jev-foundation-models
Swift 6 bridge that runs System One decision models behind Apple's Foundation Models API, on device with Laya or via hosted endpoints.
Overview
System One for Apple Foundation Models is a Swift 6 bridge that plugs decision models into Apple's Foundation Models framework (LanguageModel, LanguageModelExecutor and @Generable), so apps can evaluate typed @Generable structs and enums against application state. It supports several backends: on-device Core ML versions of Laya (322M multilingual and 421M English models) on the Neural Engine and GPU with no network calls; self-hosted laya-serve speaking the Jev-compatible /v1/systemone protocol; Cloudflare's open Clef and Clef-Flash multimodal models; OpenAI's Decisions API; and hosted TypeSafe Jev with retry, cancellation and confidence routing policies.
The README includes a security advisory never to embed cloud API keys in mobile app binaries, recommending on-device models or a server-side proxy. A list summary notes a protocol-based model interface and six tests. Latency claims in the README are the author's.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which case of this @Generable enum fits the app state?(inferred) | — |
| Yes / no | Is this condition true of the app state?(inferred) | — |
How it works
- Define @Generable types for the decision
- Choose a backend: on-device Laya, laya-serve, Clef, OpenAI Decisions or Jev
- Evaluate against app state through the Foundation Models API
- Route on confidence with RoutingPolicy
Strengths
- Fully on-device option with no network
- One API across several decision backends
- Clear guidance on key safety in mobile apps
Caveats
- Cloud backends need a server-side key proxy for apps
- Small test suite
At a glance
- Type
- Project
- Status
- beta
- Language
- Swift
- Runtime
- Apple platforms (Foundation Models, Core ML)
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Apple platforms: a Swift 6 bridge that runs Jev decisions through Apple's Foundation Models on device, with a protocol-based model interface and six tests.
yibie/awesome-jev
Frequently asked questions
Can it run without a network?
Yes, with the on-device Core ML Laya models.
Which hosted backends are supported?
TypeSafe Jev, OpenAI Decisions, Cloudflare Clef and self-hosted laya-serve.
Can I put my TypeSafe key in the app?
No; the README warns never to embed cloud keys in mobile binaries.
Related projects
Go harness for System One models: hosted Jev, open Kev in-process via llama.cpp, and Cloudflare Clef, with typed helpers and a CLI.
ServicebetaGoTypeScript library for one-line local decisions with open models (kev-0.6b, kev-4b, open-jev), no API keys or cloud.
ProjectbetaTypeScriptZero-dependency npm package that runs Laya natively or as WebAssembly in Node, Bun and browsers, drop-in for Jev's /v1/systemone.
ProjectbetaJavaScriptLocal-first desktop AI workbench for llama.cpp, vLLM and SGLang that includes a JEV typed-decision page and /v1/systemone gateway.
ServicebetaRuns the open LAYA decision model under Bun with native ONNX inference behind a TypeSafe-compatible API, plus an English/Persian playground.
LibrarybetaTypeScript- ollaya★ 1,305
Rust daemon and CLI that pulls open decision models like Laya and serves them locally behind a TypeSafe-compatible /v1/systemone.
ServicebetaRust