GPT-4 API unable to give opinion and gives the standard template response while chatgpt gives correct response

API

Sent Prompt: Now that you’ve seen all parts, you can provide best part with rating out of 100.
Received Response: I’m sorry, but as an AI, I don’t have the ability to see or rate parts. However, I can help you find information or reviews about specific parts if you provide me with more details.


Chatgpt-4

Prompt: Now that you’ve seen all parts, you can provide best part
response: Based on the provided transcript, there are several intriguing and potentially viral moments. Given the context of social media virality, the following excerpts seem the most noteworthy due to their controversial or emotionally charged nature:

1 Like

Welcome to the forum.

We need more info to help and be able to reproduce in the Playground.

What settings (temperature, etc.) are you using? What is your system prompt for the API and any other user or account messages.

Sorry about that, Below is the link for playground OpenAI Platform

I am still playing around prompt that I give in my code and finding best prompt to get optimal output.

but I used same prompt that is in playground and results from my code and playground are very different

1 Like

No problem.

Taking a quick look, I would recommend putting some of that first USER message as the system prompt maybe.

And here it is with a temperature of 1.0…

The length of the text (how many parts) may be too much, so you might want to experiment with trying to do it for each part then have it choose between all three?

Let us know how it goes.

Yes that was the issue I just asked in prompt to give best portion for individual part and it worked like charm.

I’m assuming it’s not remembering what I have provided in the past like it remember in chatgpt chats.

1 Like

Yeah, the API is stateless, so you need to include all the content you want in each call.

Glad you got it sorted out. Hope you stick around. We’ve got a great community growing here!

1 Like