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

Ollajev

Ollama-style runtime that pulls System One decision models from Hugging Face and serves them locally behind /v1/systemone.

Librarybeta
View on GitHub ↗

Overview

Ollajev is like Ollama for decision models. It pulls System One decision models from Hugging Face and serves them locally behind the same /v1/systemone API as TypeSafe's Jev, so the stock typesafe-sdk works by changing TYPESAFE_BASE_URL. You send one state and any number of typed questions and get a probability for each in one forward pass, with no generated text.

Version 0.4 makes Vega the default model, described as a trained head on a frozen Qwen3.5 in 0.8B and 4B sizes with a 73,728-token context and image input. It also supports images, audio and video with Vega, Clef-Omni, d1-omni and Clef on Apple Silicon via MLX. The CLI mirrors Ollama (pull, list, run, serve, ps, rm) with a terminal model manager and a browser playground. It installs via script, Homebrew or uv, compiling llama.cpp during install.

What the decision model decides

TypeQuestionAnswers
ChoiceAny choice question in the /v1/systemone format—
Yes / noAny yes/no questionyes · no
ScoreAny score question—

How it works

  1. Install and run ollajev to pick and download a model.
  2. ollajev serve starts the local API and playground.
  3. Point TYPESAFE_BASE_URL at it and use the TypeSafe SDK.
  4. Idle models are unloaded to make room for others.

Strengths

  • Drop-in local replacement for the hosted endpoint.
  • Multimodal inputs with supported models.
  • Familiar Ollama-style workflow.

Caveats

  • Model accuracy varies and differs from hosted Jev.
  • Install compiles llama.cpp and may pull large PyTorch builds.

Getting started

curl -fsSL https://raw.githubusercontent.com/nvkudva/ollajev/main/install.sh | sh, then ollajev serve.

At a glance

Type
Library
Status
beta
Language
Python
Runtime
local server, CLI, browser playground
Has code
yes

Use cases and integrations

How curated lists describe it

  • Run System One decision models from Hugging Face locally behind the Jev /…

    wh000wh000/awesome-jev-live

  • Run System One decision models from Hugging Face locally behind the Jev / System One API, Ollama style.

    hellogumbo/awesome-jev

  • Pulls System One decision models from Hugging Face and serves them, Ollama style, behind TypeSafe's routes and shapes, so the stock SDK works by changing `TYPESAFE_BASE_URL`.

    MrJev/awesome-jev

Frequently asked questions

Does the TypeSafe SDK work with Ollajev?

Yes, set TYPESAFE_BASE_URL to the local server; routes and shapes are the same.

Which model is default?

Vega, in 0.8B and 4B sizes, as of version 0.4.

Can it take images?

Yes, with Vega and other supported multimodal models.

Related projects

  • 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
  • jev

    TypeScript library for one-line local decisions with open models (kev-0.6b, kev-4b, open-jev), no API keys or cloud.

    ProjectbetaTypeScript
  • 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.

    ProjectbetaSwift
  • laya-system-one

    Zero-dependency npm package that runs Laya natively or as WebAssembly in Node, Bun and browsers, drop-in for Jev's /v1/systemone.

    ProjectbetaJavaScript
  • OmniStudio

    Local-first desktop AI workbench for llama.cpp, vLLM and SGLang that includes a JEV typed-decision page and /v1/systemone gateway.

    Servicebeta
  • usejev

    Runs the open LAYA decision model under Bun with native ONNX inference behind a TypeSafe-compatible API, plus an English/Persian playground.

    LibrarybetaTypeScript