JevCode / 生态

Jev 覆盖地图

这里收录 571 个项目，每条写明 Jev 在其中做的具体决策，并给出可点开的证据链接。验证标签说明证据强度，不是评分。

复核于: 2026-10-29

官方资源

TypeSafe agent skills
把 Jev 的原语、模式与评估写法打包成 agent 可加载的技能包，让代码 agent 在生成调用时先确定答案类型。
证据https://github.com/typesafe-ai/skills
验证标签public-sourcemaintainer-runbook
System One adapter (Python)
以同一套客户端接口换到底层 LLM 上比较 Jev 与聊天模型的决策质量，客户端代码不改。
证据https://github.com/typesafe-ai/system-one-adapter-python
验证标签public-sourcemaintainer-runbook

SDK 与客户端

TypeSafe JavaScript SDK
在 TypeScript 侧把调用返回收窄成声明的答案类型，使决策结果不需要再做运行时解析。
证据https://github.com/typesafe-ai/typesafe-sdk-js
验证标签public-sourcemaintainer-runbook
laya-php
在 PHP 侧把 Jev 兼容的类型化决策映射进原生类型，接管 Laravel 的判定点。
证据https://github.com/marcreichel/laya-php/blob/main/src/Laya.php#L197
验证标签public-sourcecall-site-verified
go-jev
Go 客户端把 Jev 的类型化答案封装成库，并附带 jev-cli 供 UNIX 管道直接调用。
证据https://github.com/mattn/go-jev/blob/main/jev.go#L19
验证标签public-sourcemaintainer-runbookcall-site-verified
zio-typesafe-ai
Scala 3 / ZIO 库对 Jev 做端到端类型化封装，避免调用方直接处理原始 JSON。
证据https://github.com/jamesward/zio-typesafe-ai/blob/main/src/main/scala/com/jamesward/zio_typesafe_ai/TypeSafeAI.scala#L3
验证标签public-sourcecall-site-verified
jev-java
Java 17+ 客户端库，把 Jev System One 决策引擎包成惯用 Java 接口。
证据https://github.com/Olti1947/jev-java/blob/main/src/main/java/io/github/Olti1947/jev/JevClient.java#L137
验证标签public-sourcecall-site-verified
jev-sdk-java
面向 Java 21 的类型安全客户端，请求 TypeSafe AI 的 Jev 决策 API。
证据https://github.com/luigivis/jev-sdk-java/blob/main/src/main/java/com/luigivismara/jev/JevClient.java#L156
验证标签public-sourcecall-site-verified
typesafe-sdk (Rust)
Rust 客户端库，除默认异步用法外还提供 blocking 特性供同步调用。
证据https://github.com/codeitlikemiley/typesafe-sdk-rust/blob/main/benches/client_overhead.rs#L15
验证标签public-sourcemaintainer-runbookcall-site-verified
kunobi-jev
Rust 客户端在 Jev 之外还接入其他托管决策模型，用同一套类型化调用切换提供方。
证据https://github.com/kunobi-ninja/kunobi-jev/blob/main/src/blocking.rs#L27
验证标签public-sourcecall-site-verified
tinyjevclient
轻量 Rust 客户端，只覆盖 System One API 的类型化调用面。
证据https://github.com/tinyhumansai/tinyjevclient/blob/main/README.md#L3
验证标签public-sourcereadme-claim
typesafe.zig
Zig 绑定把 Choice、Noul、Score 三类问题合成一次调用，直接返回类型化答案。
证据https://github.com/mattneel/typesafe.zig/blob/master/src/Client.zig#L353
验证标签public-sourcecall-site-verified
openai-scala-client
通用 LLM 客户端把决策模型作为并列的一类 API，Jev 的 System One 与 OpenAI 的 Decisions 同层接入。
证据https://github.com/cequence-io/openai-scala-client/blob/master/typesafe-client/src/main/scala/io/cequence/openaiscala/typesafe/JsonFormats.scala#L125
验证标签public-sourcecall-site-verified
decisionkit-dotnet
.NET 决策引擎把提供方做成可替换层，Jev 是其受支持后端之一。
证据https://github.com/iamjonatha/decisionkit-dotnet/blob/main/README.md#L3
验证标签public-sourcereadme-claim
typesafeai-dotnet-sdk
社区维护的 .NET 独立客户端，直接对接 System One HTTP API，README 明示非官方身份。
证据https://github.com/saibimajdi/typesafeai-dotnet-sdk/blob/main/src/TypeSafe.Sdk/ITypeSafeClient.cs#L34
验证标签public-sourcecall-site-verified
Bodila51/grok-bot-jev
README 自述：A Python router around the TypeSafe SDK's system_one call.
证据https://github.com/Bodila51/grok-bot-jev/blob/main/jevopus/jevopus.py#L22
验证标签call-site-verified
boozedog/pi-codemode
README 自述：When a TypeSafe API key is present, Codemode injects guest jev.ask into QuickJS and the type checker. Without a key, Codemode is unchanged: no jev global, no HTTP, and no @typesafe-ai/sdk dependency.
证据https://github.com/boozedog/pi-codemode/blob/master/src/jev/client.ts#L5
验证标签call-site-verified
chaitin/Decis
README 自述：official typesafe-sdk already speaks — so pointing the SDK at your own host is the whole
证据https://github.com/chaitin/Decis/blob/master/playground/__init__.py#L3
验证标签call-site-verified
codeitlikemiley/typesafe-sdk-rust
README 自述：typesafe-sdk = { version = "0.3", features = ["blocking"] }
证据https://github.com/codeitlikemiley/typesafe-sdk-rust/blob/main/benches/client_overhead.rs#L15
验证标签call-site-verified
dakdevs/decide-mcp
README 自述：Defaults to TypeSafe AI Jev through AI Gateway, using AI SDK 7's experimental_evaluate. Supports custom decision policies, multiple bias profiles, automatic profile routing, and individual tools per profile.
证据https://github.com/dakdevs/decide-mcp/blob/main/src/providers.ts#L5
验证标签call-site-verified
deepanwadhwa/OpenDecision
README 自述：A Python API, a FastAPI server, and a TypeSafe SDK compatible endpoint.
证据https://github.com/deepanwadhwa/OpenDecision/blob/main/demos/images/measure_latency.py#L79
验证标签call-site-verified
emrickgarrett/OneVOneJev
README 自述：AI: @typesafe-ai/sdk (jev-latest) with heuristic fallback
证据https://github.com/emrickgarrett/OneVOneJev/blob/main/server/src/jev.ts#L7
验证标签call-site-verified
iapp-technology/openthai-systemone
README 自述：mirrors TypeSafe AI's POST /v1/systemone (Jev), so SDK code written for TypeSafe can point at this model.
证据https://github.com/iapp-technology/openthai-systemone/blob/main/demo/doom_demo.py#L25
验证标签call-site-verified
InsaneArts/typesafe-sdk-swift
README 自述：package(url: "https://github.com/InsaneArts/typesafe-sdk-swift", from: "0.2.0")
证据https://github.com/InsaneArts/typesafe-sdk-swift/blob/main/README.md#L10
验证标签readme-claim
kunobi-ninja/kunobi-jev
README 自述：Rust client for typed decision models: TypeSafe Jev, Cloudflare Clef, Liquid d1,
证据https://github.com/kunobi-ninja/kunobi-jev/blob/main/src/blocking.rs#L27
验证标签call-site-verified
lambertsj/beatjev
README 自述：It's a Cloudflare Worker. public/ is served as static assets (plain HTML/CSS/JS, no build step). src/worker.js handles POST /api/classify and calls Jev through @typesafe-ai/sdk, so the API key stays server-side.
证据https://github.com/lambertsj/beatjev/blob/main/src/worker.js#L6
验证标签call-site-verified
logan-markewich/jeff
README 自述：Use the official typesafe-sdk by pointing TYPESAFE_BASE_URL at jeff.
证据https://github.com/logan-markewich/jeff/blob/main/bench/eval_accuracy.py#L70
验证标签call-site-verified
luigivis/jev-sdk-java
README 自述：A small, type-safe Java 21 client for the TypeSafe AI Jev (System One) decision API.
证据https://github.com/luigivis/jev-sdk-java/blob/main/src/main/java/com/luigivismara/jev/JevClient.java#L156
验证标签call-site-verified
MoLeMo-Lab/mojev
README 自述：wire-compatible with the TypeSafe Python SDK's system_one API.
证据https://github.com/MoLeMo-Lab/mojev/blob/master/mojev/apitest.py#L48
验证标签call-site-verified
nshkrdotcom/typesafe_sdk
README 自述：The TypeSafe-specific provider/API SDK. It owns TypeSafe authentication,
证据https://github.com/nshkrdotcom/typesafe_sdk/blob/main/codegen/typesafe_sdk/codegen/source/openapi.ex#L20
验证标签call-site-verified
Olti1947/jev-java
README 自述：Idiomatic Java 17+ client library for TypeSafe AI's Jev System One decision engine.
证据https://github.com/Olti1947/jev-java/blob/main/src/main/java/io/github/Olti1947/jev/JevClient.java#L137
验证标签call-site-verified
Premo-Cloud/typesafe-sdk-java
README 自述：implementation("io.github.premo-cloud:typesafe-sdk:0.6.0")
证据https://github.com/Premo-Cloud/typesafe-sdk-java/blob/main/typesafe-sdk/src/main/java/io/github/premocloud/typesafe/TypeSafeClient.java#L54
验证标签call-site-verified
saibimajdi/typesafeai-dotnet-sdk
README 自述：This is a community SDK. It is an independent client for the TypeSafe AI System One HTTP API,
证据https://github.com/saibimajdi/typesafeai-dotnet-sdk/blob/main/src/TypeSafe.Sdk/ITypeSafeClient.cs#L34
验证标签call-site-verified
ticofab/scala-jev-sdk
README 自述：libraryDependencies += "io.github.ticofab" %% "scala-jev-sdk" % " "
证据https://github.com/ticofab/scala-jev-sdk/blob/master/README.md#L14
验证标签readme-claim
tinyhumansai/tinyjevclient
README 自述：tinyjevclient is a typed Rust client for TypeSafe AI's System One API and
证据https://github.com/tinyhumansai/tinyjevclient/blob/main/README.md#L3
验证标签readme-claim
vicfei/awesome-jev-prompts
README 自述：import { choice, noul, score, TypeSafeClient } from "@typesafe-ai/sdk";
证据https://github.com/vicfei/awesome-jev-prompts/blob/main/README.md#L96
验证标签readme-claim

Jev 类模型

ollaya
用 Rust 实现与 Jev 同形的判定服务，让服务端不依赖托管模型也能拿到类型化答案。
证据https://github.com/ollaya-dev/ollaya/blob/main/crates/ollaya-server/src/http.rs
验证标签public-sourcecall-site-verified
JevAny
以模型层复刻 Jev 的判定接口，供无法直连托管服务的环境使用。
证据https://github.com/weitianxin/JevAny/blob/main/jevany/model.py
验证标签public-sourcecall-site-verified
eikos
把判定步骤从生成链路里拆出来，交给类型化模型单独返回结论。
证据https://github.com/caiovicentino/eikos/blob/main/eikos/serve.py#L343
验证标签public-sourcecall-site-verified
abhishek085/open-spark-jev
README 自述：An open-source, local implementation inspired by TypeSafe's Jev and System One models, built for NVIDIA DGX Spark. The model is called spark-s1 .
证据https://github.com/abhishek085/open-spark-jev/blob/main/open_spark_jev/experimental/control_es.py#L11
验证标签call-site-verified
alvarobartt/sys1
README 自述：Blazing fast structured decisions for System One open-models self-hosted with a TypeSafe AI compatible API, written in Rust.
证据https://github.com/alvarobartt/sys1/blob/main/benchmark.py#L791
验证标签call-site-verified
DECRUX9812/openjev
README 自述：An open, local, zero-cost reimplementation of the Jev decision layer for job postings.
证据https://github.com/DECRUX9812/openjev/blob/main/README.md#L3
验证标签readme-claim

