Who it is forAn iOS or macOS developer who wants a private assistant feature that works offline.
The problemShipping an AI feature usually means a server bill and a privacy policy for user prompts.
The outcomeA Swift app that downloads a model and replies on device.
What you need first
- Rust, Python (uv), Swift or TypeScript (pnpm)
- An Apple device to benefit from unified memory
- Disk space for the downloaded model
Steps
- Add
.package(url: "https://github.com/trymirai/uzu.git", from: "0.6.13")to your Package.swift dependencies - Create the engine with EngineConfig.create() and Engine.create(config:)
- Load a model by identifier and iterate engine.download(model:) to show progress
- Start engine.chat and call session.reply with your messages
- Show message.text() in your UI
Prompt for your agent
Add trymirai/uzu to my Swift app <app> and build a chat screen that downloads the Qwen 3.5 0.8B mirai model on first launch and replies on device, following the README Swift example.
Caveat. This card is written from the trymirai/uzu README, not from a test run by Deduped. Before you start, make sure you have Rust, Python, Swift or TypeScript. Check the repo for newer instructions before you rely on these steps.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for trymirai/uzu and "Add on-device AI chat to a Swift app with Uzu"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer