The Goal is to create an API between a program and a custom GPT, by custom I mean the free for all GPTs under the “Explore GPTs” section in the html-app or app. Though we can copy the instructions over to an assistant, the results differ in responses. Since custom GPTs have their own address they too should have an API. Since none exist I tried a script to launch the custom GPT address in a browser, search for the entry field and enter a value from the script. If you notice on a browser the search field to prompt doesn’t light up until it receives text. I was able to get this far. I Would enter a value from the script then the send arrow on the GPT would activate. I couldn’t send it.
The goal is to test apps from the developer side, without using the ASSISTANT since the results differ or might require further calibration when copying the instructions from the Custom GPT.
An API that routes to the custom GPT and returns results like the universal API does, would be fine. In short, a CustomGPT API should be created for each Custom GPT created by the user.
This would speed up workflow, testing apps. Once it is monetizing then some type of token rate can be applied. At the very least the amount of tokens should be what a regular account allows before locking up.
I build a custom GPT, I have an API, I can now work with its output in a script.
I was going to scrape the browser had I succeeded in sending that prompt from the script. Then I could test development.
Over and Out…