框架与集成

DeciFlow
以类型化决策节点编排流程，分支条件由 Jev 返回的枚举值直接决定。
证据https://github.com/call518/DeciFlow
验证标签public-sourcedemo-only
jev4pg
在 PostgreSQL 里把文本谓词交给 Jev 判定，保留概率与来源证据，精确的连接与聚合仍由 SQL 完成。
证据https://github.com/Sheltercosmo/jev4pg/blob/main/sdd/evaluators.py#L39
验证标签public-sourcecall-site-verified
Gut Check
由 Jev 评估 Home Assistant 实体的健康度与升级风险，低置信度时不允许自动改动。
证据https://github.com/funkadelic/ha-gutcheck/blob/main/custom_components/gutcheck/budget.py#L23
验证标签public-sourcecall-site-verified
zod-jev
用一次 parseAsync 把整个 Zod schema 的规则合成单个 Jev 请求，再把概率转成 Zod issue，调用方仍用原生 Zod API。
证据https://github.com/jomatsu/zod-jev/blob/main/scripts/check-package.mjs#L35
验证标签public-sourcecall-site-verified
spring-ai-typesafe
基于 Spring RestClient 与 Jackson 3 实现 Jev API，并提供 Spring AI 集成。
证据https://github.com/spring-ai-community/spring-ai-typesafe/blob/main/typesafe-java-sdk/src/main/java/org/springaicommunity/typesafe/TypeSafeClient.java#L47
验证标签public-sourcecall-site-verified
GenieRobot/typesafe-ai-rails
README 自述：Community Rails integration for TypeSafe AI's System One API, built on the
证据https://github.com/GenieRobot/typesafe-ai-rails/blob/main/lib/typesafe/rails/client.rb#L55
验证标签call-site-verified
irzix/nestjs-agentic
README 自述：| @nestjs-agentic/jev | Jev (TypeSafe System One) action gates, output rails, and evaluation judges |
证据https://github.com/irzix/nestjs-agentic/blob/main/README.md#L70
验证标签readme-claim
liuhongrui087-art/jev-routed-agent
README 自述：A multi-step reasoning agent built with LangChain v1 + Jev + Flask + Ollama.
证据https://github.com/liuhongrui087-art/jev-routed-agent/blob/master/services/router.py#L22
验证标签call-site-verified
Para-FR/casse-brique-typesafe
README 自述：A brick breaker you can play in the browser, built with Next.js 16, React 19, TypeScript and Tailwind CSS 4. The paddle can be handed over to an AI pilot powered by TypeSafe AI's Jev model.
证据https://github.com/Para-FR/casse-brique-typesafe/blob/main/src/app/api/paddle/route.ts#L108
验证标签call-site-verified
spring-ai-community/spring-ai-typesafe
README 自述：API (jev), built on Spring RestClient and Jackson 3, plus Spring AI integrations that
证据https://github.com/spring-ai-community/spring-ai-typesafe/blob/main/typesafe-java-sdk/src/main/java/org/springaicommunity/typesafe/TypeSafeClient.java#L47
验证标签call-site-verified

Agent 工具

jev-cli
极薄命令行封装：输入 JSON，输出类型化决策，不做额外抽象。
证据https://github.com/lhotwll217/jev-cli/blob/main/src/client.ts#L25
验证标签public-sourcecall-site-verified
pi-typesafe
把 classify、triage、compare、score 等小判断批量交给 Jev，一次调用返回校准过的数值供 agent 使用。
证据https://github.com/DevMortimer/pi-typesafe/blob/main/src/ask.ts#L32
验证标签public-sourcecall-site-verified
agent-reflex
编码 agent 的前置守卫插件，把危险 shell 命令判定交给可替换的决策模型，Jev 是其中之一。
证据https://github.com/psygreg/agent-reflex/blob/main/plugins/hermes/classifier.py#L34
验证标签public-sourcecall-site-verified
APUS-AI-Lab/fast-browser-use
README 自述：Open-Source Reverse Engineering of Jev · Pure Local Browser-Use Agent Skill
证据https://github.com/APUS-AI-Lab/fast-browser-use/blob/main/README.md#L29
验证标签readme-claim
BeLazy167/typesafe-mod
README 自述：A Claude Code function hook that sends two kinds of decision to TypeSafe's Jev
证据https://github.com/BeLazy167/typesafe-mod/blob/main/hooks/suggest.ts#L27
验证标签call-site-verified
dbreunig/building-with-jev-skill
README 自述：This repo holds an agent skill for writing and improving programs that call Jev, TypeSafe's System One model.
证据https://github.com/dbreunig/building-with-jev-skill/blob/main/README.md#L3
验证标签readme-claim
EliaAlberti/jev-rules
README 自述：claude plugin update jev-rules@jev-rules
证据https://github.com/EliaAlberti/jev-rules/blob/main/README.md#L62
验证标签readme-claim
geilt/typesafe-cli
README 自述：Small Python CLI + agent skill for TypeSafe System One (Jev).
证据https://github.com/geilt/typesafe-cli/blob/main/README.md#L3
验证标签readme-claim
getexcited/stepwarden
README 自述：It is built on two early-access dependencies (Claude Code function hooks and Jev),
证据https://github.com/getexcited/stepwarden/blob/main/lib/jev.ts#L30
验证标签call-site-verified
HyunjunJeon/pi-quiet-ask
README 自述：TypeSafe Jev as the pi coding agent's quiet decision layer.
证据https://github.com/HyunjunJeon/pi-quiet-ask/blob/main/bench/src/jev_bench/providers/jev.py#L12
验证标签call-site-verified
iamjonatha/jev-book-tags
README 自述：JEV Book Tags is a calibre plugin that uses TypeSafe AI JEV to suggest subject and genre tags for books in a calibre library. It preserves existing tags and lets you review suggestions before applying them.
证据https://github.com/iamjonatha/jev-book-tags/blob/main/plugin/client.py#L7
验证标签call-site-verified
jbt95/jev-toolkit
README 自述：A small stdio MCP server for TypeSafe/Jev with three tools: typesafe_ask for
证据https://github.com/jbt95/jev-toolkit/blob/main/src/core/paths.ts#L13
验证标签call-site-verified
Jhonnyr97/JevGuard
README 自述：Then run /plugin configure jevguard@jevguard to set the Jev API key and
证据https://github.com/Jhonnyr97/JevGuard/blob/master/README.md#L19
验证标签readme-claim
KamilPostrozny/pi-jev-code
README 自述：Single-agent Jev coprocessor for Pi coding sessions. Your normal Pi model remains the only coding agent; there are no planners, subagents, reviewers, task graphs, or orchestration.
证据https://github.com/KamilPostrozny/pi-jev-code/blob/main/src/index.ts#L6
验证标签call-site-verified
kbhuw/jev-sift
README 自述：Get a key from the TypeSafe dashboard and provide it as JEV_API_KEY in the MCP server's environment. TYPESAFE_API_KEY also works.
证据https://github.com/kbhuw/jev-sift/blob/main/README.md#L11
验证标签readme-claim
King4s/Jev-AI-Skill
README 自述：Jev-AI-Skill is one skill and one MCP server for Claude Code, Codex and Hermes. Jev,
证据https://github.com/King4s/Jev-AI-Skill/blob/main/README.md#L10
验证标签readme-claim
KorenKrita/cliproxy-embeddings-rerank-forward
README 自述：A native plugin for CLIProxyAPI that forwards OpenAI-compatible /v1/embeddings, Cohere-style rerank, and Jev decision requests (OpenRouter Decisions API and TypeSafe System One API) to upstream providers. Supports multip
证据https://github.com/KorenKrita/cliproxy-embeddings-rerank-forward/blob/main/README.md#L3
验证标签readme-claim
leepokai/jev-guard
README 自述：jev-guard key writes ~/.jev-guard/config.json (mode 0600). Every adapter reads that file, so it works for GUI hosts that never see your shell profile. Environment variables win when present: JEV_API_KEY, AI_GATEWAY_API_K
证据https://github.com/leepokai/jev-guard/blob/main/README.md#L73
验证标签readme-claim
leonaaardob/fast-dev-compaction
README 自述：Codex plugin that wraps session compaction with Jev decisions: before Codex
证据https://github.com/leonaaardob/fast-dev-compaction/blob/main/src/request.ts#L3
验证标签call-site-verified
luobosibing2/dsh-jev-plugin
README 自述：Native DeepSeek Harness (DSH) plugin with Jev or Luna Decisions as its judgment model.
证据https://github.com/luobosibing2/dsh-jev-plugin/blob/main/bench/completion_skill/completion_recovery/runner.py#L56
验证标签call-site-verified
mrnugget/jev-shell-history
README 自述：(TypeSafe). As you type, the plugin looks at your last 100 distinct history
证据https://github.com/mrnugget/jev-shell-history/blob/main/src/cli.ts#L5
验证标签call-site-verified
ndolinschi/mcpmatch
README 自述：User goal → match the best MCP server/capability from a catalog (two-stage when large) via TypeSafe Jev.
证据https://github.com/ndolinschi/mcpmatch/blob/main/src/app/api/decide/route.ts#L13
验证标签call-site-verified
psygreg/agent-reflex
README 自述：Plugin for coding agent workflow optimization and safeguarding using any system one decision model, from Jev to self-hosted LM Studio, Ollama or llama.cpp.
证据https://github.com/psygreg/agent-reflex/blob/main/plugins/hermes/classifier.py#L34
验证标签call-site-verified
ratel-ai/ratel
README 自述：ratel_jev_plugin(url=None, api_key_env=None, model=None): Jev (TypeSafe AI). Defaults: https://api.typesafe.ai, TYPESAFE_API_KEY, jev-latest.
证据https://github.com/ratel-ai/ratel/blob/main/src/rankers/python/README.md#L27
验证标签readme-claim
Ravinder82/jev-flash-router
README 自述：Zero-token-output decision router MCP server powered by TypeSafe Jev.
证据https://github.com/Ravinder82/jev-flash-router/blob/main/README.md#L10
验证标签readme-claim
shimo4228/jev-skill-router
README 自述：jev-skill-router is a reference implementation for people building with TypeSafe's Jev, a model that writes no text and answers typed questions with probabilities. It is a Claude Code hook: on every prompt it asks Jev wh
证据https://github.com/shimo4228/jev-skill-router/blob/main/README.md#L15
验证标签readme-claim
tamaratran/fast-jev-compaction
README 自述：Claude Code plugin that replaces the compaction summary with Jev decisions:
证据https://github.com/tamaratran/fast-jev-compaction/blob/main/src/request.ts#L3
验证标签call-site-verified
tamaratran/jev-pruner
README 自述：A Claude Code plugin that uses TypeSafe's Jev to trim noisy Bash output after
证据https://github.com/tamaratran/jev-pruner/blob/main/src/jev.ts#L1
验证标签call-site-verified
thruwire/foreman
README 自述：above slower coding agents. It supports TypeSafe AI's Jev and OpenAI's Decisions API.
证据https://github.com/thruwire/foreman/blob/main/src/foreman/decision_models/jev.py#L47
验证标签call-site-verified
ussyverse/hermes-jev-router
README 自述：An opt-in Hermes plugin that uses TypeSafe Jev to assess task complexity and deterministic policy to select an allowed model within configured cost, token, context, latency-estimate and capability constraints.
证据https://github.com/ussyverse/hermes-jev-router/blob/main/jev.py#L77
验证标签call-site-verified
wobsoriano/oxlint-plugin-jev
README 自述：A rule is a yes/no question about a function, a call, a JSX element, or a whole file. Each match is sent to Jev with the question, and the plugin reports an error when the yes-probability clears your cutoff.
证据https://github.com/wobsoriano/oxlint-plugin-jev/blob/main/src/index.ts#L10
验证标签call-site-verified
xienda/dsh-jev-verify
README 自述：Jev — TypeSafe AI's System One decision model — as a first-class plugin for
证据https://github.com/xienda/dsh-jev-verify/blob/main/bench/batch-e2e.mjs#L28
验证标签call-site-verified

