I have a (private) GPT that performs certain actions on source code.
Wouldn’t it be useful if there was an FB “AI Agent” action that I can configure to connect through ChatGPT’s API to my GPT (or perform a generic prompt) and have it perform it’s action and apply the results to a file or similar?
Not exclusively for OpenAI, but for other LLMS that offer API access.
That way you could use FB on an even broader scale 
Maybe as a first step, some HTTP/REST action. “HTTP Get File” is quite limited …
I just found that there is that Action Studio. I totally missed that until today. I guess there we can build our own actions, even with HTTP interaction. Are there already some “REST-Style” fbap examples or shared contributions?
Hi Olaf
No plans for ChatCTP or any other AI system at this - TBH I am not even sure what that would achieve?
As for http/rest - that is something we are planning - the main issue we have is that we don’t have a good way of handling json - we’ll likely need a new dynamic object variable type. I did some R&D on this a while back - it is not a small task.
Right now our focus is on replacing the Javascript engine (VBScript will be deprecated) - that is another massive task.
We will have an 8.5 update out very soon (working on docs and testing) that automatically takes care of the 24h2 problem (plus some new features) while we work on the new javascript implementation.
Ok, I understand the Json and Javascript challenges.
That AI integration is of course not for “hey, we can do AI too”. I have a process that I run manually, currently, where at one point a GPT post-processes a file (it’s an HTML file, produced by some Figma export). That AI processing works very well, but of course I would like to automate that.
You could call something like this