How does @Generable improve AI responses?

@Generable is a Swift macro that allows the model to return structured Swift types instead of raw strings, providing compile-time safety and easier data handling.

What is the token limit for Apple's on-device models?

The current hardware-constrained limit is 4,096 tokens, which includes instructions, the prompt, and the generated output combined.

Which iOS version is required for the FoundationModels framework?

These patterns and the framework are designed for iOS 26 and later versions.

Can the model perform actions within my app?

Yes, by implementing the Tool protocol, you can define custom actions that the model can invoke to perform domain-specific tasks.

Does this skill support cloud-based AI processing?

No, this skill is specifically focused on Apple's on-device FoundationModels, which prioritize privacy and offline availability by keeping data on the device.

Apple On-Device Foundation Models

Name: Apple On-Device Foundation Models
Author: hieuck

byhieuck

0•

Mobile Development

Integrates Apple's on-device FoundationModels framework to build privacy-first, offline-capable AI features for iOS applications.

The foundation-models-on-device skill provides developers with standardized patterns and implementation guides for Apple's native language model framework in iOS 26+. It simplifies the integration of Apple Intelligence by covering critical workflows including model availability checks, session management, and advanced features like structured data extraction via the @Generable macro. By following these patterns, developers can implement high-performance text generation, custom tool calling, and real-time snapshot streaming while adhering to hardware constraints like the 4,096-token context window and ensuring all data remains securely on the user's device.

主な機能

01Custom tool calling for domain-specific AI actions

02Model availability and hardware eligibility validation

03Real-time snapshot streaming for responsive SwiftUI integrations

04Context window management for the 4,096 token limit

05Structured output generation using @Generable and @Guide macros

060 GitHub stars

ユースケース

01Implementing custom AI tools to perform local app actions via natural language

02Developing privacy-preserving text summarizers that work entirely offline

03Extracting structured data from natural language for automated form filling

主な機能

01Custom tool calling for domain-specific AI actions

02Model availability and hardware eligibility validation

03Real-time snapshot streaming for responsive SwiftUI integrations

04Context window management for the 4,096 token limit

05Structured output generation using @Generable and @Guide macros

060 GitHub stars

ユースケース

01Implementing custom AI tools to perform local app actions via natural language

02Developing privacy-preserving text summarizers that work entirely offline

03Extracting structured data from natural language for automated form filling