浏览器与计算机操作

AndrewPrifer/jimothy
README 自述：Train small, insanely fast local classifiers from Jev-compatible examples. Run locally in your browser or Node.js.
证据https://github.com/AndrewPrifer/jimothy/blob/master/README.md#L3
验证标签readme-claim
Bodila51/muse-jev-playbook
README 自述：Use TypeSafe AI's Jev as a cheap, fast decision layer inside an AI agent workflow — triage, classify, score, and gate work before expensive steps (browser, deep research, retries, subagents).
证据https://github.com/Bodila51/muse-jev-playbook/blob/main/README.md#L3
验证标签readme-claim
browser-use/jev-ultrafast
README 自述：
证据https://github.com/browser-use/jev-ultrafast/blob/main/jev_ultrafast/model.py#L119
验证标签call-site-verified
droidrun/mobile-jev
README 自述：One goal. A real Android phone. Jev makes the decisions.
证据https://github.com/droidrun/mobile-jev/blob/main/scripts/mobile-agent/policy.mjs#L224
验证标签call-site-verified
ethanplusai/jev-chat-for-twitch
README 自述：Sign up at typesafe.ai and create an API key. The key stays in your browser profile. You pay TypeSafe directly for what you use.
证据https://github.com/ethanplusai/jev-chat-for-twitch/blob/main/README.md#L28
验证标签readme-claim
filedcom/playjev
README 自述：Stagehand, but with Jev. Natural-language browser automation for Playwright, powered by Jev.
证据https://github.com/filedcom/playjev/blob/main/src/core/playjev.ts#L305
验证标签call-site-verified
forvela/jev-agent-browser
README 自述：Give a parent agent a bounded browser task. Jev chooses the next typed action. agent-browser executes it. Ambiguity, repetition, or a stuck state becomes a structured handoff back to the parent.
证据https://github.com/forvela/jev-agent-browser/blob/main/src/decision.js#L1
验证标签call-site-verified
idovmamane/dejevu
README 自述：Jev? Déjà vu. Browser agents that run on instinct, no Jev needed. One look at the page. One call to any open model. One action.
证据https://github.com/idovmamane/dejevu/blob/main/dejevu/policy.py#L209
验证标签call-site-verified
jkudish/jev-browser
README 自述：
证据https://github.com/jkudish/jev-browser/blob/main/src/questions.ts#L4
验证标签call-site-verified
Liyucheng1997/332_lab-jev-chat
README 自述：采集：一个 App 一个适配器（app/src/main/java/com/jev/probe/capture/ChatAppAdapter.kt），服务按前台包名分发；适配器只负责把当前窗口变成「标题 + 消息列表（谁说的、说了什么）」，下游判断 / 悬浮窗 / 填入全部通用。微信适配器读聊天气泡（com.tencent.mm:id/bkl），按气泡位置判断谁说的；QQ 适配器读正文节点（com.tencent.mobileqq:id
证据https://github.com/Liyucheng1997/332_lab-jev-chat/blob/main/README.md#L50
验证标签readme-claim
michaelswissa/jevry
README 自述：The historical beta.1 study ran three five-turn sessions per implementation, with the same fixture tasks and resolved Claude model. Jevry also used Jev for browser decisions.
证据https://github.com/michaelswissa/jevry/blob/main/desktop/engine.ts#L205
验证标签call-site-verified
ndolinschi/swarmrouter
README 自述：Pick which agent/skill handles a task (research / code / browser / support / writer) + confidence — visual swarm map powered by TypeSafe Jev.
证据https://github.com/ndolinschi/swarmrouter/blob/main/src/app/api/decide/route.ts#L7
验证标签call-site-verified
phyous/tsai-civ2
README 自述：An original Civilization II browser harness for TypeSafe Jev, with live decision
证据https://github.com/phyous/tsai-civ2/blob/main/civ2/session.py#L486
验证标签call-site-verified
pst2154/Nemotron_Jev
README 自述：Ask typed questions about text or JSON and inspect model-derived probability distributions in a browser. One container runs the model, the original Decision Lab explorer, and a TypeSafe-shaped API.
证据https://github.com/pst2154/Nemotron_Jev/blob/master/README.md#L3
验证标签readme-claim
rokcso/bluenoise
README 自述：BlueNoise collects no telemetry or analytics. It does not read cookies, access account credentials, call X APIs, or take account actions such as blocking, muting, following, or posting. Settings and rules are stored thro
证据https://github.com/rokcso/bluenoise/blob/main/README.md#L34
验证标签readme-claim
Sac-Y/Jev-cu
README 自述：把 Computer Use 的「下一步点哪里」交给 Jev（TypeSafe System One）：Jev 从界面文字候选中选元素、动作、完成度与风险，Codex Computer Use 负责读取界面与执行，本地策略门槛拦截敏感操作。只传文字，不传截图。
证据https://github.com/Sac-Y/Jev-cu/blob/main/scripts/jev-decide.mjs#L19
验证标签call-site-verified
yikangy873-gif/jev-desktop
README 自述：Jev Desktop adds a bounded decision loop to Codex Computer Use for browser tabs and native macOS apps.
证据https://github.com/yikangy873-gif/jev-desktop/blob/main/plugins/jev-desktop/scripts/bridge-client.mjs#L14
验证标签call-site-verified
Ying-Kai-Liao/jev-browser
README 自述：jev-browser runs locally. It has no server of its own and collects no telemetry. To decide
证据https://github.com/Ying-Kai-Liao/jev-browser/blob/main/README.md#L362
验证标签readme-claim

应用

Jev RAG
在检索后由 Jev 判定段落是否足以支撑答案，再决定是否放行生成。
证据https://github.com/aifabrice/jev-rag
验证标签public-sourcereadme-claim
jev-research-pipeline
把研究流程的每一步收敛判定交给 Jev，避免开放式文本在多步之间漂移。
证据https://github.com/shimo4228/jev-research-pipeline/blob/main/src/jev_research_pipeline/pipeline/drift.py#L21
验证标签public-sourcecall-site-verified
Cribrix
由 Jev 先路由查询、再筛检索段落、最后核验断言，之后才放出答案。
证据https://github.com/david96182/cribrix/blob/main/cribrix/clients/jev.py#L4
验证标签public-sourcecall-site-verified
JEV Book Tags
由 Jev 判定书籍的题材与证据充分度，越过阈值才写入 Calibre 标签。
证据https://github.com/iamjonatha/jev-book-tags/blob/main/plugin/client.py#L7
验证标签public-sourcecall-site-verified
advocaat
通过网关以 typesafe-ai/jev 作为模型标识发请求，把 Jev 当作可插拔的判定后端。
证据https://github.com/pithings/advocaat/blob/main/src/api.ts#L127
验证标签public-sourcecall-site-verified
arimanyus/warrenduffer
README 自述：Dashboard: SSH tunnel or Tailscale to http://127.0.0.1:8080. Set the capital Jev can deploy in the header. Kill = cancel all entries and flatten; Resume clears it.
证据https://github.com/arimanyus/warrenduffer/blob/main/src/context.ts#L42
验证标签call-site-verified
dabit3/jev-experiments
README 自述：sent only to api.typesafe.ai from the native process; it is never written to disk by the app
证据https://github.com/dabit3/jev-experiments/blob/main/jev-ax-pilot/README.md#L153
验证标签readme-claim
DevMortimer/pi-typesafe
README 自述：A tool for the agent. typesafe_evaluate hands small structured judgments (classify, triage, compare, score) to Jev and returns calibrated numbers in one batched call.
证据https://github.com/DevMortimer/pi-typesafe/blob/main/src/ask.ts#L32
验证标签call-site-verified
GeekLinkDev/jev-subtitle-translator
README 自述：An open-source SRT translator and subtitle QC tool for LLM-powered workflows. Use it to translate SRT files, preserve every subtitle cue, and automatically check translations with Jev.
证据https://github.com/GeekLinkDev/jev-subtitle-translator/blob/main/README.md#L7
验证标签readme-claim
iamaamir/system-one
README 自述：Use TypeSafe Jev today, Reflex locally, any compatible /v1/systemone endpoint, or your own SystemOneProvider tomorrow. Your application keeps the same choice, noul, and score code.
证据https://github.com/iamaamir/system-one/blob/main/pi-system-one/src/debug.ts#L7
验证标签call-site-verified
kraayenjon/jev-linkedin-saved-classifier
README 自述：open saved posts -> read every card -> one Jev call each -> live dashboard -> board + csv + json
证据https://github.com/kraayenjon/jev-linkedin-saved-classifier/blob/main/src/jev_linkedin_saved/cli.py#L38
验证标签call-site-verified
mattneel/typesafe
README 自述：and its Jev model. Ask typed Choice, Noul and Score questions about your application's state in
证据https://github.com/mattneel/typesafe/blob/master/lib/typesafe.ex#L252
验证标签call-site-verified
mattneel/typesafe.zig
README 自述：Jev model. Ask typed Noul, Choice and Score questions about your application's state in one
证据https://github.com/mattneel/typesafe.zig/blob/master/src/Client.zig#L353
验证标签call-site-verified
ndolinschi/spendbrake
README 自述：Budget + plan → continue / downgrade_model / stop — cost×value dashboard powered by TypeSafe Jev.
证据https://github.com/ndolinschi/spendbrake/blob/main/src/app/api/decide/route.ts#L7
验证标签call-site-verified
nekuda-ai/WindTunnel
README 自述：tasks are scored by inspecting the application, and published final answers are redacted. The Jev release includes
证据https://github.com/nekuda-ai/WindTunnel/blob/main/experiments/jev/frozen/arms/decision-providers.mjs#L60
验证标签call-site-verified
ReallyArtificial/jev-by-example
README 自述：Ten runnable experiments in building agents that know what to do next—and when to ask. Each combines a small Jev judgment with application logic you can read, change, and test.
证据https://github.com/ReallyArtificial/jev-by-example/blob/main/src/client.mjs#L4
验证标签call-site-verified
ronadin2002/jev-cua
README 自述：A small floating bar for voice and text commands. Jev picks the next action from your Mac’s live controls, the app executes it, and the loop continues toward your request.
证据https://github.com/ronadin2002/jev-cua/blob/main/Sources/Core.swift#L21
验证标签call-site-verified
tostechbr/partway
README 自述：It asks Jev, TypeSafe's System One model, about every word you say. Jev answers with typed decisions and probabilities, not text, in about 200 ms, so the app can act mid-sentence and wait only when it has to.
证据https://github.com/tostechbr/partway/blob/main/README.md#L22
验证标签readme-claim
yusukebe/hono-jev-router
README 自述：app.on('jev', ' ', handler) registers a semantic route.
证据https://github.com/yusukebe/hono-jev-router/blob/main/README.md#L51
验证标签readme-claim
Zefan-Cai/Open-Jev
README 自述：Open-Jev-27B-v1.1 is here. Give your app a decision, with probabilities.
证据https://github.com/Zefan-Cai/Open-Jev/blob/main/README.md#L3
验证标签readme-claim

