hello, the problem has existed for a long time now, as soon as you add a photo to a thread, the assistant crashes randomly. (thread.run.failed)
It would be nice to stabilize the situation once and for all. Thanks.
Yep, OpenAI is aware of ongoing API errors, including issues with image uploads, and they’re actively investigating. Hopefully, they’ll have it sorted soon! You can check their real-time updates here: (https://status.openai.com/).
Just to say, for me all is working ok now, hope this will stay like that !
upload image > ok
time to respond > ok
function call > ok
ok one hour latter error “run failed” again… and it start talking to me in chinese lol
Good to hear it was working earlier—though that’s definitely strange that it started failing again and switched to Chinese.
Try refreshing the page or starting a new chat session. Long-running sessions sometimes get unstable.
Log out and back in to force a fresh session allocation. This can reset any potential model-switching issues.
If using the API, check if you’re hitting rate limits or token caps. This could cause automatic fallback behavior.
Clear your cache or try a different browser/device. If the issue is session-related, this might fix it.
If this happens again, track when it occurs (time, session length, actions before the error). If there’s a pattern, it could indicate a throttling or load-balancing issue.
Hopefully, one of these fixes stabilizes things! Let me know if the problem persists, and we can dig deeper.
My test are with the assitant api in streaming mode: for now, i get a lot less error. No more error after upload image, and chinese appear only once. But still getting some random error after function call… For exemple my function is called correctly, but usually, just after, the model say something like “The action is done thank you” or “Your data are well saved” etc… but now i have the error “thread.run.failed”.
Good to hear that things are improving, even if there are still some lingering issues! Thanks for sharing the details—it helps narrow things down.
Since function calls are executing correctly, but the AI’s response sometimes fails, this could be related to how the model processes post-function messages.
A few things to check:
Try adding a slight delay before the assistant responds after function execution. Sometimes, immediate streaming output can fail due to timing issues.
Log API response headers and error codes—if OpenAI is returning specific failure patterns, this could help identify the cause.
Test in non-streaming mode to see if the problem happens only in streaming responses.
See if errors correlate with long session times or idle time before execution. If failures only happen after prolonged use, session expiration might be a factor.
If possible, try shorter function call outputs. Sometimes, large responses can hit unexpected system limits.
If the issue persists, OpenAI might still be stabilizing things in the background—testing at different times of day might help identify if this happens more during peak usage. Hope this helps.