Handling multiple replies in my assistant


I’ve created an assistant, now if i have instructions saying i will send multiple items to classify and i send 2 items separated in a json format
in the web version it works fine i get 2 responses in a json back

In the API however i only get 1 response, what am i doing wrong? if i try to read all the messages i just get a bunch of newlines, isn’t the output always in position 0?

it depends i think. it might send the result for both inputs in 1 response. if you use function calling, assuming your function can only handle 1 input, it will give 2 responses (parallel).

1 Like

Maybe i"m still new to this but i use prompt engineering to classify text, it’s not a function so why would that be relevant in this case?

On the web version if i prompt it to return several items in a single response it works fine so why can’t it be the same in the API? it’s the same instructions after all.

Check the thread you create through the api in the OpenAI backend. (Platform.openai.com) you might be doing something wrong or different inadvertently like overwriting the Assistant instructions in a new thread. I always find the thread history very helpful

Thanks, i checked but i don’t see an issue, i see several cases there in the thread but only 1 response same as i get in the code

However as i mentioned before if i try that in the regular web chat not the playground then it works fine…

Feel free to share more details / examples to have us take a look?