游戏与模拟

iammusham/jev-snake
README 自述：The project includes a browser-based human baseline and a pygame experiment client. Jev mode uses the same Python engine, replacing only the source of each requested direction.
证据https://github.com/iammusham/jev-snake/blob/main/snake_game/jev_agent.py#L65
验证标签call-site-verified
lukaske/jev-doom-agent
README 自述：The vendored GPL Chocolate Doom fork lives in vendor/chocolate-doom. browser_doom_bridge.c exposes the real engine's health, armor, ammo, damage, coordinates, kills, line-of-sight monsters, and visible pickups to JavaScr
证据https://github.com/lukaske/jev-doom-agent/blob/main/README.md#L18
验证标签readme-claim
valentynkit/jev-plays-pokemon-red
README 自述：ticks and only calls Jev where the game actually branches. Between the bedroom and Oak's
证据https://github.com/valentynkit/jev-plays-pokemon-red/blob/main/src/jpp/policy.py#L20
验证标签call-site-verified
wuyoscar/jev-skill
README 自述：| Demo | Doom & Wikiracing | Watch the official game demos; no standalone repo linked on this page. | Author demo |
证据https://github.com/wuyoscar/jev-skill/blob/main/README.md#L143
验证标签readme-claim

演示与 Playground

Trio-Spark demos
用演示场景展示类型化判定在交互流程里的落点，而非以生成文本作为结果。
证据https://github.com/machinefi/trio-spark
验证标签public-sourcedemo-only
juancamiloqhz/roverlab
README 自述：A browser-based planetary rover sandbox for experimenting with autonomous decisions using TypeSafe AI.
证据https://github.com/juancamiloqhz/roverlab/blob/main/server/decisions.ts#L1
验证标签call-site-verified
sdras/jev-webmcp-extension
README 自述：Inspecting requests in the playground. Select “Open in playground” to load the request's state and questions into the TypeSafe playground.
证据https://github.com/sdras/jev-webmcp-extension/blob/main/src/jev.js#L4
验证标签call-site-verified
wondertwins/jev-benchmark
README 自述：Everything is here: the code, the labeled data, every raw request and response (runs/raw/), the metrics, and an interactive playground. Model served as jev-1.13.0 on 2026-09-16.
证据https://github.com/wondertwins/jev-benchmark/blob/main/jevchess/experiments.py#L22
验证标签call-site-verified

基准与研究

Decision Models as Guardrails
把类型化判定当作护栏而非评分器，用固定选项约束下游可执行的动作集合。
证据https://github.com/raxITlabs/decision-models-as-guardrails/blob/main/infra/gcp/laya_server.py#L118
验证标签public-sourceauthor-reported-metriccall-site-verified
anandi1989/awesome-jev-usecases
README 自述：ts2.tech: "TypeSafe AI Raises $40 Million for Jev, but Its 445× Cost Claim Is Still Self-Tested": *The skeptical audit of self-tested benchmarks, plus funding verification.*
证据https://github.com/anandi1989/awesome-jev-usecases/blob/main/README.md#L234
验证标签readme-claim
caiovicentino/eikos
README 自述：evaluation/eval_jev_suite.py and evaluation/eval_laya_suite.py run the same suites through Jev (API, evaluation
证据https://github.com/caiovicentino/eikos/blob/main/eikos/serve.py#L343
验证标签call-site-verified
codeitlikemiley/system-one-adapter-rust
README 自述：A drop-in replacement for TypeSafe's system_one evaluation API, backed by LLM
证据https://github.com/codeitlikemiley/system-one-adapter-rust/blob/main/src/client.rs#L15
验证标签call-site-verified
erboland/jev-fund
README 自述：jev-fund: Open-Source Paper Hedge Fund
证据https://github.com/erboland/jev-fund/blob/main/README.md#L32
验证标签readme-claim
frankda/jev-poly-crypto-demo
README 自述：This version implements a live market-data adapter, Jev integration, paper trading and a dashboard. It does not connect a wallet or submit real orders. TRADING_MODE=live is rejected at startup.
证据https://github.com/frankda/jev-poly-crypto-demo/blob/main/scripts/check-jev.ts#L2
验证标签call-site-verified
fstandhartinger/jevbench
README 自述：Jev is TypeSafe AI's decision model; JevBench also evaluates Jev-compatible systems. The canonical JevBench board is https://benchmarkheaven.com/jev-models.
证据https://github.com/fstandhartinger/jevbench/blob/main/README.md#L5
验证标签readme-claim
grmkris/robo-harness
README 自述：Decision runs (bun run jev) let an evaluation model or rules baseline pick bounded joint steps; chat and a decision run never move the arm at the same time.
证据https://github.com/grmkris/robo-harness/blob/main/apps/server/src/decision/strategies.ts#L273
验证标签call-site-verified
kylemclaren/jevpdf
README 自述：Cheap enough to ask about every line. Jev costs $0.042 per million input tokens, and output is free. Checking every line of a 15-page paper costs well under a cent.
证据https://github.com/kylemclaren/jevpdf/blob/main/server/jev-upstream.ts#L5
验证标签call-site-verified
micytao/vllm-playground
README 自述：Decision Models (Experimental) — A brand-new sandbox tab for structured-read decisions powered by DiffusionGemma and Jev / System One. Gallery of 12 use cases across 4 tabbed categories (Noul, Choice, Score, Combination)
证据https://github.com/micytao/vllm-playground/blob/main/README.md#L37
验证标签readme-claim
nexibeo/jev-cookbook
README 自述：Practical, tested recipes for TypeSafe's Jev, the fast decision model on OpenRouter. Fifteen real-world jobs, each with a runnable script, a small labelled dataset and measured results: support triage, database indexing,
证据https://github.com/nexibeo/jev-cookbook/blob/main/README.md#L5
验证标签readme-claim
nickthompson480/typesafe-ai-playground
README 自述：You need Python 3.10 or newer, a modern browser, and a TypeSafe API key to run evaluations. There are no packages to install and no frontend build step.
证据https://github.com/nickthompson480/typesafe-ai-playground/blob/main/server.py#L17
验证标签call-site-verified
pedronauck/skills
README 自述：jev-engineering — Design and build use cases, POCs and production systems with TypeSafe's Jev decision model (System One): design modes and use-case archetypes, decision state, typed questions, Jev combined with generati
证据https://github.com/pedronauck/skills/blob/main/README.md#L80
验证标签readme-claim
theSekyi/jevusecases
README 自述：A directory of real projects, filtered by what they replace, whether they've been benchmarked, and whether there's code to copy — not a marketing site for Jev itself.
证据https://github.com/theSekyi/jevusecases/blob/main/README.md#L5
验证标签readme-claim
unicodeveloper/jevocks
README 自述：Jevinik uses Jev, TypeSafe AI's System One evaluation model, to make typed, probabilistic decisions without calling a general-purpose generative LLM.
证据https://github.com/unicodeveloper/jevocks/blob/main/README.md#L11
验证标签readme-claim
willkelly/jev-evaluation
README 自述：An adversarial evaluation of jev, the decision model sold by TypeSafe. It
证据https://github.com/willkelly/jev-evaluation/blob/main/jeveval/config.py#L18
验证标签call-site-verified

