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

laya-rust

Pure-Rust Laya inference on candle for CPU, Metal or CUDA, with a CLI and a local web console showing close calls.

Modelbeta
View on GitHub ↗

Overview

laya-rust provides Rust inference for Laya on candle with no Python, torch or ONNX export step, running on CPU, Metal or CUDA. You download the Apache-2.0 Laya checkpoint (about 847 MB, ungated) or its multilingual or typed-decisions variants, which share the same layout; the README recommends the fine-tuned typed-decisions variant as the default for typed decisions.

The CLI takes a state file and a questions JSON file and prints Jev-shaped output: a choice with probabilities and confidence, a score with legend, distribution and confidence, and a noul probability, plus Laya's act-probability field and token usage. A feature-gated web console serves a single page with input on the left and the decision on the right, showing each answer's verdict, probability, confidence and the runner-up so close calls are visible. It is aimed at developers who want a small, dependency-light Rust binary for Laya.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich department should handle this ticket?billing · technical · sales
ScoreHow urgent is this ticket?not urgent · low · normal · high · critical outage
Yes / noDoes this need a human?—

How it works

  1. Download Laya weights from Hugging Face
  2. Build with cargo
  3. Run the CLI with a state file and questions JSON
  4. Or serve the local web console

Strengths

  • No Python or ONNX step
  • CPU, Metal and CUDA support
  • Console highlights close calls via the runner-up

Caveats

  • Manual weight download
  • Personal project without published benchmarks in the excerpt

Getting started

Download the weights, cargo build --release, then ./target/release/laya --state-file examples/ticket.txt --questions examples/questions.json.

At a glance

Type
Model
Status
beta
Language
Rust
Runtime
candle CLI and local web console
Has code
yes

Use cases and integrations

How curated lists describe it

  • Pure-Rust inference for Laya on candle: ModernBERT-large with its RL decision head, no Python at serving time.

    MrJev/awesome-jev

Frequently asked questions

Which checkpoint should I use?

The README suggests the fine-tuned typed-decisions variant.

Does it need Python?

No, it is pure Rust on candle.

Which hardware does it support?

CPU, Apple Metal and CUDA.

Related projects

  • 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
  • gojev

    Go harness for System One models: hosted Jev, open Kev in-process via llama.cpp, and Cloudflare Clef, with typed helpers and a CLI.

    ServicebetaGo
  • hearim

    Go gateway exposing the Jev /v1/systemone contract over ordinary LLM backends by reading single-token label logprobs.

    ServicebetaGo
  • intern-decision-mlx

    Runs Intern-Decision-0.8B on Apple Silicon behind a local /v1/systemone endpoint that accepts screenshots for agent decisions.

    ServicebetaPython