Inspiration

I wanted to build a simple note helper that would be fast, private and always available. I was inspired by the fullmoon chat app made by Mainframe - it showed me that running LLMs locally on iPhone is actually possible, useful, and fun.

What it does

Local AI Note is a writing helper with AI that runs on your device. Nothing leaves your phone. You can write notes, ask for help with ideas, summarize text, and edit for clarity. It works offline and keeps everything private.

How we built it

I used a local LLM with Qwen3-1.7B-4bit. The whole model runs on the device so nothing needs to connect to the internet.

I built the app with Swift and SwiftUI. Used RevenueCat for handling purchases of course. I created custom buttons and UI elements to make the interface feel smooth. The tricky part was getting the LLM to run properly on iOS and managing memory usage.

Challenges we ran into

Making it work was hard at first. I had to create a smooth onboarding flow that asks users to download the model without confusing them.

After I got it working, the app kept crashing when moved to background. I had to learn a lot about properly killing background tasks and iterate many times.

The UX was tricky too. Should it generate new text on every word you type? Every symbol? How should the help window update to feel natural? Took a lot of testing to get it right.

Accomplishments that we're proud of

I made it fast, got the flow smooth, and the concept works well. It feels natural to use.

What we learned

This is my first app with a local LLM but it won't be my last. This opens up a lot of opportunities for building things that work offline and keep data private.

What's next for Local AI Note

I want to build more local LLM apps. I'll work on optimizing the prompts and fine-tuning the model to make it more useful for writing tasks.

Built With

Share this project:

Updates