The Code Interpreter API generate right code, But the outputs are empty.
I run the code generated by API on my local jupyter, it runs well.
When I run the same instruction on the code interpreter WebUI, it also have not execution result.
4 Likes
I have the same question…
No solution has been found yet.
1 Like
I think we have found a solution.
In prompt, Let it use the print function to output the final result
I think it is a bug…
Someone had the same question last year in the community
Anybody have any updates on this?
I’ve spent countless hours trying different techniques, including trying to wait until the thread run completely finishes and then listing out the steps to try and parse the output, even specifically asking the model to run print statements as mentioned. The outputs are always an empty list, without fail. Absolutely nothing is returning the outputs, they are always an empty list.
Even worse, I know for a fact that on the backend the model is receiving the output, because it can answer questions using the output it gets, but I can’t actually see the output returned by the code interpreter, meaning I can’t display it to the user for them to see in my chat application.
I have no choice but to develop my own sandbox Python code interpreter and use it via function calling, which really sucks as this is so close to being complete.
1 Like