You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've noticed an interesting regression with Code Interpreter on Azure OpenAI in the past few weeks.
Regardless of model, we used to be able to send a prompt with a message like draw me a 5 phase sine wave starting with each phase voltage by n * pi and have it generate the image 100% of the time.
A couple of weeks ago we start getting messages around failing in the plotting environment. These messages have different wording but the same meaning.
eg
It seems there was an issue with the plotting environment.
The default plotting backend ran into an environment issue, so I’ll switch to a non-interactive backend and regenerate the figure cleanly
This happens across both gpt-4o and gpt-5.x models
Sample request
{
"metadata": {
"frequency_penalty": "0"
},
"temperature": 0.5,
"top_p": 0.95,
"model": "gpt-4o",
"max_output_tokens": 16384,
"tools": [
{
"type": "code_interpreter",
"container": {
"type": "auto",
"file_ids": []
}
}
],
"input": [
{
"type": "message",
"role": "system",
"content": [
{
"type": "input_text",
"text": "You are an AI assistant designed to provide accurate, comprehensive, and useful responses. Your primary goal is to deliver value through assistance.\n\n\r\n\r\n## INSTRUCTIONS:\r\n- You have access to a Python execution environment (Code Interpreter).\nUse Python to perform mathematical calculations, manipulate data programmatically, or generate new files.\n\nDo NOT use Python for reading or describing content that is already visible in images/PDFs or simple extraction of values you can see visually\n\nExplain your plan briefly before running code. Keep outputs concise.\r\n"
}
]
},
{
"type": "message",
"role": "user",
"content": [
{
"type": "input_text",
"text": "draw me a 5 phase sine wave starting with each phase voltage by n * pi"
}
]
}
],
"stream": true
}
Sample response fragments
EDB
event: response.output_text.done
data: {"type":"response.output_text.done","content_index":0,"item_id":"msg_0f830460ec85e9fe0069fbb8b51da081968c213807300f7bc4","logprobs":[],"output_index":2,"sequence_number":374,"text":"It seems there was an issue with the plotting backend. Let me address this and generate the sine wave visualization again."}
event: response.output_item.done
data: {"type":"response.output_item.done","item":{"id":"msg_077f71b4408544c80069fbba79c0e081959805ce6d30c87488","type":"message","status":"completed","content":[{"type":"output_text","annotations":[],"logprobs":[],"text":"The default plotting backend ran into an environment issue, so I’ll switch to a non-interactive backend and regenerate the figure cleanly."}],"phase":"commentary","role":"assistant"},"output_index":2,"sequence_number":259}
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I've noticed an interesting regression with Code Interpreter on Azure OpenAI in the past few weeks.
Regardless of model, we used to be able to send a prompt with a message like
draw me a 5 phase sine wave starting with each phase voltage by n * piand have it generate the image 100% of the time.A couple of weeks ago we start getting messages around failing in the plotting environment. These messages have different wording but the same meaning.
eg
It seems there was an issue with the plotting environment.The default plotting backend ran into an environment issue, so I’ll switch to a non-interactive backend and regenerate the figure cleanlyThis happens across both gpt-4o and gpt-5.x models
Sample request
Sample response fragments
Beta Was this translation helpful? Give feedback.
All reactions