What are the limitations of on-device models compared to cloud models?

On-device models typically have a tighter context window (around 4,096 tokens) and require explicit availability checks to ensure the device is eligible and the model is ready.

What iOS version is required for these FoundationModels?

These patterns and implementation guides are specifically designed for iOS 26 and later, leveraging the latest Apple Intelligence framework capabilities.

How does the @Generable macro improve AI integration?

The @Generable macro provides compile-time safety and structured output, allowing you to generate native Swift types directly from the model instead of parsing raw strings.

Does this skill require a cloud backend or internet connection?

No, the primary focus of this skill is implementing on-device models that process data locally, ensuring user privacy and allowing features to work offline.

iOS On-Device Foundation Models

Name: iOS On-Device Foundation Models
Author: unju-ai

byunju-ai

0•

Mobile Development

Integrates Apple's FoundationModels framework into iOS applications for private, offline, and structured on-device language model interactions.

This skill provides specialized patterns and best practices for leveraging Apple Intelligence on-device using the FoundationModels framework in iOS 26+. It enables developers to implement text generation, structured data extraction via the @Generable macro, and custom tool calling—all while maintaining user privacy by processing data entirely on-device. The skill covers essential workflows including model availability checks, multi-turn conversation sessions, and real-time snapshot streaming for building responsive, AI-powered SwiftUI interfaces without cloud dependencies.

主要功能

01Custom tool calling for domain-specific AI logic and local actions

02Type-safe structured output using @Generable and @Guide macros

030 GitHub stars

04Real-time snapshot streaming for progressive UI updates in SwiftUI

05On-device text generation and summarization via Apple Intelligence

06Privacy-first implementation guidelines for offline AI capabilities

使用场景

01Extracting structured data like calendar events or profiles from natural language input

02Implementing real-time, streaming AI interfaces within iOS apps for better UX

03Building privacy-preserving AI assistants that function without an internet connection

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add unju-ai/ecc foundation-models-on-device

For use in Claude.ai and ChatGPT

主要功能

01Custom tool calling for domain-specific AI logic and local actions

02Type-safe structured output using @Generable and @Guide macros

030 GitHub stars

04Real-time snapshot streaming for progressive UI updates in SwiftUI

05On-device text generation and summarization via Apple Intelligence

06Privacy-first implementation guidelines for offline AI capabilities

使用场景

01Extracting structured data like calendar events or profiles from natural language input

02Implementing real-time, streaming AI interfaces within iOS apps for better UX

03Building privacy-preserving AI assistants that function without an internet connection

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add unju-ai/ecc foundation-models-on-device

For use in Claude.ai and ChatGPT