feder-cr/jev

jevos: an open-source alternative to Jev and Microsoft-Decision-1. Yes/no decisions, choices and scores with probabilities, on your laptop's CPU.

What it solves

Jev is designed for high-speed, local decision-making. It allows users to send a text and a set of questions (yes/no, multiple choice, or scoring) and receive the probability of each answer instantly, without relying on cloud services or expensive GPUs.

How it works

The project provides a single binary that runs on a laptop CPU. It uses an optimized model (available in OpenVINO int8 or GGUF formats) to process text and return answers. It exposes a TypeSafe Jev API via an HTTP server, allowing multiple questions to be asked about a single piece of text in one request to minimize processing time.

Who it’s for

Developers who need to integrate fast, automated classification or decision-making logic into their applications while keeping data local and avoiding per-token costs.

Highlights

  • CPU-Optimized: Runs on standard laptop CPUs with response times between 28ms and 130ms.
  • Zero Dependencies: Distributed as a single binary with no Python requirement for runtime.
  • Multi-Question Requests: Can process several different questions (e.g., sentiment, routing, and policy checks) about the same text in a single pass.
  • Local and Free: Operates entirely on the machine, eliminating cloud costs and privacy concerns.
  • Large Context: Supports a context window of up to 8,192 tokens.

Related

  • Project
    razorback16/openjev

    OpenJev is an open-source decision server that provides fast, structured, and calibrated answers (yes/no, choice, score) from diffusion-based models by reading their internal probability distributions.

  • Dispatch
    Jeeves: Improving Decision Models with Reasoning and Diffusion Drafting

    Jeeves is a 9B reasoning classifier based on Qwen3.5-9B that outperforms Jev and Kev-9B on out-of-domain tasks by integrating a reasoning chain before decision-making.

  • Project
    Rizzo-AI-Academy/rizzo-flow

    A local, open-source implementation of the Jev programming model that provides typed decisions and probabilities from an LLM without generating text tokens.

  • Dispatch
    Jev in 25 Lines of Python – How a Minimal Local Classifier Works

    The post shows that a 25‑line Python script can replicate Jev’s classification behavior by loading a GGUF LLM, prompting it with labeled choices, and converting the final token logits into calibrated probabilities.

  • Project
    PostHog/jeeves

    Jeeves is a reasoning-enhanced decision model based on Qwen3.5 that improves classification accuracy by generating a reasoning chain before outputting a final, calibrated decision.