Assistant API server error

API
, ,
1

Hi,
I’m trying the assistant API in my playground, I do NOT have any function calling, retrieval, or code interpreter enabled. Just a rather long prompt. Does anyone know why it always ran into server error?
It can’t even respond to Hi, seems only happens to this specific assistant. API status and other assistants are working.

User

hi

Run failed

Sorry, something went wrong.

“status”: “failed”,
“started_at”: 1701671989,
“expires_at”: null,
“cancelled_at”: null,
“failed_at”: 1701671991,
“completed_at”: null,
“last_error”: {
“code”: “server_error”,
“message”: “Sorry, something went wrong.”
},
“model”: “gpt-4-1106-preview”,

image
image769×201 4.96 KB

1 Like
2

+1, I have been facing this issue extensively too.

3

For me, it seems reducing the complexity of the prompt helps, maybe conflicting instruction can cause server error. In general, it’s still hard to debug if it’s prompt issue or OpenAI infra issue.

1 Like
4

Think you’re right. Having trouble really only with a ~600 token instruction set. Haven’t had much issue with a ~230 set (yet).

2 Likes
5

Hai,
did you find any solution for this error?
please provide me some correct answer

thanks

6

Try openairetro/examples/hello-world at main · icdev2dev/openairetro · GitHub updating with your prompt.

At least you will know if the error is in the data plane or the execution plane.

7

We are facing the same problem. Can OpenAI team at least clarify what might cause this to happen?