Can my iOS app call Jev directly from Swift?

Technically yes through JevFoundationModels, but do not ship it that way. The package warns API keys must never sit in app binaries — proxy through your own backend or run Laya on-device.

What runs fully offline?

LayaOnDevice: 322M multilingual and 421M English decision models on Core ML and the Neural Engine with zero network calls. Pick the OnDevice package trait to exclude all cloud code.

How do my @Generable types map to decisions?

Bool becomes noul, enums and strings become choice, @Guide ranges become score, with confidence and probabilities on response metadata — evaluated in 15-150 ms.

Do I have to link all three backends?

No. Swift 6.1 package traits select only what you link: OnDevice for Core ML, Remote for Jev cloud plus laya-serve HTTP, or All for everything.

How do package traits control which backends get built?

Swift 6.1 Package Traits let you enable only what you need: OnDevice activates Laya for Core ML, Remote activates Jev plus LayaServe, Clef, and OpenAI, and All activates every backend. You can also name model boundaries directly, such as Jev, Laya, LayaServe, Clef, and OpenAI.

What does the ProxyTransport solve for mobile apps?

It routes mobile cloud requests through your own authenticated reverse proxy protected by Apple App Attest and Firebase App Check, so the TypeSafe key never ships in the app binary. The docs point to a mobile security guide and tech note 0010 for the pattern.

Which targets require a key or network?

LayaOnDevice needs neither. LayaFoundationModels needs local or LAN HTTP and an optional token. ClefFoundationModels needs Workers AI or a local runner and an optional Cloudflare token. OpenAIFoundationModels and JevFoundationModels both need cloud HTTPS and a key.

CommunityJev Models & Reimplementations

System One Foundation Models

Native Swift 6 package that evaluates @Generable structs and enums through System One decision models in 15-150 ms, with on-device Laya, self-hosted laya-serve, and TypeSafe Jev cloud backends.

See highlights

Open github.com

Stars: 74
Language: Swift
License: APACHE-2.0
Published by: Community
Author: peterfriese
Last verified: Oct 2026

peterfriese/system-one-foundation-models

User repository on GitHub

View on GitHub

A lightweight, native Swift 6 bridge integrating TypeSafe AI's Jev System One decision model into Apple's Foundation Models framework.

Forks: 3
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A lightweight, native Swift 6 bridge integrating System One decision models into Apple's Foundation Models framework (LanguageModel, LanguageModelExecutor, @Generable). Evaluate strongly typed @Generable structs and enums against application state in 15–150ms with zero hallucinations, calibrated probabilities, and full Apple Intelligence API compatibility across:

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Swift

Highlights

Three backends behind Swift 6.1 package traits: LayaOnDevice on Core ML and the Neural Engine, LayaServe over HTTP, JevFoundationModels for TypeSafe cloud.
Maps Foundation Models types to primitives: Bool to noul, enums to choice, @Guide ranges to score, metadata to confidence and probabilities.
Laya runs 322M and 421M parameter models fully offline; the package warns never to embed cloud keys in app binaries. Apache-2.0 licensed.

Watch out

Needs recent Xcode with iOS/macOS 27+; on-device use means bundling the Core ML model. Copy reviewed against the repo README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

gut

Python helper that turns judgment calls into one-liners like likely, classify and rate, defaulting to Jev while also running on local or OpenAI-compatible backends.