// the find
rudrankriyam/Foundation-Models-Framework-Lab
A practical lab for building, testing, and evaluating apps with Apple's Foundation Models framework.
A showcase app and toolset for Apple's on-device Foundation Models framework, targeting iOS/macOS 26+ with Xcode 27. It covers the full surface area of the new API: structured generation, tool calling, RAG, voice, HealthKit, adapters, and multilingual support. Aimed at developers trying to learn or prototype with Apple Intelligence before the public SDK stabilizes.
The breadth of working examples is genuinely useful — structured generation with @Generable, tool calling with nine real system integrations (calendar, contacts, HealthKit, etc.), and a CLI with Homebrew distribution are all things you'd otherwise have to build from scratch. The AppBench evaluation suite with deterministic graders and reproducible fixtures is the right way to benchmark an on-device model and not something you see in most demo repos. CI automation that pushes to TestFlight on every main commit is a nice touch for a sample app. The agent skill files (SKILL.md + references) are a practical addition for Claude Code and Codex users who want to build on top of this without reading the source.
This requires iOS/macOS 26 and Xcode 27 beta — software that doesn't exist yet in stable form. Anyone trying to use this today is signing up for beta OS pain and potential API churn on every Xcode seed. The RAG implementation leans on LumoKit/VecturaKit, third-party packages with minimal adoption, which adds fragile dependencies to an already-unstable foundation. The repo has absorbed three previously separate repos (CLI, AppBench, Adapter Studio), and the directory structure reflects that history — it's sprawling and the boundaries between Foundation Lab, FoundationLabCore, and FoundationModelsKit require study to understand. The health AI features (predictive analytics, personalized health plans) are demo-quality at best; the @Generable models have no validation and the HealthKit queries would need significant hardening for anything production.