EVERY ENTRY HAS RECEIPTS

AWESOME
JEV

Open-source directory

Every project here really calls Jev in its source · refreshed daily

788projects
182runnable here
765code receipts

ALL 788 PROJECTSEvery project

experiments/browser-chaos/src/jev-driver.ts
questions: {
  stuck: boolean,
  destructive: boolean,
}
runnable
@mizchi17

jev-playground

TypeSafe AI の System One モデル Jev を MoonBit から触るためのプレイグラウンド。

Benchmarks & researchTypeScript
src/evals/judges/system_one.py
questions: {
  does_pass: boolean,
}
runnable
@danielgshea12

jev-as-a-judge

Using Jev as an evaluator.

Benchmarks & researchPython
extensions/eutrya-adaptive-extension/src/router.mjs
questions: {
  route: choice,
  toolFree: boolean,
}
runnable
@hellozenstrategist-lab8

eutrya

Jev-native AI security harness for autonomous research, multi-agent swarms, persistent hunt boards, and long-running agent workflows. CLI-first, open source, and built for authorized security research.

Benchmarks & researchJavaScript
tests/test_with_model.py
questions: {
  a: boolean,
  b: boolean,
}
runnable
@aabolfazl6

typesafe-local

Inspired by TypeSafe Ai, Ask a local LLM typed questions, get calibrated probabilities instead of text. Structured output without generation or parsing. MLX / Apple Silicon.

Benchmarks & researchPython
test/tool.test.js
questions: {
  urgent: boolean,
}
runnable
@arunav255

jev-mcp

Connect JEV to MCP clients and compare its judgments against general-purpose LLMs using shared datasets and measurable accuracy.

Benchmarks & researchJavaScript
src/decisions.ts
questions: {
  instructions: boolean,
}
runnable
@thejorgg5

omp-jev

TypeSafe Jev routing for Oh My Pi, with an opt-in checkpoint orchestrator and editable XDG configuration. Requires Bun = 1.3.14 and OMP = 18.2.3.

Benchmarks & researchTypeScript
src/jevchat/state.py
questions: {
  done: boolean,
}
runnable
@adhyaay-karnwal3

jev-chat

A chatbot from typed Jev decisions: hierarchical speculative decoding over System One probabilities.

Benchmarks & researchPython
jev-browse.mjs
questions: {
  done: boolean,
  blocked: boolean,
}
runnable
@kyrylosyzonenko3

jev-browse

jev-browse is an unofficial project and isn't affiliated with TypeSafe or Vercel.

Benchmarks & researchJavaScript
web/home.js
questions: {
  what: choice,
}
runnable
@4anti2

jev-broadcast-lab

Operator lab for TypeSafe Jev. Chess Arena, closed-schema booths, Stockfish HUD for review only.

Benchmarks & researchJavaScript
experiments/probe.py
questions: {
  department: choice,
}
runnable
@FirasSX9142

Janus

Measure when to use Jev and other models on your data, then route accordingly.

Benchmarks & researchPython
src/eval/builtin.ts
questions: {
  task_completion: score,
  instruction_following: boolean,
  grounded_in_evidence: boolean,
  // +5
}
runnable
@memovai2

openevals

Affordable for parallel online agent evals and observability. Powered by JEV.

Benchmarks & researchTypeScript
examples/doom/server/src/session-jev-trace.test.ts
questions: {
  action: choice,
}
runnable
@superradcompany2

multiverse-of-madness

Jev and Microsandbox explore alternate game futures with a reusable TypeScript learning harness

Benchmarks & researchTypeScript
src/gateway.mjs
questions: {
  label: choice,
}
runnable
@acoyfellow1

nightglass

Owned, deterministic classifier for checking whether agent claims are supported by evidence, with optional Jev comparison through Cloudflare AI Gateway.

Benchmarks & researchJavaScript
firewall/judge.py
questions: {
  is_urgent: boolean,
  severity: score,
}
runnable
@AnshChoudhary1

typesafe-ai-firewall

Shadow-mode validation harness for a pre-execution firewall on AI agent tool calls (TypeSafe/Jev). Real run, findings in report.md.

