questions: { requested: boolean, issued: boolean, route: choice, // +1 }
NanoJev
A nano replica of Jev: parallel decisions, dynamic candidates, and an end-to-end training pipeline.

Every project here really calls Jev in its source · refreshed daily
questions: { requested: boolean, issued: boolean, route: choice, // +1 }
A nano replica of Jev: parallel decisions, dynamic candidates, and an end-to-end training pipeline.
questions: { department: choice, return_reason: choice, requested_resolution: choice, // +2 }
tiny Jev-like model built on top of Qwen2.5-0.5B you can train and run on your MacBook
questions: { queue: choice, escalate: boolean, urgency: score, // +1 }
A small open decision model: state + typed questions -> calibrated probabilities. A Jev / System One re-creation on Qwen3.5.
questions: { intent: choice, needs_reasoning: score, }
The open-source System One decision model. Sub-15ms, non-autoregressive, local drop-in alternative to TypeSafe Jev.
questions: { department: choice, urgent: boolean, frustration: score, }
OpenDecision is an open-source semantic decision engine like typesafe's jev.
questions: { binary: boolean, rating: score, }
Replicating Jev with a local LLM
questions: { department: choice, frustration: score, is_urgent: boolean, }
Jev-compatible System 开源Jev
questions: { enabled: boolean, color: choice, progress: score, }
Local JEV-style decisions with DiffusionGemma on Apple Silicon, with benchmarks and coding-agent examples.
questions: { composition: choice, legibility: boolean, }
Image-native typed decisions with shared visual encoding and Qwen3-VL
questions: { billing: boolean, urgent: boolean, team: choice, // +1 }
Jev-like model inference engine + Jev-compatible API
"adaptation": "Native Noul replaced with binary Yes/No Choice; state/question unchanged",
Semantic ifs from open models, on a 3090 at home. Independent; not affiliated with Jev or TypeSafe.
"""Jev-compatible HTTP API: POST /v1/systemone and GET /v1/models.
Open, Jev-compatible System One decision server on DiffusionGemma
@app.post("/v1/systemone", response_model=SystemOneResponse)A self-hosted drop-in replacement for TypeSafe's jev, powered by GliFormer.
"jev": j_val,
Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)
v = sub.add_parser("serve", help="HTTP server with the model loaded once (POST /score, /v1/systemone)")Open Jev implementation with custom finetuning
"jev": j_val,
Non-autoregressive decision engine on ModernBERT (151M) with calibrated uncertainty (RLCD), TypeSafe AI Jev benchmark audit, and in-browser WebGPU playground
keywords = ["jev", "typesafe-jev", "open-source-jev", "jev-alternative", "system-one", "system-one-model", "llm", "classification", "structured-decisions", "typ
Open-source alternative to TypeSafe's Jev: a System One style model layer that gives typed, calibrated decisions from any open-weights LLM in one forward pass (HF + vLLM), with honest benchmarks
for k, name in [("noul", "Noul (yes/no)"), ("score", "Score (rubric level)"), ("choice", "Choice")]:Open replica of TypeSafe's Jev: typed calibrated decisions in one forward pass, on Gemma 4 E2B / Gemma 3 270M (Modal)
# ---------------------------------------------------------------- TypeSafe Jev 互換 (POST /v1/systemone)
openvons (open-Jev): 有限選択肢に確率で答える判断層 — テキスト / 画像 / 日本語音声コマンド
TYPESAFE = os.environ.get("TYPESAFE_BASE_URL", "https://api.typesafe.ai").rstrip("/")Can we run something like Jev on a 3090 at home?
from .schema import Choice, Noul, Question, Score, question_from_dict
An optimized inference engine to turn LLMs into Jev-like machines: optimized for quick, lightweight, and accurate decision-making, classification, and scoring
"app": "jev",
Turning an LLM model into a Jev like System.
if self.path != "/v1/system_one":
OpenJev: an independent Jev-inspired System One decision API based on TypeSafe.ai concepts. Choice, score and noul primitives, local mock server, Python and TypeScript SDKs. Real inference planned; not affiliated with TypeSafe AI.
"""System One / "Jev"-style scorer.
Open-source Jev-style System One decision model. Gemma 3 270M with a scoring head — fast, calibrated decisions in a single forward pass. No text generation. Inspired by TypeSafe.ai's Jev.
"adaptation": "Native Noul replaced with binary Yes/No Choice; state/question unchanged",
Run SemIf (Jev-style semantic-if decisions) on a CPU — no GPU. Reads typed option probabilities straight from an open model in one forward pass, plus a web UI.
Requires TYPESAFE_API_KEY and: pip install typesafe-sdk
Fake autoregressive language model powered by TypeSafe Jev
Endpoint: POST https://openrouter.ai/api/alpha/decisions (model ~typesafe/jev-latest).
Jev notes: fourteen decisions I ran through it
import { TypeSafeClient, TypeSafeDecisionModel } from "@effect/ai-typesafe";A demonstration of the Jev System 1 model in Effect, matching vulnerabilities to their underlying CWEs
"jev" => Some(Self::Jev),
Argent + Jev
j = row.get("jev")open-Jev LM arm: Qwen2.5-0.5B + LoRA reproducing a hosted decision model's judgment at 92.9% on hand-labelled gold - trained overnight on a 6-vCPU CPU-only host, $0/call. Paper, corpora, harnesses, receipts.
from typesafe_sdk import Choice, TypeSafeClient
Turning the Jev classifier model into an autoregressive next token predictor
"""FastAPI server exposing Jev-compatible POST /v1/systemone."""
Toy local System One–style decision API (Jev-shaped). Not affiliated with TypeSafe.
"http://127.0.0.1:8768/v1/systemone",
Run jev-browser on a fully local JEV-style decision engine (no cloud API). Warm-browser fork, VRAM guard, measured benchmarks, run traces.
from .engine import FoqEngine, FoqConnectionError, SystemOneResponse
⚡ Foq — the FREE, local, open-source alternative to Jev. Typed System 1 decisions in ~25 ms — no waitlist, no cloud, no per-token cost. foq.fr
POST /v1/systemone {state, model, questions} -> {model, answers, usage}An open training and inference stack for Jev-style decision models. Train models to score dynamic candidate branches from a shared prefix, with support for high-cardinality choice, calibration, and fast batched inference.