What is the @Generable macro used for?

The @Generable macro provides compile-time safety by allowing the model to return structured Swift types instead of raw strings, simplifying data extraction.

Can I run multiple requests on a single session simultaneously?

No, sessions are designed for one request at a time. You should check the isResponding property or create separate sessions for concurrent tasks.

How do I handle devices that don't support Apple Intelligence?

The skill includes patterns for checking model availability states, such as .deviceNotEligible or .appleIntelligenceNotEnabled, to provide graceful UI fallbacks.

Does this require an internet connection?

No, this skill focuses on Apple's on-device framework, allowing AI features to function entirely offline for better privacy and reliability.

What is the context limit for on-device FoundationModels?

Apple's on-device models currently have a 4,096 token limit, which includes the instructions, prompt, and generated output combined.

FoundationModels On-Device AI

Name: FoundationModels On-Device AI
Author: oabdelmaksoud

byoabdelmaksoud

0•

Mobile Development

Implements Apple's native FoundationModels framework to build privacy-focused, on-device AI features for iOS 26+ applications.

This skill provides comprehensive patterns and implementation guides for integrating Apple's native on-device language models into iOS applications. It empowers developers to handle model availability lifecycles, execute multi-turn conversations, and generate type-safe structured data using the @Generable macro. By focusing on on-device execution, it enables AI features that work entirely offline while maintaining strict user privacy. The skill also covers advanced topics like custom tool calling for domain-specific actions and real-time snapshot streaming for responsive SwiftUI integrations, all within the constraints of on-device hardware limits.

Key Features

01Privacy-first patterns for offline AI execution without cloud dependency

02Model availability and hardware eligibility lifecycle management

030 GitHub stars

04Custom tool calling integration for domain-specific AI tasks

05Real-time snapshot streaming for fluid SwiftUI state updates

06Type-safe structured data generation using @Generable and @Guide macros

Use Cases

01Building offline-capable text summarization or creative writing assistants

02Implementing privacy-sensitive data processing that never leaves the user's device

03Extracting structured calendar events or contact details from natural language input

Key Features

01Privacy-first patterns for offline AI execution without cloud dependency

02Model availability and hardware eligibility lifecycle management

030 GitHub stars

04Custom tool calling integration for domain-specific AI tasks

05Real-time snapshot streaming for fluid SwiftUI state updates

06Type-safe structured data generation using @Generable and @Guide macros

Use Cases

01Building offline-capable text summarization or creative writing assistants

02Implementing privacy-sensitive data processing that never leaves the user's device

03Extracting structured calendar events or contact details from natural language input