Benchmarks & researchPython
experiments.py
questions: {
  is_regulation: boolean,
  mentions_deadline: boolean,
  mentions_fines: boolean,
  // +5
}
runnable
@copyleftdev1

jev-labs

Never confidently wrong: a TLA+-verified consensus kernel around TypeSafe's Jev, run through 1,680 chaos-tested pharmacy decisions with zero wrong verdicts. Film, code, and every captured call.

Benchmarks & researchPython
packages/lab/src/scenarios.ts
questions: {
  topic: choice,
  requests_credentials: boolean,
  sender_identity_mismatch: boolean,
  // +4
}
runnable
@danielhirt1

jev-lab

Experiments on TypeSafe Jev (System One decision model) via OpenRouter: repeatability, perturbation, and LLM baseline comparison

Benchmarks & researchTypeScript
berserk/global-payment-incident/demo.py
questions: {
  failure_domain: choice,
  customer_impact: score,
  systemic_outage: boolean,
}
runnable
@JGalego1

Jevs-Garage

A garage full of tiny experiments for building critical systems with System One & Jev 🔧🧠⚡

Benchmarks & researchPython
samples/comment_triage/triage.py
questions: {
  impact: score,
}
runnable
@pavy231

jev_typesafeai_test

TypeSafe AI (jev) 시범 사용 프로젝트

Benchmarks & researchPython
test_jev.py
questions: {
  is_urgent: boolean,
  needs_human: boolean,
  is_financial: boolean,
}
runnable
@TanayPadar1

gpt-vs-jev

Compare GPT generated language with JEV structured Noul decisions on the same input.

Benchmarks & researchTypeScript
jev_visual/cli.py
parser = argparse.ArgumentParser(description="Local visual Choice/Noul/Score judgments")
@hr98w152

jev-visual

An educational Jev-like visual inference experiment on Apple Silicon: shared context, direct candidate scoring, and local visual demos.

Benchmarks & researchPython
src/smartmoney_cub_harness/jev/direct.py
base_url: str = "https://api.typesafe.ai",
@myc057624

SmartMoney-Cub

Read-only trading journal and review harness: Jev typed judgments, agent integration, and a reproducible finance benchmark. No orders, no advice.

Benchmarks & researchPython
scripts/common/jev_client.py
"""回傳一個已設定好 key 的 TypeSafeClient;拿不到 key 就直接退出、說清楚為什麼。
@Zaious19

jev-capability-atlas

Independent, evidence-based map of when TypeSafe's Jev actually holds up vs. breaks down — real API-call receipts, not a leaderboard. 中文為主的雙語 repo。

Benchmarks & researchPython
public_eval_benchmark.py
jev_raw=label['jev'], jev_reported=label['jev_reported'])
@JoshuaSP17

open-jev

Typed JSON inference with DiffusionGemma, with Every and Jev benchmark results

Benchmarks & researchPython
lib/racers.mjs
const json = await postJson(fetchImpl, "https://api.typesafe.ai/v1/systemone", key, jevRequest(batch, model));
@goodrahstar16

jev-column-race

Jev vs Gemini 3.8 Flash: labelling 1,000 app reviews, 4.1× faster and 7× cheaper

Benchmarks & researchJavaScript
src/jev_benchmarks/adapters/jev.py
from typesafe_sdk import Choice, TypeSafeClient
@AbdelStark11

jev-benchmarks

Probability-aware evaluation for typed decision models: calibration, selective risk, latency, and reproducible benchmarks.

Benchmarks & researchPython
src/config.mjs
id: i ? 'deepseek' : 'jev',
@NanmiCoder9

jev-arena

Jev 模型介绍与实测:通过 Choice / Score / Noul 将自然语言转为带类型的判断与概率,用于分类、评分和路由;支持与 DeepSeek 等模型对比评论打标、速度与结果,含 CSV/Excel 导入、原速回放与离线报告。

Benchmarks & researchJavaScript
src/jevcal/providers/typesafe.py
DEFAULT_BASE_URL = "https://api.typesafe.ai"
@abhixhek8

jevcal

Stop guessing confidence thresholds: calibrate, threshold, and drift-check typed decision models (TypeSafe Jev) against an LLM teacher.

