Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

sys1rust

Runs Laya System 1 models on a Mac's GPU in Rust with custom Metal kernels, serving the same /v1/systemone API as laya serve.

ModelbetaApache-2.0
View on GitHub ↗

Overview

sys1rust runs Laya's System 1 decision models on Apple Silicon GPUs, written in Rust with its own Metal kernels and no Python. You send text and questions and get each answered with a choice, a score or a yes probability. Its server, sys1rust serve, speaks the /v1/systemone API of laya serve, so Jev and Laya clients send it the same requests. It can be installed by pasting a one-line instruction into a coding agent, which follows AGENTS.md, or with a curl install script that verifies checksums, with an optional LaunchAgent to run at login.

The first serve downloads the typed-decisions model (846 MB). Each answer carries two confidence numbers: answer_confidence (the probability of the most likely option, recommended for trust decisions) and confidence (1 minus normalised entropy), plus an action.act_probability from a second model output for acting versus escalating. The README includes speed comparisons (a macOS 26 build is about 3x faster on M5) and results, and is Apache-2.0 licensed.

What the decision model decides

TypeQuestionAnswers
Yes / noIs money involved?—
ChoiceChoice questions over supplied labels—
OtherShould the caller act on this answer or escalate? (act_probability)—

How it works

  1. Install via script or coding agent
  2. sys1rust serve downloads the model and listens locally
  3. Send /v1/systemone requests
  4. Use answer_confidence and act_probability to decide

Strengths

  • No Python; native Metal kernels
  • Same API as laya serve
  • Two clearly defined confidence measures

Caveats

  • Apple Silicon and macOS 14+ only
  • 846 MB model download

Getting started

curl -fsSL https://raw.githubusercontent.com/krishhgg/sys1rust/main/install.sh | sh && sys1rust serve

At a glance

Type
Model
Status
beta
Language
Rust
Runtime
macOS local server (Metal)
Has code
yes
License
Apache-2.0

Use cases and integrations

How curated lists describe it

Frequently asked questions

Which confidence should I threshold on?

answer_confidence, the probability of the most likely option, per the README.

Is it compatible with Jev clients?

Yes, it serves the /v1/systemone API of laya serve.

Does it need Python?

No; it is Rust with its own Metal kernels.

Related projects

  • jev-style

    Open 0.8B and 2B Qwen3.5 decision models with a local systemone-compatible server, MCP tools, agent skills and a Claude Code guard.

    ServicebetaPython
  • jev-rs

    Rust engine that answers noul, choice and score from any LLM in one prefill via logprobs, wire-compatible with /v1/systemone and MCP.

    ModelbetaRust
  • jev-mcp

    MCP server exposing Jev-style decision models served locally by Ollama's /v1/systemone as classify, yes/no and score tools.

    MCP serverbetaTypeScript
  • decidealot

    Docker service exposing local Laya, Von, a CLM head or hosted Jev through one TypeSafe-compatible HTTP and MCP endpoint.

    MCP serverbetaPython
  • edgejev

    Converts open Jev reproductions (Laya, Kev, NanoJev, PlayJev) to INT8 ONNX for offline CPU decisions at about 15.6 ms per question.

    ServicebetaPython
  • go-system-one

    Native Go runtime serving Gemma 4 12B as a Jev-like decision model with hand-tuned PTX kernels and no CGo or CUDA toolkit.

    ProjectbetaGo