Apple Intelligence - Foundation Model integrations

I made a quick introduction to Foundation Model (Apple Intelligence) integration in Drafts v48.1. This is still in its early phases, and I’m interested in feedback from those that experiment with it – I expect it to get a lot better and more capable in the coming months.

It can already do quite a bit, allowing querying, creating, modifying drafts, running actions and more. That said, it’s a bit finicky and somewhat limited by the limitations of the on-device models, so it will take some time to sort out what ways to utilize this functionality.

Example Actions

Docs

4 Likes

Nice progress. I’m looking forward to more in Drafts and Apple’s future local models.

For inspiration, you might want to take a look at what Mindnode Next have done using the local models.

Looking forward to playing with this at some point, but I have to admit that I’ve been holding off full adoption of iPadOS26, so haven’t been able to take advantage of the Apple Intelligence advances. It’s the least bullish I’ve ever been about adopting an i(Pad)OS update, far as I can remember. I’m sure I’ll get there eventually…

This is great and I expect tremendous capabilities from this model integrated to Drafts. I dream of being able to dialog in natural language with all my drafts. I integrated the class into my system and began to experiment with it. Nice tool, very promising. I was often stuck mostly because of exceeding tokens or bugs, but I know this is an experimental class. I’m looking forward to trying what’s coming.

1 Like

I am playing around with these things a lot lately.

If I use the Draft tools and eg.g ask something like „please find all active project drafts“ I get this error:

The number of tokens varies a bit but it is way too much anyways.

I think this is kind of expected if the Drafts Library is bigger and I’m not sure if you can do something about this.

Basically I’m unable to use it a lot besides doing stuff with the current draft right?

Yes. This is mostly useful for things with a limited scope. Definitely best for things that analyze one draft for things like summarization or tagging.

I expect that will get better over time and Apple will increase token limits, but that’s what it is for now.

1 Like