I’d be interested in setting up the highest quality models to run locally, and I don’t have the budget for a GPU with anywhere near enough VRAM, but my main server PC has a 7900x and I could afford to upgrade its RAM - is it possible, and if so how difficult, to get this stuff running on CPU? Inference speed isn’t a sticking point as long as it’s not unusably slow, but I do have access to an OpenAI subscription so there just wouldn’t be much point with lower quality models except as a toy.

The issue is that, in the function passed to reduce, you’re adding each object directly to the accumulator rather than to its intended parent. These are the problem lines:

if (index == array.length - 1) {
	accumulator[val] = value;
} else if (!accumulator.hasOwnProperty(val)) {
	accumulator[val] = {}; // update the accumulator object
}

There’s no pretty way (that I can think of at least) to do what you want using methods like reduce in vanilla JS, so I’d suggest using a for loop instead - especially if you’re new to programming. Something along these lines (not written to be actual code, just to give you an idea):

let curr = settings;
const split = url.split("/");
for (let i = 0; i < split.length: i++) {
    const val = split[i];
    if (i != split.length-1) {
        //add a check to see if curr[val] exists
        let next = {};
        curr[val] = next;
        curr = next;
    }
    //add else branch
}

It’s missing some things, but the important part is there - every time we move one level deeper in the URL, we update curr so that we keep our place instead of always adding to the top level.