I don’t know if this qualifies as API or not, but I can’t come up with anywhere better to put a Playground UI question.
I work from a tablet. I can use an API testing tool to submit an API call and get results, but it’s incredibly inconvenient. What I have typically done is use the Playground UI to submit story ideas and either copy the results I like out to my world-building wiki or more often vary my actual prompt and save various results as ‘prompts’ and later copy a huge bundle of them en masse to my wiki.
The GPT4 interface in the playground (and API) uses this tripartite system/ assistant/ user format, and in the playground there is no easy for way to export the results. You can’t save a ‘prompt’, either actual prompts to play with variations on a theme, or results. If a prompt uses any guiding towards the result you’re looking for I find I can only copy/paste the most recent output. The results tend to be interspersed with each other so you have to edit the results.
It is a singular pain in the arse and I can’t imagine this is the intended result. Is there any intent to move GPT4 playground into something more consistent with the DaVinci playground UI?
(As a side issue, if I could find more info on the messages [ {“role” … } ] function and what is the distinction between each role, that would also be grand.)