Benchmarks & researchPython
src/jev_rag_benchmark/manifest.py
TRACKED_PACKAGES = ["httpx", "numpy", "pydantic", "PyYAML", "typer", "typesafe-sdk"]
@erendikmenn6

jev-rag-benchmark

Reproducible benchmark for measuring Jev reranking quality, latency, and cost in RAG

Benchmarks & researchPython
jevbench/adapters/__init__.py
* TypeSafe-compatible /v1/systemone (native typed interface).
@fstandhartinger5

jevbench

JevBench v1 - a benchmark for Jev-class typed decision models: smart, cheap, fast, reliable, open.

Benchmarks & researchPython
jevbench/medqa_run.py
from typesafe_sdk import Choice, RetryPolicy, TypeSafeClient
@mahlernim5

jev-korean-benchmark

Reproducible early-access evaluation of Jev on Korean understanding and medical text, with runtime and cost evidence

Benchmarks & researchPython
main.py
import typesafe_sdk
@opaielsheikh5

ai-elo-ranker

High-speed recursive AI Elo tournament engine powered by Jev and Swiss matchmaking

Benchmarks & researchPython
src/api.ts
export const ENDPOINT = "https://api.typesafe.ai/v1/systemone";
@y0usaf5

jev-lm

A word-level language model whose output layer is Jev: n-gram drafter, Noul chunk verification, bits-per-token eval

Benchmarks & researchTypeScript
jev_agent.py
self.model = os.environ.get("JEV_MODEL", "typesafe-ai/jev")
@Astro-Han4

jev-harness

A coding agent that filters every tool result through Jev before the model sees it, with an A/B harness measuring pass@1 and cost against the unfiltered control

Benchmarks & researchPython
src/jev.mjs
import { experimental_evaluate as evaluate } from 'ai';
@caiovicentino4

jev-align

Calibrated alignment verifier for LLM responses and agent plans — powered by Jev

Benchmarks & researchJavaScript
determinest/clients.py
for name,url,key in [('openrouter','https://openrouter.ai/api/v1/models',auth['OPENROUTER_KEY']),('typesafe','https://api.typesafe.ai/v1/models',auth['TYPESAFE_
@gemanor4

jev-code-review-benchmark

Comparing Jev, Gemini Flash, and Claude Fable on Python code review rules: cost, speed, accuracy, and consistency. Includes results, charts, and reproducible experiments.

Benchmarks & researchPython
demo/server.mjs
// api.typesafe.ai allows no localhost CORS origins (and keys belong server-side).
@lbotinelly4

jev-little-airways

A show-and-tell capability study for Jev, TypeSafe's System One decision model.

Benchmarks & researchHTML
src/jse/cli.py
"jev": jev[k], "llm": llm[k], "label": None}
@zhuyansen4

jev-search-rerank-eval

Does a TypeSafe Jev rerank beat embedding search? Graded relevance eval (9,831 pairs, 164 zh/en queries) over the Agent Skills Hub catalog, with the judge-circularity bias measured.

Benchmarks & researchPython
src/confidence.ts
import type { Questions } from "@typesafe-ai/sdk";
@AntonioCoppe3

jev-harness

Decision harness for TypeSafe Jev — confidence gates, shadow mode, recipes, and evals. Claude CLI 48.9s → Jev 1.3s on the same row-filter job.

Benchmarks & researchTypeScript
behavior_study.py
req=urllib.request.Request('https://api.typesafe.ai/v1/systemone',data=body,headers={'Authorization':'Bearer '+key,'Content-Type':'application/json','Cache-Cont
@RINNECODER3

jev-behavior-study

Independent Jev 1.13.0 behavior study: report, controlled prompt experiments, raw results, and offline verification.

Benchmarks & researchPython
app/server.py
from typesafe_sdk import AsyncTypeSafeClient
@0xnairb2

research_desk

TypeSafe Jev demonstration for new analyzation — experimenting with Jev for fast analysis of news and tickers

Benchmarks & researchPython
src/rag_jev/cli.py
from typesafe_sdk import TypeSafeError
@EmreKaplaner2

rag-jev

Make room for useful evidence. Inspectable context selection for RAG, with Jev reranking and open benchmark studies.

