When I upload a file (even a small one) and perform any manipulation, even a simple one, then leave the chat and do something else, keeping the browser open and chatting with another GPT, if I return to the previous GPT about an hour later (or sometimes less) and ask it to perform any file manipulation, it loses access and throws the error: AceInternalException. The problem is even worse: it loses complete access to the DataAnalyst environment and cannot even execute a ‘Hello World’ script in Python. This issue doesn’t resolve itself, and the only solution is to start a new chat. This makes it impractical to work with even a small document. I’ve had to create an account with an alternative service to OpenAI to continue working.
3 Likes
This is expected behavior. The python environment is temporary and is recycled after a period of inactivity.
1 Like
How is it expected, I mean it losses access permanently, and the conversation can no longer continue executing any kind of code, not a simple sum. I don’t understand, are you sure we are talking about the same thing?, this didn’t happen before of course. I can’t continue working because of this.
1 Like
It’s expected because it’s not a permanent environment.
That is correct. If the environment has been recycled, then it’s gone, you’ll no longer be able to access files or variables from the session.
I think you dont understand, the GPT can’t execute nothing, not talking about past code submited, I mean, it will no longer execute even a new hello word in python and stay in that state:
And will give the following errors:
AceInternalException
Encountered exception: <class ‘ace_common.ace_exception.AceInternalException’>.
GetDownloadLinkError
Encountered exception: <class ‘file_service_client.client.GetDownloadLinkError’>
I can no longer work due to this problem, It didn’t happen before. I had to open a new paid account in other company to continue.
I tend to agree this is a newish bug.
I’m fine with the temp execution environment being torn down after x.
The issue is worse than this, once the initial environment is torn down no further execution is possible in that conversation, I expect a new execution environment to be created as required after the old one was torn down.
It also seems the execution environment leaves a lock on /mnt/data prevent additional files in the conversation.
Yes, is terrible, don’t understand why this hasn’t generated enough attention. This affects mainly developers.
1 Like
You can use Open Interpreter on your own machine and it won’t do this (it does frequently crash because of bad AI output though, and lose your session that way)
you can configure an action to give the GPT access or how?
I’m having the same issue. Will OpenAI fix it?
+1. This its really annoying. Its totally understandable to have a temporal environment but when this environment closes the assistants API should have the ability to create a new one. Dont understand why this works as it.
Having the same issue here. The error is AceInternalException
Encountered exception: <class ‘ace_common.ace_exception.AceInternalException’>.
Code_interpreter tool is not working at all.
Creating new assistants/ Re-uploading files / new threads do not initiate any new environment, it keeps giving this internal error across all instances.
(like code_interpreter environment got deleted completely and not being able to reinitialize )
In case you face this issue, solution is to clean the global file storage of code_interpreter.