I was hesitant between the Prompting forum and this one, hoping I’m in the right place.
I use the API to generate lists of first names + last names which I then display on a web page via a PHP application.
Due to performance constraints (API response time), I limit the response to 20 results.
I’m using the GPT 3.5 Turbo API in JSON mode.
My issue is that I want to offer “pagination” so a “fetch more results” button to get another 20 results… and here I’m encountering a problem: how can I get the API to send me 20 DIFFERENT results?
Currently, when I initiate the “Fetch more results” action, I get new values but also sometimes one or many values that had already been presented in the previous response.
What should I do?
Should I provide the list already given in the 2nd prompt?
Is it possible to “continue” the conversation in which case the API will remember the responses it has already given?
Should/can we use a seed to indicate the responses already given?
Or is there a different way to do this?