Benchmarks & researchPython
internal/bench/run.go
// Probability is the Noul answer: how likely Jev thinks the positive is.
@Gaurav-Gosain2

jev-sec-bench

Blind security benchmarks for Jev, TypeSafe's System One model: prompt injection and vulnerable code detection, built on jev-go

Benchmarks & researchGo
src/trade_jev/policies.py
from typesafe_sdk import AsyncTypeSafeClient, Choice
@justinhe162

trade-jev

Backtest Jev (TypeSafe) as a BUY/SELL/HOLD trader on NQ L10 order-book data

Benchmarks & researchPython
src/jev_frontend_qa/core/model_client.py
Calls the real ``https://api.typesafe.ai/v1/systemone`` endpoint. The
@Nainish-Rai2

jev-frontend-qa

Evidence-driven frontend QA built on Jev Ultrafast and Browser Harness, with a synthetic todo demo.

Benchmarks & researchPython
run.py
d, wall, srv = post("api.typesafe.ai", "/v1/systemone", os.environ["TYPESAFE_API_KEY"], {"model": "jev-1.13.0", "state": state, "questions": questions})
@OmarMujahid2

jev-decision-bench

An independent benchmark of TypeSafe's Jev, a model that does not write text. You send it some content and a list of typed questions (yes/no, pick one option, rate on a scale) and

Benchmarks & researchPython
jevlab/client.py
api.typesafe.ai; the CLI, the verifier and the lab harness all go through it.
@SamuelSacco2

jev-exploration

Jev (TypeSafe) exploratory thread: claim audit, live demos, and runnable code

Benchmarks & researchPython
scripts/jev_eval.mjs
import { experimental_evaluate as evaluate } from 'ai';
@scienthoon2

jev-ood-calibration

Independent calibration test of TypeSafe's Jev on a task it cannot have seen: 900 rule-generated support tickets (choice / score / boolean) plus 3 public benchmarks via Vercel AI Gateway. Raw responses, ECE with noise floor, temperature refit, per-type sign of miscalibration. Reproducible for ~$0.06.

Benchmarks & researchPython
src/components/Playground.tsx
const curlCmd = `curl -i -X POST https://api.typesafe.ai/v1/systemone \\
@sysadarsh2

zerosweep

Autonomous System-One Triage Engine & Benchmark powered by TypeSafe AI (Jev). 75ms inference, $0 output tokens, and RLCD epistemic safety gates.

Benchmarks & researchTypeScript
jevchess/experiments.py
from typesafe_sdk import Choice, Noul, Score
@wondertwins2

jev-benchmark

Benchmarks and a playground for TypeSafe's Jev (System One) model: chess, and who-is-the-player-talking-to for speech-to-text game NPCs

Benchmarks & researchPython
src/jnc/prior.py
MODEL = "~typesafe/jev-latest"
@zhuyansen2

jev-news-cold-start

Cross-domain check on MIND news: a zero-shot Jev headline prior is worth ~500 labelled articles, adds +0.069 ρ as features, and lifts a Thompson-sampling cold start by 25%.

Benchmarks & researchPython
evaljev/runners.py
JEV_URL = "https://api.typesafe.ai/v1/systemone"
@4esv1

jev-eval

Benchmark TypeSafe Jev against any OpenRouter model on your own labelled classification data: accuracy, calibration, latency, cost

Benchmarks & researchPython
run_bench.py
MODEL = "typesafe/jev-1.13"
@alperenerol1

jev-1.13-mini-benchmark

Mini benchmark of TypeSafe's jev-1.13 structured decision model (OpenRouter Decisions API) on labeled support-triage: noul/choice/score, consistency, cost, lessons learned

Benchmarks & researchPython
lib/client.ts
const BASE_URL = "https://api.typesafe.ai";
@amr050081

jev-sandbox

Test bench for TypeSafe's Jev

Benchmarks & researchTypeScript
net_floor.py
typesafe = env("TYPESAFE_BASE_URL", "https://api.typesafe.ai").rstrip("/")
@anisselbd1

jev-phishing-bench

Jev (TypeSafe) vs Claude Haiku 4.5 on 2 000 phishing emails: accuracy, calibration, latency, cost. Reproducible benchmark.

Benchmarks & researchPython
generate_heatmap.py
API_URL = "https://api.typesafe.ai/v1/systemone"
@CompleteTech-LLC-AI-Research1

jev-311-heatmap

NYC 311 complaint heatmaps with TypeSafe JEV: reproducible pipeline, live research results, and interactive geographic visualizations.