其它清单

0x7067/claude-jev
README 自述：The run sends your past prompts to api.typesafe.ai. Start with --sample 250
证据https://github.com/0x7067/claude-jev/blob/main/eval/README.md#L130
验证标签readme-claim
0xArx/jevegis
README 自述：retrieved documents) and Trust & Safety moderation, both powered by TypeSafe's Jev
证据https://github.com/0xArx/jevegis/blob/main/src/lib/engine.ts#L1
验证标签call-site-verified
0xBakeer/arbiter
README 自述：Serve typed-decision models — Laya or your own — on your GPU or your Mac, with a Jev-compatible
证据https://github.com/0xBakeer/arbiter/blob/main/integrations/claude-code/hooks/eval.py#L57
验证标签call-site-verified
0xNatoshi/jev-codex-router
README 自述：Jev chooses a model and thinking effort together for each model call, including
证据https://github.com/0xNatoshi/jev-codex-router/blob/main/poc/backtest_savings.py#L157
验证标签call-site-verified
0xtrou/rubikjev
README 自述：You scramble like a menace. Jev — the in-house AI solver — judges the chaos, roasts you, and cooks the solve live.
证据https://github.com/0xtrou/rubikjev/blob/main/src/app/api/solve/route.ts#L127
验证标签call-site-verified
2FastLabs/agent-squad
README 自述：⚡ Fast and cheap — sub-second decisions billed on input tokens only (output tokens free, see pricing), versus a full LLM call per routing decision.
证据https://github.com/2FastLabs/agent-squad/blob/main/README.md#L157
验证标签readme-claim
3clyp50/a0-typesafe-ai
README 自述：Ask Jev focused questions and get typed answers with probability cards in Agent Zero's chat. One typesafe_query call can combine Choice, Noul, Score, and Box questions over the same text and images.
证据https://github.com/3clyp50/a0-typesafe-ai/blob/main/extensions/webui/get_tool_message_handler/typesafe.js#L2
验证标签call-site-verified
6Mikao9/jev-agent-design-with-topk-logits-choices
README 自述：Jev is the first backend we use for live experiments; the runtime API is
证据https://github.com/6Mikao9/jev-agent-design-with-topk-logits-choices/blob/main/jev_agent/jev_client.py#L57
验证标签call-site-verified
Aayan-DEV/aayans-yc-indexor
README 自述：Put a TypeSafe key in .env.local as TYPE_SAFE_KEY. Get one at typesafe.ai.
证据https://github.com/Aayan-DEV/aayans-yc-indexor/blob/main/README.md#L30
验证标签readme-claim
AbdelStark/bicameral
README 自述：Sent to TypeSafe (POST /v1/systemone via TypeSafeClient.systemOne): redacted, truncated JSON state for the current reflex (goal, proposed tool, compact diffs, recent action summaries) and the pack's questions.
证据https://github.com/AbdelStark/bicameral/blob/main/packages/s1-runtime/src/backends/typesafe.ts#L1
验证标签call-site-verified
abe75ch/tiltmeter
README 自述：Notices your Jev decisions tilting before anything visibly breaks.
证据https://github.com/abe75ch/tiltmeter/blob/main/tiltmeter.py#L270
验证标签call-site-verified
AboveColin/HA-Jev
README 自述：back. Jev is a decision model, not a chat model. It answers a typed question with a
证据https://github.com/AboveColin/HA-Jev/blob/main/README.md#L12
验证标签readme-claim
ad-si/cai
README 自述：[TypeSafe]: https://docs.typesafe.ai/models
证据https://github.com/ad-si/cai/blob/main/readme.md#L50
验证标签readme-claim
adarshmishra07/jcm-router
README 自述：level per message, using TypeSafe's Jev classifier. Trivial questions go to Haiku, everyday work to
证据https://github.com/adarshmishra07/jcm-router/blob/main/scripts/eval.ts#L12
验证标签call-site-verified
adhyaay-karnwal/jev-chat
README 自述：A research decoder that treats Jev (a System One decision model) as if it were a language model.
证据https://github.com/adhyaay-karnwal/jev-chat/blob/main/src/jevchat/client.py#L7
验证标签call-site-verified
affirmitv/ghosthands
README 自述：Configuration: GH_PLANNER=jev, GH_JEV_MODEL (default typesafe/jev-1.13; ~typesafe/jev-latest
证据https://github.com/affirmitv/ghosthands/blob/main/README.md#L221
验证标签readme-claim
afshinm/laya-mps
README 自述：Run Jev-style typed decisions locally on your Mac with low RAM usage and fast responses.
证据https://github.com/afshinm/laya-mps/blob/main/README.md#L3
验证标签readme-claim
Ai-trainee/awesome-jev
README 自述：prism-liquidity-agent — ★69 · TypeScript · Jev 评分 18 · 工具与产品 — Solana 流动性代理：可选的 Jev 影子判定用于存款选择、有毒流与市场状态压力。
证据https://github.com/Ai-trainee/awesome-jev/blob/main/README.md#L390
验证标签readme-claim
ai4curation/jevotron
README 自述：in tabular, structured, and text files, SQLite, and DuckDB, powered by Jev.
证据https://github.com/ai4curation/jevotron/blob/main/README.md#L8
验证标签readme-claim
AiPersonacademy/jev-resume-disqualifier
README 自述：Automated Candidate Knockout Engine for High-Volume Recruiting — Powered by TypeSafe Jev System One Decision Intelligence.
证据https://github.com/AiPersonacademy/jev-resume-disqualifier/blob/main/engine/decision_gateway.py#L20
验证标签call-site-verified
alexj11324/open-jev-approvals
README 自述：When the agent is about to run a tool, a hook intercepts the call, TypeSafe
证据https://github.com/alexj11324/open-jev-approvals/blob/main/internal/jev/client.go#L22
验证标签call-site-verified
allebee/pytest-jev
README 自述：All the claims about one text go to Jev in a single request. When a prompt change breaks the reply,
证据https://github.com/allebee/pytest-jev/blob/main/bench/llm_judge.py#L14
验证标签call-site-verified
allenporter/home-assistant-typesafe
README 自述：Queries the TypeSafe System One API (jev-latest) in a single network roundtrip, obtaining calibrated probabilities across all questions without hallucination or token streaming latency.
证据https://github.com/allenporter/home-assistant-typesafe/blob/main/README.md#L34
验证标签readme-claim
alterhq/typesafe-sdk-swift
README 自述：authenticated backend. Keep the TypeSafe API key in server-side secret storage. Direct API-key
证据https://github.com/alterhq/typesafe-sdk-swift/blob/main/Sources/TypeSafe/Requests.swift#L4
验证标签call-site-verified
altryne/jevify
README 自述：Use Jev during everyday agent work, and find what it can improve in your product.
证据https://github.com/altryne/jevify/blob/main/scripts/jev_client.py#L32
验证标签call-site-verified
andrelandgraf/safer-with-jev
README 自述：Public showcases of TypeSafe Jev judgments. TypeSafe Jev inspects the body, then optionally forwards the same bytes to a caller-chosen HTTPS URL.
证据https://github.com/andrelandgraf/safer-with-jev/blob/main/src/index.ts#L5
验证标签call-site-verified
andududu/jeview
README 自述：your code → Jeview (127.0.0.1:4777) → TypeSafe (api.typesafe.ai)
证据https://github.com/andududu/jeview/blob/main/README.md#L9
验证标签readme-claim
Anil-matcha/awesome-jev-by-typesafe
README 自述：▶ Watch: Jev: 8 Wild Things People Already Built With TypeSafe's New AI Model
证据https://github.com/Anil-matcha/awesome-jev-by-typesafe/blob/main/README.md#L14
验证标签readme-claim
aniruddh-krovvidi/switchboard
README 自述：A guardrail and model router for LLM gateways, built on Jev, TypeSafe AI's
证据https://github.com/aniruddh-krovvidi/switchboard/blob/main/jev.py#L3
验证标签call-site-verified
anishfn/shapeshift
README 自述：
证据https://github.com/anishfn/shapeshift/blob/main/src/lib/jev/client.ts#L2
验证标签call-site-verified
AnotiaWang/awesome-jev
README 自述：Open-source products and demos that put a decision model in a real loop. Most use Jev today.
证据https://github.com/AnotiaWang/awesome-jev/blob/main/README.md#L178
验证标签readme-claim
AntonioCoppe/jev-harness
README 自述：It is not affiliated with TypeSafe. You need a TypeSafe API key to call Jev.
证据https://github.com/AntonioCoppe/jev-harness/blob/main/recipes/agent-comm-harness/message-route.ts#L1
验证标签call-site-verified
AppitStudio/awesome-jev
README 自述：Discover Jev-powered apps, developer resources, and runnable decision examples.
证据https://github.com/AppitStudio/awesome-jev/blob/main/README.md#L3
验证标签readme-claim
AppSprout-dev/Unsga3
README 自述：TypeSafe / Jev Pareto scoring (optional)
证据https://github.com/AppSprout-dev/Unsga3/blob/main/README.md#L104
验证标签readme-claim
ARCJ137442/jev-life
README 自述：Alongside it sits a Jev-compatible broker that wraps any OpenAI- or Anthropic-compatible LLM into a Jev backend,
证据https://github.com/ARCJ137442/jev-life/blob/main/src/client/api.ts#L222
验证标签call-site-verified
arczhi/jet
README 自述：Judgment-first routing — Jev (TypeSafe System One) handles context
证据https://github.com/arczhi/jet/blob/main/README.md#L43
验证标签readme-claim
atharvamhaske/typesafe-sdk-go
README 自述：| WithBaseURL | TYPESAFE_BASE_URL | API base URL, defaults to https://api.typesafe.ai |
证据https://github.com/atharvamhaske/typesafe-sdk-go/blob/main/README.md#L115
验证标签readme-claim
audreyt/paseo-spacedock
README 自述：Gate judgment (Jev / TypeSafe-compatible)
证据https://github.com/audreyt/paseo-spacedock/blob/main/README.md#L56
验证标签readme-claim
automateyournetwork/netclaw
README 自述：| JEV_BASE_URL | https://api.typesafe.ai; /v1/systemone appended when no path supplied; full custom paths accepted |
证据https://github.com/automateyournetwork/netclaw/blob/main/mcp-servers/jev-mcp/README.md#L42
验证标签readme-claim
Avinash-jetwani/jevmem
README 自述：To TypeSafe AI, https://api.typesafe.ai/v1/systemone, or the URL in TYPESAFE_BASE_URL if you set it:
证据https://github.com/Avinash-jetwani/jevmem/blob/main/plugin/README.md#L46
验证标签readme-claim
awlevin/typesafe-computer-use
README 自述：
证据https://github.com/awlevin/typesafe-computer-use/blob/main/typesafe_computer_use/actions.py#L9
验证标签call-site-verified
awslabs/agentcore-samples
README 自述：Jev Evaluators (--with-jev)
证据https://github.com/awslabs/agentcore-samples/blob/main/01-features/06-observe-evaluate-optimize-your-agent/02-evaluate/custom-code-based-evaluation/README.md#L600
验证标签readme-claim
Ayush0054/metis
README 自述：metis-workflow is a small Python framework for workflows powered by TypeSafe AI's
证据https://github.com/Ayush0054/metis/blob/main/legacy/metis-triage/src/metis_triage/_triage.py#L154
验证标签call-site-verified
Baba88611/detroit-ai-player
README 自述：Select the dedicated TypeSafe backend with --model jev. Append these variables to your local 03_runner/.env, preserving existing configuration. Keep the real key local:
证据https://github.com/Baba88611/detroit-ai-player/blob/main/README.md#L189
验证标签readme-claim
bcharleson/jev-gtm-cookbook
README 自述：Open-source go-to-market recipes built on TypeSafe's Jev: code does the work, Jev makes the judgment calls, and each recipe runs locally on data you already own.*
证据https://github.com/bcharleson/jev-gtm-cookbook/blob/main/README.md#L5
验证标签readme-claim
beamnxw/minelog
README 自述：Jev (TypeSafe System One) picks one bounded action from a closed catalog, plus a safety noul and an urgency score. Fast. Typed. No prose.
证据https://github.com/beamnxw/minelog/blob/main/README.md#L33
验证标签readme-claim
BeatAPI/awesome-jev
README 自述：
证据https://github.com/BeatAPI/awesome-jev/blob/main/scripts/curate-star50-expansion-2026-09-23.mjs#L217
验证标签call-site-verified
BenCaunt/jev-box-tracking
README 自述：jev-box-tracking
证据https://github.com/BenCaunt/jev-box-tracking/blob/main/README.md#L1
验证标签readme-claim
bestagentkits/jev-skillful
README 自述：The prompt is sent to api.typesafe.ai as part of the routing request. This is the only kind of
