myJEV
Open System One-style playground: send state plus typed choice, score and noul questions to any of four backends via a UI or API.
Overview
myJEV by open-factoryai is an open-source playground for System One-style decisions. You send any state, such as a support ticket, clinical note, sales email or JSON blob, plus a set of typed questions to POST /api/evaluate, and get back machine-readable answers with probabilities and confidence instead of free text. The README example routes a billing ticket, scores its severity on a four-level rubric and checks whether the customer is frustrated.
It supports four backends, including a parallel mode over an OpenAI-compatible model, an offline mock backend that needs no key, local models via Ollama, and integration with Mapika/decider (RLCD). A web UI with light and dark themes lets you build questions and inspect responses, and Docker profiles cover different topologies. Because some backends approximate decisions with generative calls, the example response reports 24 output tokens and 8 parallel calls. It is useful for prototyping decision schemas before choosing a model.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which team should own this ticket? | billing · technical · account |
| Score | How severe is this ticket? | Low · Medium · High · Critical |
| Yes / no | Is the customer expressing strong frustration? | yes · no |
How it works
- Choose a backend (mock, OpenAI-compatible parallel, local, or decider).
- Send state and typed questions to /api/evaluate or the UI.
- Receive choice, score and noul answers with probabilities.
- Branch in your application on confidence.
Strengths
- Offline mock backend for UI and integration work without keys.
- Same response shape across backends.
- Can run fully local with Ollama or decider.
Caveats
- LLM-backed modes emulate decisions and add latency (example: 1.8 s).
- Calibration quality depends on the chosen backend.
Getting started
Start the stack with the Docker profiles described in the README and open the playground UI, or POST to /api/evaluate.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- web app, HTTP API, Docker
- Has code
- yes
Use cases and integrations
How curated lists describe it
Turn any state into structured, calibrated decisions.
wh000wh000/awesome-jev-live
Frequently asked questions
Does it use TypeSafe Jev?
Not necessarily; it offers its own backends, including OpenAI-compatible models, local models and Mapika/decider.
Can I try it without any key?
Yes, the mock backend returns the same response shape with no key, GPU or network.
What question types are supported?
Choice, score and noul, mirroring the System One primitives.
Related projects
Docker service exposing local Laya, Von, a CLM head or hosted Jev through one TypeSafe-compatible HTTP and MCP endpoint.
MCP serverbetaPythonConverts open Jev reproductions (Laya, Kev, NanoJev, PlayJev) to INT8 ONNX for offline CPU decisions at about 15.6 ms per question.
ServicebetaPythonNative Go runtime serving Gemma 4 12B as a Jev-like decision model with hand-tuned PTX kernels and no CGo or CUDA toolkit.
ProjectbetaGoGo harness for System One models: hosted Jev, open Kev in-process via llama.cpp, and Cloudflare Clef, with typed helpers and a CLI.
ServicebetaGoGo gateway exposing the Jev /v1/systemone contract over ordinary LLM backends by reading single-token label logprobs.
ServicebetaGoRuns Intern-Decision-0.8B on Apple Silicon behind a local /v1/systemone endpoint that accepts screenshots for agent decisions.
ServicebetaPython