Benchmarks & researchHTML
src/server.ts
import { TypeSafeClient } from "@typesafe-ai/sdk";
@dayhaysoos1

jevals

Local evaluation workbench for TypeSafe Jev

Benchmarks & researchTypeScript
scripts/eval.py
ENDPOINT = "https://api.typesafe.ai/v1/systemone"
@ElshinQ1

jevaluate

Jevaluate: evaluate before you trust. Field notes, runnable scripts and an agent skill for TypeSafe Jev: gated evals, a browser loop, a product walk with DeepSeek vision, a UI text judge and a first-click tree test. Co-authored with Claude Fable 5.1.

Benchmarks & researchJavaScript
apps/chat-lab/src/shared.ts
export const MODELS = { chat: 'google/gemini-3.7-flash', decision: 'typesafe/jev-1.13' } as const;
@fabricioism1

jev-expirements

A repository for jev-expirements

Benchmarks & researchTypeScript
eval.js
let md = `# Results\n\nGenerated by \`node eval.js\` on ${new Date().toISOString().slice(0, 10)} with \`${process.env.JEV_MODEL ?? 'typesafe/jev-1.13'}\`.\n\n`;
@finetuningsingh1

jev-chatbot

Experiment: using TypeSafe Jev as a chatbot by choosing replies one letter or word at a time

Benchmarks & researchJavaScript
runner/jev_client.py
ENDPOINT = "https://api.typesafe.ai/v1/systemone"
@integrate-your-mind1

jev-nethack

Jev x NetHack: bounded runner, research code, and completed recording releases

Benchmarks & researchPython
common/jev_client.py
"""Thin wrapper around TypeSafe AI's official Python SDK (`typesafe-sdk`,
@jimmyliao1

jev-storyboard-lab

Google ADK vs Microsoft Agent Framework for structured-output agents, with TypeSafe Jev as a vendor-neutral QC gate

Benchmarks & researchPython
src/typed_decisions/hole_eval.py
ap.add_argument("--model", default=os.environ.get("JEV_MODEL", "typesafe/jev-1.13"))
@kotoba-lang1

typed-decisions

Jev-shaped typed-decision model (state + Choice/Score/Noul questions -> calibrated probabilities, one pass) on ModernBERT / DeBERTa / LLaDA-MoE, with measured latency, accuracy, calibration and training cost

Benchmarks & researchPython
pipeline.py
API_URL = "https://api.typesafe.ai/v1/systemone"
@markfive-proto1

typesafe-vs-deepseek

TypeSafe (Jev) vs DeepSeek-flash: side-by-side speed/token/cost/accuracy comparison across invoice extraction, email classification, and reranking

Benchmarks & researchPython
src/typesafe.rs
.unwrap_or_else(|_| "https://api.typesafe.ai/v1/systemone".into()),
@markjaquith1

typesafe-ai-playground

A playground for experiments around Jev, TypeSafe's System One model.

Benchmarks & researchRust
parsing-experiments/src/jev_eval/runners/jev.py
import typesafe_sdk as sdk
@naveenreddy611

jev-experiments

experiments with system one model jev

Benchmarks & researchPython
src/config.py
JEV_URL = "https://api.typesafe.ai/v1/systemone"
@PistachioAIHQ1

jev-synergy-screening

Jev (TypeSafe System One) × ASReview SYNERGY abstract screening demo — Choice/Noul vs gold labels

Benchmarks & researchPython
src/app.js
if (kind === 'typesafe') return 'https://api.typesafe.ai/v1/systemone';
@pokertools-arena1

pokertools-arena.github.io

A browser-first AI poker benchmark. Seat Jev and OpenAI-compatible models at the same no-limit Texas Hold'em table, watch every card and decision as a spectator, and let the tournament run autonomously until one model wins.

