If the requests are not in English/not latin letters, the answer is returned in unicode codes

prompt=“Привіт (Hello in ukr lang)”, answer >> “text”: “\n\n\u041f\u0440\u0438\u0432”

What to do here?) And shouldn’t this be solved on the OpenAI side, since there are probably a lot of non-Latin queries?

Works on the web interface:

I suspect the problem is you need to your program to convert the output so it turns escaped unicode codes to the actual characters. I am not familiar enough to describe the exact action that needs to be taken but I think it should be rather simple.

Pretty sure you just need to add logic to decode the escaped unicode characters before printing.

1 Like

Thanks for the answer. yes, it works in the web interface… but not for the api

ThioJoe thanks again for the answer!) Yes, you can probably do as you said “pip install unidecode” for example. I found the following when entering all JSON there is an error, but when entering the response text everything is fine)
print(completion) - Unicode digits
print(completion.choices[0].message.content) - works fine