证据https://github.com/bestagentkits/jev-skillful/blob/main/packages/cli/src/core/eval/replay.ts#L22
验证标签call-site-verified
Bewinxed/jevgpt
README 自述：TypeSafe AI's jev-latest is a decision model: it does not generate text,
证据https://github.com/Bewinxed/jevgpt/blob/master/src/jevgpt/sampler.py#L29
验证标签call-site-verified
bhaiG-de/jev-design-test
README 自述：Prerequisites: Node 20+, and a Jev (TypeSafe) API key.
证据https://github.com/bhaiG-de/jev-design-test/blob/master/vite.config.ts#L5
验证标签call-site-verified
BillionsBobby/JevRouter
README 自述：
证据https://github.com/BillionsBobby/JevRouter/blob/main/src/provider.ts#L103
验证标签call-site-verified
bladedevoff/stuntd
README 自述：With the paid Jev API in front, the same demo runs with it as the teacher instead: put
证据https://github.com/bladedevoff/stuntd/blob/main/examples/banking/README.md#L42
验证标签readme-claim
bodepudimuneendra-netizen/laya-jev-GraphRAG
README 自述：│ LAYER 2 — DECISION (Laya / Jev) ← The CPU │
证据https://github.com/bodepudimuneendra-netizen/laya-jev-GraphRAG/blob/main/graphrag_neo4j_laya/graphrag/models/jev.py#L7
验证标签call-site-verified
bohutang/sift
README 自述：Search results stay native: no Jev scoring, local filters, labels, or collapsed posts on /search. The normal feed and post pages still use your filtering settings.
证据https://github.com/bohutang/sift/blob/main/background.js#L4
验证标签call-site-verified
brainstormity/Jev-X-Sentiment-Analysis
README 自述：An on-demand crypto market intelligence and decision-support terminal powered by TypeSafe AI's System One model (Jev).
证据https://github.com/brainstormity/Jev-X-Sentiment-Analysis/blob/main/app/services/typesafe_service.py#L3
验证标签call-site-verified
brandonbryant12/transcript-scorecard
README 自述：Server-side OpenAI Decisions and TypeSafe JEV adapters for incremental transcript scoring
证据https://github.com/brandonbryant12/transcript-scorecard/blob/main/apps/api/src/classifier.ts#L1
验证标签call-site-verified
brnyxx/jev-ra
README 自述：
证据https://github.com/brnyxx/jev-ra/blob/main/jev_ra/config.py#L11
验证标签call-site-verified
buberlo/jev-trader
README 自述：jev-trader
证据https://github.com/buberlo/jev-trader/blob/main/README.md#L1
验证标签readme-claim
bugkiwi/elons-job
README 自述：TypeSafe 请求使用 https://api.typesafe.ai/v1/systemone；API 返回不符合预期、超时、401、429 或 5xx 时 Fail Open。
证据https://github.com/bugkiwi/elons-job/blob/main/README.md#L46
验证标签readme-claim
BunsDev/typesafe-ai-playground
README 自述：Small experiments. Clear decisions. This is an independent community project, not an official TypeSafe AI product, production agent harness, or claim that a typed answer is correct.
证据https://github.com/BunsDev/typesafe-ai-playground/blob/main/lib/serverJev.ts#L52
验证标签call-site-verified
BunsDev/typesafe-ui
README 自述：Reusable React components and interactive interface patterns for TypeSafe AI projects, built with shadcn/ui, Base UI, and Tailwind CSS.
证据https://github.com/BunsDev/typesafe-ui/blob/main/README.md#L3
验证标签readme-claim
burnigtm/jev-mcp
README 自述：codex mcp add jev --env TYPESAFE_API_KEY=ts_... -- node /absolute/path/to/jev-mcp/dist/index.js
证据https://github.com/burnigtm/jev-mcp/blob/main/src/errors.ts#L1
验证标签call-site-verified
BYK/jev-mcp
README 自述：Most Jev integrations expose "ask the model a question". The hard part in practice is not asking —
证据https://github.com/BYK/jev-mcp/blob/main/src/tools/ask.ts#L37
验证标签call-site-verified
bytelabs-oss/clash-jev
README 自述：play around with it. I built clash-jev, a bot that uses Jev to make near real-time decisions on live
证据https://github.com/bytelabs-oss/clash-jev/blob/main/clash_jev/policy.py#L183
验证标签call-site-verified
caiovicentino/eikos-arena
README 自述：MIT) against Jev (TypeSafe's decision API). Every 5 minutes both get the same market snapshot and the same 28
证据https://github.com/caiovicentino/eikos-arena/blob/main/arena/arena.py#L31
验证标签call-site-verified
can1357/jegrep
README 自述：src/jev.rs 0.97 · 458 lines shown
证据https://github.com/can1357/jegrep/blob/main/README.md#L122
验证标签readme-claim
CaptainCore/captaincore
README 自述：Optional integrations -- fathom_api_key (analytics) and typesafe_api_key (the TypeSafe decision API behind captaincore typesafe and scan --triage; TYPESAFE_API_KEY in the environment works too)
证据https://github.com/CaptainCore/captaincore/blob/master/typesafe/client.go#L188
验证标签call-site-verified
carldaws/hunch
README 自述：Point Hunch at a System One endpoint. Jev is available through
证据https://github.com/carldaws/hunch/blob/main/README.md#L27
验证标签readme-claim
carlosedm10/agi-jev-containment
README 自述：What is Jev in this project? Jev is TypeSafe System One (jev-1.13.0): a semantic classifier that answers typed questions (criticality L0–L5, intent, concealment, …) over short-term tape + long-term key nodes. It does not
证据https://github.com/carlosedm10/agi-jev-containment/blob/main/README.md#L70
验证标签readme-claim
cbruyndoncx/AskJev-MCP
README 自述：Ask everything you want to know about a state in one call rather than several — TypeSafe bills per
证据https://github.com/cbruyndoncx/AskJev-MCP/blob/main/src/index.ts#L6
验证标签call-site-verified
cequence-io/openai-scala-client
README 自述：Decision models give typed answers with calibrated probabilities. Two APIs serve them - TypeSafe AI's System One
证据https://github.com/cequence-io/openai-scala-client/blob/master/typesafe-client/src/main/scala/io/cequence/openaiscala/typesafe/JsonFormats.scala#L125
验证标签call-site-verified
Chandler-Sun/chat2jev
README 自述：Convert OpenAI-compatible Chat Completions requests into TypeSafe System One (Jev) State / Questions, compare generated text with structured judgments, and publish reusable question sets as proxy routes.
证据https://github.com/Chandler-Sun/chat2jev/blob/main/src/components/answers.tsx#L49
验证标签call-site-verified
CharlesFeng0314/JEV_sees
README 自述：JEV is fast, structured, and built to make decisions.
证据https://github.com/CharlesFeng0314/JEV_sees/blob/main/jev_sees/__init__.py#L3
验证标签call-site-verified
Charlyhno-eng/jev-document-classification
README 自述：To use JEV, you need a Vercel account and an AI Gateway API key.
证据https://github.com/Charlyhno-eng/jev-document-classification/blob/main/README.md#L25
验证标签readme-claim
cheeaun/jevmoji
README 自述：Type up to 40 characters. Get related emojis scored by TypeSafe’s Jev — many matches, not just one.
证据https://github.com/cheeaun/jevmoji/blob/main/README.md#L10
验证标签readme-claim
CheshiAI/Cheshi
README 自述：Previous conversation recall is off by default, including for existing TypeSafe
证据https://github.com/CheshiAI/Cheshi/blob/main/README.md#L86
验证标签readme-claim
chez-shanpu/typesafeai-go
README 自述：// A TypeSafe AI compatible server running locally without authentication.
证据https://github.com/chez-shanpu/typesafeai-go/blob/main/option.go#L9
验证标签call-site-verified
christos-antono/NPCTalk
README 自述：| Actions and spoken orders | TypeSafe | TYPESAFE_API_KEY |
证据https://github.com/christos-antono/NPCTalk/blob/main/README.md#L57
验证标签readme-claim
chy4pro/jev-for-chrome
README 自述：| OpenRouter | https://openrouter.ai/api/alpha/decisions | typesafe/jev-1.13 |
证据https://github.com/chy4pro/jev-for-chrome/blob/main/README.md#L112
验证标签readme-claim
cobanov/awesome-jev
README 自述：New: Ollaya runs open decision models locally behind a Jev-compatible API.
证据https://github.com/cobanov/awesome-jev/blob/main/README.md#L8
验证标签readme-claim
cobusgreyling/Jev
README 自述：Jev Showcase — TypeSafe System One, not chat
证据https://github.com/cobusgreyling/Jev/blob/main/README.md#L1
验证标签readme-claim
cocktailpeanut/jevthoven
README 自述：There is no offline composer, audio model, or hidden song library. Every note traces to a persisted Jev decision (or a labeled fixture/manual edit).
证据https://github.com/cocktailpeanut/jevthoven/blob/main/app/scripts/live-smoke.mjs#L15
验证标签call-site-verified
codaaiteam/jev-mcp
README 自述：Official TypeSafe: get a key at console.typesafe.ai. It's used against api.typesafe.ai by default.
证据https://github.com/codaaiteam/jev-mcp/blob/main/README.md#L30
验证标签readme-claim
Cognipeer/agent-sdk
README 自述：Against the real Jev API
证据https://github.com/Cognipeer/agent-sdk/blob/main/examples/jev-guardrail/README.md#L24
验证标签readme-claim
colliber/duckdb-jev
README 自述：the extension talking to api.typesafe.ai.
证据https://github.com/colliber/duckdb-jev/blob/main/docs/README.md#L4
验证标签readme-claim
columnar-tech/jevaro
README 自述：A Python batching proxy for Jev that returns
证据https://github.com/columnar-tech/jevaro/blob/main/README.md#L3
验证标签readme-claim
compozy/yoshi
README 自述：Jev judges new context once, above a configurable size gate. Yoshi applies the
证据https://github.com/compozy/yoshi/blob/main/README.md#L7
验证标签readme-claim
Contrastive-LM/CLM
README 自述：It performs on par with Jev across computer-use, gaming and tool-calling
证据https://github.com/Contrastive-LM/CLM/blob/main/src/clm/__init__.py#L6
验证标签call-site-verified
CorieW/JevTest
README 自述：To run the same suite using Jev, copy .env.example to .env.local, set TYPESAFE_API_KEY, then run:
证据https://github.com/CorieW/JevTest/blob/main/src/jev.ts#L122
验证标签call-site-verified
CTNicholas/jev-workflow-builder
README 自述：Create an account on typesafe.ai
证据https://github.com/CTNicholas/jev-workflow-builder/blob/main/README.md#L14
验证标签readme-claim
danielgshea/jev-as-a-judge
README 自述：This project tests a third option: Jev, TypeSafe AI's System One decision model. We compare Jev with GPT-5.6 Luna, GPT-5.6 Terra, and Claude Sonnet 4.6 on the same fixed agent runs, measuring binary accuracy, score relia
证据https://github.com/danielgshea/jev-as-a-judge/blob/main/README.md#L5
验证标签readme-claim
DanRWilloughby/snifftest
README 自述：judgment rules too, put a TypeSafe key in TYPESAFE_API_KEY and drop
证据https://github.com/DanRWilloughby/snifftest/blob/main/src/jev.ts#L50
验证标签call-site-verified
danvega/jev-spring-boot-starter
README 自述：| jev.api-key | TYPESAFE_API_KEY fallback | TypeSafe API key |
证据https://github.com/danvega/jev-spring-boot-starter/blob/main/README.md#L145
验证标签readme-claim
Das-rebel/a3m-router
README 自述：model="jev-auto" — ML decision head, calibrated probabilities, ~2ms warm
证据https://github.com/Das-rebel/a3m-router/blob/main/README.md#L120
验证标签readme-claim
daseinlabs/open-jev
README 自述：make systemone # TypeSafe-style request against /v1/systemone
证据https://github.com/daseinlabs/open-jev/blob/main/demo/doom/client.py#L66
验证标签call-site-verified
davila7/jev-explained
README 自述：Learn how TypeSafe's Jev makes typed, probabilistic decisions — by running it.
证据https://github.com/davila7/jev-explained/blob/main/src/lib/providers.ts#L21
验证标签call-site-verified
DECRUX9812/typesafe-skill-router
README 自述：A TypeSafe API key (TYPESAFE_API_KEY) — create one at
证据https://github.com/DECRUX9812/typesafe-skill-router/blob/main/__init__.py#L30
验证标签call-site-verified
delexw/claude-code-trace
README 自述：Jev and provider API keys live in the OS credential store (macOS Keychain, Windows Credential
证据https://github.com/delexw/claude-code-trace/blob/main/src-tauri/src/efficiency/jev.rs#L12
