[Critical] Over 25% Assistant API Request Timeout Randomly

API
72

Issues are sitll persisting.
run_EtdthPa3uzSsDpbH5Jsh6ckM/steps
{
“object”: “list”,
“data”: [
{
“id”: “step_NHVXyXqM5JkzMOERasCP88SO”,
“object”: “thread.run.step”,
“created_at”: 1709528338,
“run_id”: “run_EtdthPa3uzSsDpbH5Jsh6ckM”,
“assistant_id”: “asst_gVGXCMmQwWVb3R6b1JbttS23”,
“thread_id”: “thread_XWeDox7vrthrPsmxIYNcJULJ”,
“type”: “message_creation”,
“status”: “cancelled”,
“cancelled_at”: 1709528354,
“completed_at”: null,
“expires_at”: null,
“failed_at”: null,
“last_error”: null,
“step_details”: {
“type”: “message_creation”,
“message_creation”: {
“message_id”: “msg_0OhBQlnXlQekPx6PZinMjEZa”
}
},
“usage”: {
“prompt_tokens”: 4245,
“completion_tokens”: 1337,
“total_tokens”: 5582
}
},
{
“id”: “step_GobvsGBCMTj68A8qnGdRIFXe”,
“object”: “thread.run.step”,
“created_at”: 1709528305,
“run_id”: “run_EtdthPa3uzSsDpbH5Jsh6ckM”,
“assistant_id”: “asst_gVGXCMmQwWVb3R6b1JbttS23”,
“thread_id”: “thread_XWeDox7vrthrPsmxIYNcJULJ”,
“type”: “message_creation”,
“status”: “completed”,
“cancelled_at”: null,
“completed_at”: 1709528338,
“expires_at”: null,
“failed_at”: null,
“last_error”: null,
“step_details”: {
“type”: “message_creation”,
“message_creation”: {
“message_id”: “msg_79d5EJj4YmbcGaHBXdtTIirK”
}
},
“usage”: {
“prompt_tokens”: 2868,
“completion_tokens”: 1375,
“total_tokens”: 4243
}
}
],
“first_id”: “step_NHVXyXqM5JkzMOERasCP88SO”,
“last_id”: “step_GobvsGBCMTj68A8qnGdRIFXe”,
“has_more”: false
}
had to stop the run because it seemed to be caught in a loop

73

Kind of same issue mine get stucks at “in_progress”

74

I’ve noticed that the error rate is decreasing, yet the issue is still there. It seems we may need to implement our own error handling in the future along with Assistant API.
Run id: run_WetCN6hsmZW9B8lv0MAc8GgF

75

The problems still continue. Please resolve it as soon as possible.

run_NtsH18bIX10zpqczkamoDSJr

76

run_IPwiwOPTiy7AlsE6ETqFWHnB

77

The issue is back for most of our calls across all assistants.

run.id: run_P7dXhKdIHZr4ytOezF8fm6it

78

Seeing the same thing. Assistants that were working yesterday now aren’t. And when I try to inspect the thread on the platform, I get an endless spinner. Clearly something on OAI’s end, but, as always, nothing on the status page.

79

And we’re back… again, no status update. Assistant API is the red headed step child.

80

Apologies for the trouble – we’re looking into this.

81

We’ve resolved most of these issues — once again, apologies for the choppiness over the past couple of days. We take these incidents seriously and will be doing a detailed post-mortem to come up with robust remediations.

Thanks for building on the Assistants API. Please reach out if you face any further issues.

82

Please take a look: run_K9uIHhMXhPHvqy3M9uE6HhA3

The problems still continue.

83

@advancedfutureai this run looks fine on our end and it looks like it completed successfully.

84

runId: run_ujPe6hn7dq4H6WJXSGz6Z6SB

Maybe this will also get completed but it takes 12 minutes to do that which was 1 min earlier

85

2024-03-05 12:40:50,920 - DEBUG - Starting new HTTPS connection (1): api.openai.com:443
2024-03-05 12:40:51,225 - DEBUG - https://api.openai.com:443 “GET /v1/threads/thread_QBvnj3liUfNk5mHbjajV55Dg/runs/run_wHvYG1OhRQew3LqUZFG17m6y HTTP/1.1” 200 None
2024-03-05 12:40:51,229 - DEBUG - Starting new HTTPS connection (1): api.openai.com:443

the error still persists several requests without response to the assistant api.

thread_QBvnj3liUfNk5mHbjajV55Dg

86

Same issue here. OpenAI assistants have been stalling since I began working with them in January 2024. I thought it was because my apps are consuming a lot of tokens for each query, but now I realize it’s a common issue.

87

image
image1020×120 8.85 KB

In fact, following the logs, it freezes and comes back after a while varying between 5 to 12 minutes, by then the application has already timed out to return the response.

88

image
image549×886 34.6 KB

The behavior of the Playground also seems abnormal to me.

for a simple question, it seems to be repeatedly consuming the Tokens

89

Another run that times out. I’ve already increased our server timeouts multiple times to 5 minutes now. We have a polling system in the client to continously poll the run so this means the individual runs request is taking over 5 minutes to return an api response from the endpoint.

“run_uP0rpICu7UTRAD5J9jOBNvP9”

90

I am still seeing this issue. I don’t think it has been resolved.

91

I’m sorry that you’re having this issue, I have been having the same problem. I cannot finish my project with this insane inconsistency. We ALL HAVE TO ACCEPT, that this system cannot handle the load. It simply cannot. It simply cannot. Maybe someday I hope. I have friends who have Enterprise accounts and they DO NOT have these issues at all. I don’t like making assumptions, but it is likely that OpenAI is prioritizing time to the Enterprise people who pay BIG money? Oh well … my dream is trashed.