Benchmarks & researchJavaScript
internal/jev/client.go
APIURL       = "https://api.typesafe.ai/v1/systemone"
@rogeriochaves1

jev-experiments

Results page: https://claude.ai/code/artifact/f00ee126-9554-4e2f-b2e7-1fc86c066aa9

Benchmarks & researchGo
run.py
URL = "https://api.typesafe.ai/v1/systemone"
@ruffood1

jev-reality-check

Jev (TypeSafe AI) 可复现实测:算术、计数、日期、零幻觉、把握度校准,中英对照

Benchmarks & researchPython
benchmarks/detectionai.py
from typesafe_sdk import AsyncTypeSafeClient, RetryPolicy
@sandroandric1

JevGram

AI detection in research papers with Jev

Benchmarks & researchPython
src/jev.ts
import { choice, TypeSafeClient, type EntryType } from '@typesafe-ai/sdk';
@scarif-labs1

jev-software-decision-benchmark

Reproducible benchmark evaluating JEV as a software decision primitive for dependency-update automation under distribution shift.

Benchmarks & researchTypeScript
src/run.py
req=urllib.request.Request('https://api.typesafe.ai/v1/systemone',data=json.dumps(p).encode(),headers={'Authorization':'Bearer '+self.key,'Content-Type':'applic
@simonmesmith1

jev-bbq-experiment

Reproducible evaluation of TypeSafe Jev on all 58,492 BBQ questions: accuracy, stereotype bias, uncertainty, cost and latency.

Benchmarks & researchR
server.js
import { experimental_evaluate as evaluate } from 'ai';
@sstehniy1

jev-calculator

iOS 6-inspired Jev calculator demo with a lifetime API budget

Benchmarks & researchTypeScript
src/transport.rs
.post("https://api.typesafe.ai/v1/systemone")
@Tech-Byte-Frontier1

jevgate

File-scoped maintainability review with TypeSafe Jev

Benchmarks & researchRust
apps/api/src/playground/config.py
jev_api_url: str = "https://api.typesafe.ai/v1/systemone"
@tedliou1

decision-model-playground

A local playground for comparing Laya and Jev decision models with article recommendations.

Benchmarks & researchPython
evaluator.js
const JEV_URL = process.env.JEV_URL || 'https://api.typesafe.ai/v1/systemone';
@Thanh-Mathieu951

jev-model-tokengate

Every token passes the gate before the screen.

Benchmarks & researchJavaScript
src/jevbench/backends/jev.py
DEFAULT_ENDPOINT = "https://api.typesafe.ai/v1/systemone"
@TokenTrim1

jev-agent-failure-benchmark

Benchmarking Jev (Typesafe.ai) against a strong LLM on the Who&When Pro agent-failure-attribution benchmark (text subset).

Benchmarks & researchPython
jev_router/jev.py
"""Client for TypeSafe's Jev decision model (POST /v1/systemone).
@TokenTrim1

jev-routing-experiment

Benchmarking TypeSafe's Jev decision model as a cost-efficient LLM router on RouterArena

Benchmarks & researchPython
src/jcp/prior.py
MODEL = "~typesafe/jev-latest"
@zhuyansen1

jev-cold-start-prior

Can a TypeSafe Jev prior read from a README on day one predict which new agent-skill repos gain stars? Zero-shot Jev ≈ a text model trained on ~150 labels; best used as a feature. Prospective test running.

Benchmarks & researchPython
src/jip/label.py
MODEL = "~typesafe/jev-latest"
@zhuyansen1

jev-issue-pulse

Can a Jev-labelled GitHub issue stream catch a broken release before the fix? No at daily cadence (null, n=7). Per issue, Jev matches triage labels far better than keywords or sentiment.

Benchmarks & researchPython
src/jsp/label.py
MODEL = "~typesafe/jev-latest"
@zhuyansen1

jev-support-pulse

Does a Jev-labelled support-tweet stream spike before a brand admits an outage? At equal false alarms it catches 17 vs 10 incidents (volume), ~4h ahead; a good keyword list is almost as good.

Benchmarks & researchPython
scripts/run_baseline.py
python scripts/run_baseline.py --model jev --base-url https://api.typesafe.ai/v1 --api-key-env TYPESAFE_API_KEY
@zsavage81

padflow-jev-evals

Typed-decision benchmark from PadFlow (land development SaaS): schemas, anonymized labeled rows, and a runner for confidence-calibrated models like TypeSafe Jev.

Benchmarks & researchPython
worker.js
const TYPESAFE_ENDPOINT = 'https://api.typesafe.ai/v1/systemone';
@zzzzzec1

jevsort

Jev-powered integer sorting experiment: serial selection versus parallel rank prediction.

Benchmarks & researchHTML