Skip to content

local-jev (ronitgupta138)

Rust binary marketed as an offline drop-in for Jev, claiming under 18 ms CPU decisions with an identical REST API.

Project

Overview

This local-jev is a Rust project, published on crates.io, presented as a 100% offline, zero-cost drop-in alternative to TypeSafe's Jev. It claims to run Choice, Score, Noul and Route decisions locally on CPU in under 18 ms from single-pass logits, exposing an identical REST API so agents such as Claude Code, Hermes, MCP tools or browser-use can switch by setting TYPESAFE_BASE_URL to the local server.

The README argues against cloud decision APIs on cost, privacy and latency, but its comparison figures for hosted Jev (for example $0.002 per call) differ from pricing quoted by other projects, and the excerpt does not say which model performs the inference or show accuracy results. Treat its performance claims as unverified until benchmarks and model details are documented.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich option fits the state?—
Yes / noAny yes/no question—
ChoiceWhich route should this request take? (Route primitive)—

How it works

  1. Run the local-jev binary
  2. Set TYPESAFE_BASE_URL to the local server
  3. Send Jev-format requests
  4. Receive typed answers offline

Strengths

  • Offline, no API key
  • Single native binary

Caveats

  • Model and accuracy not documented in the excerpt
  • Cost comparison with Jev appears inaccurate
  • Marketing-heavy README

Getting started

Install from crates.io (local-jev) and export TYPESAFE_BASE_URL="http://localhost:8080/v1".

At a glance

Type
Project
Language
Rust
Runtime
native HTTP server binary
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

Which model does it run?

The README excerpt does not say.

Is it really as accurate as Jev?

No accuracy results are shown, so this is unverified.

How do I switch from Jev?

Set TYPESAFE_BASE_URL to the local server's address.

Related projects

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

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

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

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

    ServicebetaPython