Foundation Models

Build a small agentic SwiftUI feature, profile it with the Foundation Models Instrument, and use the trace to fix prompt bloat, repeated tool calls, latency, and unsafe debugging data.

Build a practical SwiftUI image analysis flow with Foundation Models vision attachments, structured output, token budgeting, optional OCR and barcode tools, and a careful Private Cloud Compute fallback.

Build a Foundation Models router that preserves drafts, chooses local or Private Cloud Compute intentionally, and exposes quota and availability in SwiftUI.

Build a release gate for Foundation Models features with evaluations, tool trajectory checks, and layered security mitigations.

Use Core Spotlight as the retrieval layer for a Foundation Models chat feature, with item donation, delegate hydration, batched search results, query tokens, and retrieval evaluation.

Foundation Models Dynamic Profiles let one LanguageModelSession switch instructions, tools, models, and reasoning level as app state changes. That is the missing shape for agentic Swift features.

Teach an on-device Foundation Models session to call a small Swift tool, validate the arguments, and return app data without turning your SwiftUI view into an AI grab bag.

Use Foundation Models streaming APIs to update SwiftUI as an on-device response is generated, with cancellation, availability checks, and a production-shaped view model.

Foundation Models lets Swift apps ask the on-device language model for structured output. Here is a small SwiftUI search suggestion model using availability checks, @Generable, and @Guide...