Kev
Open-source, self-hosted System One decision engine (NotJev : Kev) with a Jev-style API on top of Ollama or any OpenAI-compatible model.
Overview
NotJev : Kev (arjun988) is an open-source System One decision engine with a Jev-style wire format. You send context and choice, score or noul questions to a self-hosted server's /v1/systemone endpoint and get calibrated probabilities instead of text, for example a billing versus technical choice with confidence and a probability that a ticket should be escalated to a human.
Rather than shipping its own trained model, it lets you bring your own: Ollama, vLLM or any OpenAI-compatible endpoint, or a built-in mock to start. It is published as npm packages: a TypeScript SDK with LangChain and LlamaIndex adapters, a server with a playground, and a CLI with an MCP server. A migration guide covers moving from Jev, and a companion site has docs, benchmark write-ups and a workbench. It is independent and not affiliated with TypeSafe or OpenJev, and is a different project from jaredpalmer's Kev models.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which team should handle this? | billing · technical |
| Yes / no | Escalate to a human now? | — |
How it works
- Run kev-server with a chosen backend model or the mock
- Send state and typed questions to /v1/systemone
- Engine derives choice, score or noul probabilities from the backend model
- Your code thresholds on the answers
Strengths
- Self-hosted with any OpenAI-compatible model
- SDK adapters for LangChain and LlamaIndex
- MCP server via the CLI
- Migration guide from Jev
Caveats
- Calibration depends on the backend model you choose
- Python package not yet available
- Independent project, distinct from jaredpalmer/kev
Getting started
npm install -g @kev-ai/server @kev-ai/cli, run kev-server and open http://127.0.0.1:3000/playground/.At a glance
- Type
- Library
- Status
- beta
- Language
- TypeScript
- Runtime
- Node server, SDK, CLI and MCP server
- Has code
- yes
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
Open-source System One decision engine.
wh000wh000/awesome-jev-liveOpen-source System One decision engine. Typed choice / score / noul with calibrated probabilities. Self-host with Ollama or any OpenAI-compatible model. Apache-2.0.
hellogumbo/awesome-jevOpen System One engine for typed choice, score and noul answers with calibration.
MrJev/awesome-jev
Frequently asked questions
Does Kev include its own model?
No, you bring Ollama, vLLM or another OpenAI-compatible model, or use the mock.
Is it compatible with Jev?
It uses a Jev-style /v1/systemone format and provides a migration guide.
Is it related to jaredpalmer/kev?
No, it is a separate independent project.
Related projects
- Laya★ 32,782
Multilingual non-autoregressive decision engine answering choice, score and yes/no questions in one ~33 ms pass, with a checkpoint router.
LibrarybetaPython 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.
ServicebetaGo