验证标签call-site-verified
dengyie/ai-register-machine
README 自述：| typesafe.ai / jev | ./register.sh typesafe [count] [threads] | Python + requests + Stytch Magic-Link | typesafe-*.json (0600 API Key) |
证据https://github.com/dengyie/ai-register-machine/blob/main/README.md#L36
验证标签readme-claim
devagrawal09/jev-review
README 自述：The reviewer keeps orchestration in code and uses Jev for bounded judgments:
证据https://github.com/devagrawal09/jev-review/blob/main/src/review/codebase-judgments.ts#L3
验证标签call-site-verified
devagrawal09/stanley-code
README 自述：For built-ins, Jev answers fixed-choice questions about each piece. Using your required TypeSafe API key,
证据https://github.com/devagrawal09/stanley-code/blob/main/README.md#L94
验证标签readme-claim
Devin-AXIS/jev-dsh-decision
README 自述：问题类型是 choice（单选）、score（等级评分）、noul（是的概率）。多个互补 Skill 可使用多个独立 noul 问题。传入最小必要上下文，不传密钥或无关私密资料。网络固定发往 https://api.typesafe.ai/v1/systemone；默认模型 jev-latest。
证据https://github.com/Devin-AXIS/jev-dsh-decision/blob/main/README.md#L137
验证标签readme-claim
DevMortimer/pi-warden
README 自述：Jev requests cost money, so each 1.0 change was measured by what it saves. Each figure is the measured change on recorded sessions of one machine, not a promise for yours.
证据https://github.com/DevMortimer/pi-warden/blob/main/scripts/slop-cases.mjs#L77
验证标签call-site-verified
devtooligan/jevscan-evm
README 自述：Produce a heat map of likely bugs - in seconds, for pennies - built with TypeSafe's Jev.
证据https://github.com/devtooligan/jevscan-evm/blob/main/README.md#L10
验证标签readme-claim
dfinke/Jev
README 自述：
证据https://github.com/dfinke/Jev/blob/main/README.md#L2
验证标签readme-claim
Dicklesworthstone/skillranker
README 自述：The right skill for the next step, powered by Jev-compatible providers.
证据https://github.com/Dicklesworthstone/skillranker/blob/main/src/jev/client/tests.rs#L100
验证标签call-site-verified
Dimweaker/jev-libero
README 自述：
证据https://github.com/Dimweaker/jev-libero/blob/main/src/jev_libero/client.py#L14
验证标签call-site-verified
docxology/daf-jev
README 自述：Modular, composable Python client and decision toolkit for the TypeSafe Jev
证据https://github.com/docxology/daf-jev/blob/main/src/daf_jev/__init__.py#L23
验证标签call-site-verified
doeixd/discern
README 自述：error. Provider-neutral: runs on TypeSafe's Jev, or any other DecisionModel.
证据https://github.com/doeixd/discern/blob/main/README.md#L7
验证标签readme-claim
dominicrico/jev-router
README 自述：| jev (default) | https://api.typesafe.ai/v1/systemone | jev-latest | TYPESAFE_API_KEY or /jev key (required) |
证据https://github.com/dominicrico/jev-router/blob/main/README.md#L94
验证标签readme-claim
donvito/ai-backends
README 自述：| TypeSafe Jev | System One decision model: choice, score, and noul questions | Available |
证据https://github.com/donvito/ai-backends/blob/main/README.md#L99
验证标签readme-claim
dorkitude/webctl
README 自述：Smart web search CLI for agents, backed by Jev. Saves a lot of tokens.
证据https://github.com/dorkitude/webctl/blob/main/README.md#L3
验证标签readme-claim
Dujaydis/JevSysUno
README 自述：A reference integration for TypeSafe AI's Jev, the first "System One" model: typed
证据https://github.com/Dujaydis/JevSysUno/blob/master/README.md#L3
验证标签readme-claim
ellistev/typesafe-minecraft-demo
README 自述：An experimental Minecraft Java bot controlled through TypeSafe choices, with a live world view and the actual API input and output beside it.
证据https://github.com/ellistev/typesafe-minecraft-demo/blob/main/src/decisions.cjs#L45
验证标签call-site-verified
equationalapplications/system-one-reviewer
README 自述：| jev | --provider jev (default; SOR_PROVIDER env also works) | hosted TypeSafe System One, TYPESAFE_API_KEY | live-tested (all published numbers) |
证据https://github.com/equationalapplications/system-one-reviewer/blob/main/README.md#L236
验证标签readme-claim
eriestra/browser-use-olympics
README 自述：echo 'TYPESAFE_API_KEY=...' > ~/.config/typesafe/env # chmod 600
证据https://github.com/eriestra/browser-use-olympics/blob/main/almond-fastloop.mjs#L102
验证标签call-site-verified
Eriskii/ErisLint
README 自述：anything useful, then map Jev's answers to warnings or errors.
证据https://github.com/Eriskii/ErisLint/blob/main/src/jev.rs#L24
验证标签call-site-verified
EthanAlgoX/AIStock
README 自述：Try JEV decisions (optional): configure its separate TypeSafe API key in Settings → Models & runtime, then select JEV · Decisions only for a new trading strategy. JEV returns buy/sell/hold, probabilities, and confidence
证据https://github.com/EthanAlgoX/AIStock/blob/main/README.md#L148
验证标签readme-claim
etweisberg/jev-ui
README 自述：
证据https://github.com/etweisberg/jev-ui/blob/main/apps/demo/scripts/check-key.mjs#L5
验证标签call-site-verified
EugeneBoondock/jevsql
README 自述：
证据https://github.com/EugeneBoondock/jevsql/blob/main/src/client.mjs#L28
验证标签call-site-verified
evoke-build/evoke
README 自述：To decide against Jev, export TYPESAFE_API_KEY for the jev adapter, or OPENJEV_API_KEY for openjev. To
证据https://github.com/evoke-build/evoke/blob/main/crates/evoke-adapters/src/systemone.rs#L85
验证标签call-site-verified
f-tiger/jev-decision-gate
README 自述：Use Jev for Issue triage without building the whole integration and review workflow.
证据https://github.com/f-tiger/jev-decision-gate/blob/main/skills/jev-decision-gate/scripts/jev_decision_gate/triage.py#L15
验证标签call-site-verified
f/jev-leftpad
README 自述：Set TYPESAFE_API_KEY before using it. jev-leftpad uses jev-latest through
证据https://github.com/f/jev-leftpad/blob/main/src/index.js#L1
验证标签call-site-verified
fabricioctelles/skills
README 自述：references/jev-integration.md — integration patterns and code examples
证据https://github.com/fabricioctelles/skills/blob/main/README.md#L49
验证标签readme-claim
FBddcz/embodied-jev
README 自述：🧠 模型入口放在界面里：可接 TypeSafe Jev、Claude 原生 API、OpenAI 兼容 API，以及本地 MiniCPM5-2B。
证据https://github.com/FBddcz/embodied-jev/blob/main/README.md#L42
验证标签readme-claim
featherless-ai/simple-jev
README 自述：curl https://simple-jev-demo-api.featherless.ai/v1/models
证据https://github.com/featherless-ai/simple-jev/blob/main/README.md#L20
验证标签readme-claim
feder-cr/jev
README 自述：
证据https://github.com/feder-cr/jev/blob/main/src/main.cpp#L166
验证标签call-site-verified
FeiLiuEM/tetrajev
README 自述：Four readings. One routed decision. Two frozen open-weight models, four independent readings per task, one fit-free decision layer — Jev-class decision behaviour with no fine-tuning and no distillation.
证据https://github.com/FeiLiuEM/tetrajev/blob/main/scripts/run_jev_api.py#L5
验证标签call-site-verified
FerryCorleone/crush-monitor
README 自述：一个分析你和 Crush 或对象聊天的小工具，支持 Jev、OpenAI Decisions（测试版）和常见文本大模型。帮你读懂一点对方的情绪和想法，也看看自己的回复哪里没表达好、可以怎么调整。
证据https://github.com/FerryCorleone/crush-monitor/blob/main/server/analysis.ts#L7
验证标签call-site-verified
flitsinc/go-llms
README 自述：TypeSafe System One models (Jev): typed judgments with probabilities, driven by a JSON output schema
证据https://github.com/flitsinc/go-llms/blob/main/README.md#L9
验证标签readme-claim
FluidInference/FluidUse
README 自述：Jev-style decision model: a 322M mmBERT encoder plus decision head that answers typed
证据https://github.com/FluidInference/FluidUse/blob/main/Sources/FluidUse/PublishedCoreML/PublishedCoreMLTypes.swift#L80
验证标签call-site-verified
Foadsf/jev-for-engineers
README 自述：so twelve questions in one call cost about what one call costs. TypeSafe
证据https://github.com/Foadsf/jev-for-engineers/blob/main/jev.py#L11
验证标签call-site-verified
FrancoisChastel/jev-code
README 自述：jev-code
证据https://github.com/FrancoisChastel/jev-code/blob/main/README.md#L1
验证标签readme-claim
freepik-company/jev-mcp
README 自述：TypeSafe's System One model, through OpenRouter or TypeSafe directly.
证据https://github.com/freepik-company/jev-mcp/blob/main/internal/judgment/tasks.go#L35
验证标签call-site-verified
frostney/clean-code-review
README 自述：Point it at a public GitHub pull request, a diff or a file. Jev, TypeSafe's
证据https://github.com/frostney/clean-code-review/blob/main/README.md#L7
验证标签readme-claim
gamesonrblx/Jevbridge
README 自述：ACP and MCP adapter that bridges TypeSafe Jev with any LLM.
证据https://github.com/gamesonrblx/Jevbridge/blob/main/src/types.ts#L111
验证标签call-site-verified
gbh1234/Adaptive-rag
README 自述：INTENT_BACKEND=jev
证据https://github.com/gbh1234/Adaptive-rag/blob/main/README.md#L83
验证标签readme-claim
genspark-ai/genoffice
README 自述：Six apps, one AI panel, a file search reranked by TypeSafe Jev, an Ask AI
证据https://github.com/genspark-ai/genoffice/blob/main/apps/shell/src/main/file-index/decider.ts#L52
验证标签call-site-verified
get-convex/convex-evals
README 自述：bun run decisions dry-run --provider openrouter --model typesafe/jev-1.13 --limit-evals 2
证据https://github.com/get-convex/convex-evals/blob/main/README.md#L44
验证标签readme-claim
GhalebDweikat/winnow
README 自述：winnow replay score --judged ~/.winnow/replay/judged-typesafe.jsonl
证据https://github.com/GhalebDweikat/winnow/blob/main/sidecar/src/winnow/hooks.py#L282
验证标签call-site-verified
GhrezaKh74/JevTicktRouter
README 自述：Structured support-ticket triage built on .NET 10 and React 19, using TypeSafe Jev to make
证据https://github.com/GhrezaKh74/JevTicktRouter/blob/master/backend/JevTicketRouter.Infrastructure/Decisions/SelfHosted/SelfHostedOptions.cs#L42
验证标签call-site-verified
gi-dellav/hsom
README 自述：hsom* is a strict-allowlist harness. environment.toml defines which shell commands Jev (the TypeSafe System One classifier) may select. In one step, Jev picks: command → subcommand → flags (boolean or valued) → path slot
证据https://github.com/gi-dellav/hsom/blob/main/README.md#L5
验证标签readme-claim
githubnext/localjev
README 自述：A local, Jev-compatible POST /v1/systemone API written in TypeScript for
证据https://github.com/githubnext/localjev/blob/main/src/server.ts#L19
验证标签call-site-verified
Godefroy/micdrop
README 自述：@micdrop/typesafe - Jev, the System One model of TypeSafe, classifying each turn of the user into typed decisions in a few hundred ms
证据https://github.com/Godefroy/micdrop/blob/main/README.md#L27
验证标签readme-claim
GodsBoy/jev-agent-skill-router
README 自述：
证据https://github.com/GodsBoy/jev-agent-skill-router/blob/main/src/jev_router/transport.py#L12
验证标签call-site-verified
gopaljigaur/decide
README 自述：them through one API: TypeSafe's hosted Jev, OpenRouter, OpenAI's Decisions
证据https://github.com/gopaljigaur/decide/blob/main/decide/backends/typesafe.py#L20
验证标签call-site-verified
Govcraft/emergent
README 自述：Judgment as a Pipeline Step: jev-handler
证据https://github.com/Govcraft/emergent/blob/main/README.md#L161
验证标签readme-claim
gtaras7/typesafe-jev
README 自述：A local screening workbench for a folder of CVs, built on TypeSafe's Jev model. Point it at one CV or
证据https://github.com/gtaras7/typesafe-jev/blob/main/cv-screen/src/cli.ts#L97
验证标签call-site-verified
GuSheng107/helpme-Jev
README 自述：JEV decides. LLM explains. You decide.
证据https://github.com/GuSheng107/helpme-Jev/blob/main/README.md#L55
验证标签readme-claim
hamgua/alpha-trading-bot-okx
