about the results on your end.

Thank you and I hope this helps!

15

Still getting very similar errors! Works about 1/3 of the time.

1 Like
16

Can you compare your issue to the one reported here?

Otherwise I’d ask you to create a separate topic for this issue.

Thanks!

17

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.