Work on generating in-context ai responses in the assistant

[Mark-H]

Not quite working yet.

The basic idea is, a (hidden) prompt tells the LLM what to do and forces it to return it as JSON. The assistant then renders that a certain way and adds a button that calls the callback so it can be inserted.

Example:

MODx.AIKit.openWithContext(
     "Generate a pagetitle, introtext, and keywords for <insert relevant page content here>.",
     function(callback) {
        console.log(callback);
     }
);

The thing I'm struggling with right now is getting the LLM to return only the JSON and not, like, JSON wrapped in a code block. Might need to look into structured outputs or find another way to distinguish from usual text and a callback.