Is this just a hosted proxy in front of TypeSafe Jev?

No: it is a Jev-class reimplementation on an RWKV checkpoint with its own decision engine, speaking the same request contract so existing Jev adapters work unchanged.

What hardware and toolchain do I actually need to serve it?

A CUDA GPU with enough memory for the 1B checkpoint in fp16, plus a compiler toolchain for JIT kernels; a slow CPU-only smoke path exists but is not the intended serving route.

Why does the first start take so long or fail immediately?

Download the checkpoint file into the models directory first; weights are not vendored, and first start compiles CUDA kernels once before caching them.

Which backend and decoding mode should I choose in serve?

Use the batched backend for GPU serving with trie or single-fork decoding and optional request merging; the reference backend is for smoke tests rather than throughput.

How did the author measure accuracy, and what did it score?

Using JevBench's unchanged typesafe adapter against the local server on the public item sets: overall 0.541 across 231 decisions, with easy at 0.708, standard at 0.683, the adequacy judge set at 0.500, and hard at 0.396. It is a self-measurement on public items, and the author notes 109 held-out hard items were never touched.

What would a hosted run cost at current input prices?

About $0.006 to $0.019 per thousand decisions at $0.01 to $0.03 per million input tokens, because answers are read off a constrained softmax and output cost is zero. The estimate uses a mean of roughly 643 input tokens per decision. It is a hypothetical hosted price, not what self-hosting costs you in GPU time.

What are the licensing terms for the code and weights?

MIT for the code in this repository, with vendored parts keeping their own terms: the Albatross backend is Apache-2.0, and the RWKV weights follow the license on their model card. So the serving code is permissive, but redistribution of the checkpoint is governed separately.

CommunityJev Tools & CLIs

Rwkv Jev

RWKV-7 decision-model server exposing a Jev-compatible /v1/systemone endpoint, with constrained decoding over answer tokens and zero output tokens.

See highlights

Open github.com

Stars: 4
Language: Python
License: MIT
Published by: Community
Author: XingQiPan
Last verified: Oct 2026

XingQiPan/rwkv-jev

User repository on GitHub

View on GitHub

基于 RWKV-7 (g1k-1b) 的 Jev 式零生成决策引擎：受限 softmax 直读概率分布，支持 noul/choice/score 三原语，TypeSafe /v1/systemone 接口兼容，JevBench v1.2 公开集实测 acc 0.541 / p50 87ms

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A Jev-class decision model: you POST a piece of state and a bounded rubric to /v1/systemone, it returns a typed answer with a full probability distribution over the exact label set. No prose is generated — answers are read off a constrained softmax over the answer token(s), so output cost is zero tokens. Model: rwkv-g1k-1b-temp-6003 — an RWKV-7 G1k 1B checkpoint (.pth), from BlinkDL's temp training models (link).

Quoted from the project's README. Read the full README.

Tagscommunity github

Works forJev for Scoring

Used inJev for Software & DevTools

Languages: Python

Highlights

Implements the /v1/systemone contract with typed answers read off a constrained softmax over answer tokens.
Runs on a batched GPU backend in fp16 with one forward per answer position, plus a CPU reference backend.
Weights load from a local checkpoint file, with trie multi-token paths or single-fork label decoding.
Includes a JevBench self-measurement setup with schema validity and latency reporting on public items.

Watch out

Accuracy figures are the author's self-measurement on public items with a local GPU, not an independent benchmark of Jev itself. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

brier

Open local decision model on Qwen3-4B with LoRA adapters that answers choice, score, and noul questions with probabilities, calibrated confidence, and conformal sets, offline with no per-call cost.

Jev Tools & CLIs#community#github

Community

Mangaba-ai/brierPython

paper-radar-jev

Python CLI that pulls recent arXiv papers, scores relevance with TypeSafe and writes daily Markdown and JSON rankings with resume-safe storage.

Jev Tools & CLIs#community#github

Community

LYchoon/paper-radar-jevPython

jev-bridge

Jev-style /v1/systemone API over any OpenAI-compatible backend, reading single-token label logprobs with a restricted softmax instead of generating JSON.

Jev Tools & CLIs#community#github

Community

TOSUKUi/jev-bridgePython

pyGPT

A desktop AI assistant with a Jev / System One inline plugin for classification, routing, verification, and scoring without chat-style output.

Jev Tools & CLIs#community#python#desktop

Community

szczyglis-dev/py-gptPython

judgekit

CLI and Python library that runs declarative YAML judgment tasks - classify, route, score, verify - natively on Jev Choice/Score/Noul, on any OpenAI-compatible backend, or on local keyword rules, with confidence gates, fallbacks and per-decision cost records.

Jev Tools & CLIs#community#yaml#classification

Community

lexingtonhibiki/judgekitPython

jev-tmmluplus-eval

Harness that sits Jev for the 66-subject Traditional Chinese TMMLU+ suite, asking every item as one four-way A/B/C/D choice with calibration reporting.

Jev Tools & CLIs#community#github

Community

lianghsun/jev-tmmluplus-evalPython

From the Jev guides

Original write-ups on what this entry is for.

Jev on the Gateways: Vercel, Workers AI, and OpenRouter

Oct 12, 2026

The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.

Jev Tools & CLIs#gateway#vercel#cloudflare

Read article

Exporting Decision Models: GGUF, ONNX, and Device Runtimes

Oct 10, 2026

Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.

Jev Tools & CLIs#open-models#local#onnx

Read article

Records at Rest: Documents, Inboxes, and Risk Queues

Sep 28, 2026

Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.

Jev Tools & CLIs#classification#email#security

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai