Issue with Persistent Thread Runs Using Assistants API with Code Interpreter Tool

API
, ,
1

Hello, I’m experiencing a recurring issue with the Assistants API, particularly when using the thread feature with the code interpreter tool. My goal is to process Excel files through the assistant and receive a modified Excel file in response based on provided instructions.

Issue Description:

Initial Request Success : Initially, when I submit a thread run with an Excel file and specific instructions, the system processes the request correctly and returns the expected Excel file.
Follow-Up Request Failures : However, when continuing the thread with follow-up requests, I frequently encounter errors. The errors typically state “internal system limitations” and "I have made another attempt, but unfortunately, the system continues to prevent the successful generation and provision of the file, etc "

  • Has anyone else encountered similar issues with thread continuity and file processing using the API?
  • Any suggestions on how to overcome these ?
3 Likes
2

Thanks for reporting this! Could you share some example thread_ids so that we can take a deeper look?

3

Thank you for your response. Thread id’s are :
Issue 1 :
thread_RJEDorDADlLzL802t8d9ON1k
And i am attaching an image as well against this Issue-1 just for reference,
Issue 2 :
thread_jBwyKukNTqOZfk5SFi0psKMb
That’s another thread which is causing issue but in a different way.
If you could look into this thread, In this my assistant isn’t behaving rightly even after my instructions , assistant reading the file in the wrong direction, though i have confirmed the format of those files as well.

Note: My main issue is Issue 1 , as i faced this issue in multiple threads . On the the first run, it does work fine, although results were not 100% accurate but that’s the purpose of thread, “Follow ups”. But after first run, most of the time my assistant failed to process requests/messages.

Please look into this. Thank you.

image
image1310×741 35.8 KB

4

Hello, i have posted another post with the same issue
and here the link of it

and thread_id for this post is
thread_1LP8Yz1AV1K9iLA8WY5QS74Q

5

I am experiencing the same issue. And I can see the problem in the Playground. The first code_interpreter run is Ok, but the follow up run fails.

Snag_3ae015
Snag_3ae0151303×715 35.8 KB

6

I am also facing the same issue with both gpt-4-turbo and gpt-4o
its unable to use code interpreter and says that :It looks like the connection was unexpectedly closed again during the process. This might be due to a temporary issue with the system. I suggest retrying the operation after a few moments or ensuring there’s a stable connection before proceeding.

If this issue persists, you might consider running the code locally or using a different environment. Let me know how you’d like to proceed or if there’s anything else I can assist you with in the meantime!

7

Any updates on this problem? This seems to be a serious stability issue with OpenAI API that keep reoccurring, making OpenAI unusable for real production projects.