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

noulxp

NoulXP: open standard and reference engine for packaging calibrated System One decision models so one runtime can run them all.

ProjectbetaApache-2.0
View on GitHub ↗Live demo ↗

Overview

NoulXP (formerly OpenDXP) is an open standard for packaging single-pass decision models. Today each System One model ships its own inference code: Laya needs its package, Julia 1 its own, Decider its own server. NoulXP makes a model a package that a single engine can run: it declares one of two profiles, carries weights in ONNX or GGUF, describes input construction and calibration as data, and ships a conformance file of reference answers. An engine that reproduces that file within 0.01 probability with the same decisions runs the model faithfully.

Requests travel over HTTP (noulxp serve) or the Model Context Protocol (noulxp mcp), so applications and agents can ask any compatible model the same way. Laya, Julia 1, Decider and AnyJev have been converted and checked. The spec, JSON Schemas and reference implementation are Apache-2.0, and the System One Engine powering systemonemodels.ai runs NoulXP packages. It is for model authors and platform builders.

What the decision model decides

TypeQuestionAnswers
ChoiceChoice questions answered by any NoulXP-packaged model—
ScoreScore questions—
Yes / noNoul questionstrue · false

How it works

  1. Convert a model to a NoulXP package with weights, template and calibration data.
  2. Ship a conformance file of reference answers.
  3. Run it with noulxp serve (HTTP) or noulxp mcp.
  4. Validate any engine against the conformance file.

Strengths

  • One runtime and harness for many decision models.
  • Conformance checks make portability verifiable.
  • Apache-2.0 open specification.

Caveats

  • Spec is at version 0.2; still evolving.
  • Only two profiles are defined so far.

Getting started

pip install noulxp, then run noulxp serve or noulxp mcp with a package.

At a glance

Type
Project
Status
beta
Language
Python
Runtime
CLI, HTTP server, MCP server
Has code
yes
License
Apache-2.0

Use cases and integrations

How curated lists describe it

Frequently asked questions

What problem does NoulXP solve?

Each decision model ships its own runtime; NoulXP lets one engine run any conforming package.

How is faithfulness checked?

An engine must reproduce the package's conformance answers within 0.01 probability with the same decisions.

Which models are converted?

Laya, Julia 1, Decider and AnyJev, per VALIDATION.md.

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
  • 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
  • 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
  • opendecider

    Open calibrated decision models from 400M to 80B with Jev-compatible server, MCP server and clients, benchmarked against Jev.

    ModelbetaPython
  • 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