README 自述：Jev 快车道（TypeSafe System One）
证据https://github.com/hamgua/alpha-trading-bot-okx/blob/main/README.md#L361
验证标签readme-claim
harlan-zw/harlan-agent-kit
README 自述：api.typesafe.ai or the Vercel AI Gateway.
证据https://github.com/harlan-zw/harlan-agent-kit/blob/main/packages/advocaat/README.md#L9
验证标签readme-claim
haseeb-heaven/jev-system-one
README 自述：A polished terminal interface where OpenAI generates useful answers and TypeSafe Jev independently evaluates their relevance, reliability, and quality.
证据https://github.com/haseeb-heaven/jev-system-one/blob/master/src/jev_system_one/jev.py#L3
验证标签call-site-verified
hatt-io/jevkeep
README 自述：Before compaction, JevKeep reads the conversation and asks Jev, a relevance-scoring model accessed through the Typesafe API, which excerpts still matter to the current task.
证据https://github.com/hatt-io/jevkeep/blob/main/README.md#L17
验证标签readme-claim
hellogumbo/should-ai-kill-us-all
README 自述：We ask Jev, TypeSafe AI's System One model, whether AI should kill us all. Every ten minutes. Using the actual headlines.
证据https://github.com/hellogumbo/should-ai-kill-us-all/blob/main/functions/api/verdict.js#L10
验证标签call-site-verified
hellozenstrategist-lab/eutrya
README 自述：A standalone terminal agent with Jev in its decision loop.
证据https://github.com/hellozenstrategist-lab/eutrya/blob/main/extensions/eutrya-adaptive-extension/integration/native-v02.mjs#L9
验证标签call-site-verified
hemanth/is-lesion
README 自述：jev — TypeSafe System One (jev-1.14.0) with 1 Hugging Face visual reference exemplar per class (noul + choice + score + box)
证据https://github.com/hemanth/is-lesion/blob/main/README.md#L48
验证标签readme-claim
hev/reranker
README 自述：nDCG@10, every system permuting the same BM25 top-30 shortlist per query. Jev uses the
证据https://github.com/hev/reranker/blob/main/hev_rerank/rerank.py#L18
验证标签call-site-verified
heyjunpenn/awesome-jev
README 自述：| openai-scala-client | ★ 249 | Scala | 2026-09-23 | Scala client module for TypeSafe System One that submits Jev Noul, Choice, and Score questions and returns typed answers. |
证据https://github.com/heyjunpenn/awesome-jev/blob/main/README.md#L80
验证标签readme-claim
HorusJiang/dsh-jev-tools
README 自述：一个 DeepSeek Harness 插件：把 Jev 接进长会话，在工具结果进入上下文之前做判定。
证据https://github.com/HorusJiang/dsh-jev-tools/blob/main/README.md#L18
验证标签readme-claim
hotchpotch/jev-reranker
README 自述：| model | Explicit value → JEV_MODEL (environment/.env) → jev-latest |
证据https://github.com/hotchpotch/jev-reranker/blob/main/README.md#L434
验证标签readme-claim
hraness/algal
README 自述：or reuse a TypeSafe key. doctor --clef checks configuration without an API
证据https://github.com/hraness/algal/blob/main/README.md#L895
验证标签readme-claim
hunkim/solar-mini4-jev
README 自述：A drop-in wrapper that exposes Upstage Solar Pro4 (default; solar-mini4 optional) through the TypeSafe Jev System One API shape.
证据https://github.com/hunkim/solar-mini4-jev/blob/main/engine.py#L158
验证标签call-site-verified
huntedman/JevLint
README 自述：JevLint is powered by Jev, TypeSafe AI’s System One model. It sends selected source
证据https://github.com/huntedman/JevLint/blob/main/src/jev-client.ts#L52
验证标签call-site-verified
HyunjunJeon/jev-judgment
README 自述：TypeSafe Jev and acts on the probabilities. The
证据https://github.com/HyunjunJeon/jev-judgment/blob/main/README.md#L6
验证标签readme-claim
iamjonatha/decisionkit-dotnet
README 自述：A strongly typed, provider-independent decision engine for .NET, with TypeSafe JEV support.
证据https://github.com/iamjonatha/decisionkit-dotnet/blob/main/README.md#L3
验证标签readme-claim
iamtoomas/JevLint
README 自述：JevLint is powered by Jev, TypeSafe AI’s System One model. It sends selected source
证据https://github.com/iamtoomas/JevLint/blob/main/src/jev-client.ts#L52
验证标签call-site-verified
IAmUnbounded/save-token-jev-clean
README 自述：| TYPESAFE_API_KEY | required unless stored in the macOS Keychain (CLI and library only) | TypeSafe/Jev API key |
证据https://github.com/IAmUnbounded/save-token-jev-clean/blob/main/README.md#L185
验证标签readme-claim
iamvatsalpatel/tiershift
README 自述：The judgment comes from TypeSafe Jev , a calibrated decision model. About 180 ms. Four cents per thousand routes.
证据https://github.com/iamvatsalpatel/tiershift/blob/main/bench/experiments/gate-experiment.ts#L10
验证标签call-site-verified
iefnaf/pi-jev
README 自述：
证据https://github.com/iefnaf/pi-jev/blob/main/README.md#L9
验证标签readme-claim
ielab/llm-rankers
README 自述：: New experiments with Jev (TypeSafe AI's non-generative "System One" model) as a pointwise / pairwise / setwise / listwise reranker, including all 100 candidates in a single request. Check out the jev folder for results
证据https://github.com/ielab/llm-rankers/blob/main/README.md#L7
验证标签readme-claim
ilyamk/jev-gmail-ai-spam-filter-and-labeling
README 自述：Semantic email classification powered by Jev, with confidence-aware automation,
证据https://github.com/ilyamk/jev-gmail-ai-spam-filter-and-labeling/blob/main/README.md#L11
验证标签readme-claim
InternLM/Intern-Decision
README 自述：Intern-Decision-4B reaches 90.02% average accuracy across seven suites (Jev: 88.74%),
证据https://github.com/InternLM/Intern-Decision/blob/main/README.md#L97
验证标签readme-claim
intikhab49/open-jev-typed-decision-engine
README 自述：
证据https://github.com/intikhab49/open-jev-typed-decision-engine/blob/main/README.md#L3
验证标签readme-claim
ipenywis/laya-ultrafast
README 自述：Hosted mode still works. Set DECISION_MODEL=typesafe to use the original Jev policy unchanged.
证据https://github.com/ipenywis/laya-ultrafast/blob/main/laya_ultrafast/laya.py#L48
验证标签call-site-verified
ItIsCuthNotCup/MetaCog
README 自述：judge = SystemOneJudge.jev() # hosted; reads $TYPESAFE_API_KEY
证据https://github.com/ItIsCuthNotCup/MetaCog/blob/main/README.md#L48
验证标签readme-claim
itsmostafa/typesafe-mcp
README 自述：Give your AI agent direct access to models like TypeSafe Jev, Liquid AI d1, Cloudflare Clef, Respan Span-01, GPT-6 Luna Decisions, and Perplexity Decider Flash: quick decisions with real probabilities.
证据https://github.com/itsmostafa/typesafe-mcp/blob/main/README.md#L10
验证标签readme-claim
itsoltech/basal-rs
README 自述：
证据https://github.com/itsoltech/basal-rs/blob/main/contracts/typesafe/README.md#L4
验证标签readme-claim
j1s4nn/duomind
README 自述：Internet connection for setup and Jev API calls
证据https://github.com/j1s4nn/duomind/blob/main/README.md#L111
验证标签readme-claim
JamesANZ/us-legal-mcp
README 自述：JEV_MODEL – default jev-latest
证据https://github.com/JamesANZ/us-legal-mcp/blob/main/README.md#L234
验证标签readme-claim
jamesward/zio-typesafe-ai
README 自述：Jev / System One API — a "System One model" that answers typed, atomic
证据https://github.com/jamesward/zio-typesafe-ai/blob/main/src/main/scala/com/jamesward/zio_typesafe_ai/TypeSafeAI.scala#L3
验证标签call-site-verified
JanOstrowka/typesafe-assist
README 自述：
证据https://github.com/JanOstrowka/typesafe-assist/blob/main/custom_components/typesafe_conversation/api.py#L161
验证标签call-site-verified
jaredpalmer/kev
README 自述：Small Jev-like decision models you can train and run yourself.
证据https://github.com/jaredpalmer/kev/blob/main/README.md#L3
验证标签readme-claim
jarrodwatts/jev-trader
README 自述：With no PRIVATE_KEY it dry-runs: real book, real decisions, simulated fills. Set MODEL=jev and TYPESAFE_AI_API_KEY to use Jev; the default mock is a momentum heuristic stand-in.
证据https://github.com/jarrodwatts/jev-trader/blob/main/src/model.ts#L2
验证标签call-site-verified
jccrump/openbot
README 自述：Google, Mistral, or a local model through Ollama or LM Studio. An optional Jev
证据https://github.com/jccrump/openbot/blob/main/README.md#L7
验证标签readme-claim
jcpsimmons/jev-macos-loop
README 自述：Read https://raw.githubusercontent.com/jcpsimmons/jev-macos-loop/master/AGENTS.md
证据https://github.com/jcpsimmons/jev-macos-loop/blob/master/src/providers.mjs#L17
验证标签call-site-verified
JedimEmO/typesafe-client
README 自述：endorsed by, or supported by TypeSafe AI. For the API itself, see the
证据https://github.com/JedimEmO/typesafe-client/blob/main/crates/typesafe-client/src/answer.rs#L405
验证标签call-site-verified
jerryfane/omp-jev-compaction
README 自述：TypeSafe's Jev decision model, over either the TypeSafe API or OpenRouter.
证据https://github.com/jerryfane/omp-jev-compaction/blob/master/src/asker.ts#L21
验证标签call-site-verified
jev-chat/jev-chat-windows
README 自述：判断 · Jev —— 判断意图、紧张度，并给三条候选排序。来源选 OpenRouter（默认，key 在
证据https://github.com/jev-chat/jev-chat-windows/blob/main/README.md#L47
验证标签readme-claim
jhd3197/Prompture
README 自述：driver = get_decision_driver_for_model("typesafe/jev-latest")
证据https://github.com/jhd3197/Prompture/blob/main/prompture/drivers/kev_decision_driver.py#L4
验证标签call-site-verified
jkudish/jev-mcp
README 自述：Jev is TypeSafe's System One model: it returns typed answers with calibrated probability distributions, not generated text. A verify call is a Choice over supports / contradicts / says_nothing, so you see the whole distr
证据https://github.com/jkudish/jev-mcp/blob/main/README.md#L733
验证标签readme-claim
jmanhype/jev-dspy-lab
README 自述：and other callers that use TypeSafe decisions inside DSPy-style pipelines.
证据https://github.com/jmanhype/jev-dspy-lab/blob/main/scripts/verify_integration.py#L41
验证标签call-site-verified
joelhooks/pi-fast-jev-compaction
README 自述：Sends a compact representation of tool calls and surrounding conversation to TypeSafe's /v1/systemone endpoint.
证据https://github.com/joelhooks/pi-fast-jev-compaction/blob/main/src/core/request.ts#L8
验证标签call-site-verified
jomatsu/pi-jev-auto-mode
README 自述：decision models, with Jev (TypeSafe System One) as the default. OpenAI Decisions,
证据https://github.com/jomatsu/pi-jev-auto-mode/blob/main/src/jev/engine.ts#L95
验证标签call-site-verified
jomatsu/zod-jev
README 自述：A single parseAsync call sends every rule of that schema to Jev in a single request, then turns
证据https://github.com/jomatsu/zod-jev/blob/main/scripts/check-package.mjs#L35
验证标签call-site-verified
jonathanavis96/jev-kit
README 自述：What is Jev, and why a kit
证据https://github.com/jonathanavis96/jev-kit/blob/main/README.md#L39
验证标签readme-claim
jozefRudy/job_search
README 自述：Optional; omit either key to use the default (https://api.typesafe.ai, jev-latest).
证据https://github.com/jozefRudy/job_search/blob/main/README.md#L52
验证标签readme-claim
jucelioalencar/linearmemory
README 自述：Contracts: OpenAI Decisions and TypeSafe System One. Settings are available at GET/PUT /api/settings/search-filter; existing search snapshots are returned by GET /api/search-results/:eventId without rerunning retrieval.
证据https://github.com/jucelioalencar/linearmemory/blob/main/README.md#L175
验证标签readme-claim
jujumilk3/jev-calibration-audit
README 自述：An independent, API-only audit of the calibration of TypeSafe AI's Jev, a
证据https://github.com/jujumilk3/jev-calibration-audit/blob/main/src/jev_audit/client.py#L12
验证标签call-site-verified
justinhe16/trade-jev
README 自述：Run: a backtest that calls Jev and stores its answers and trades in runs/ /. It's the only step that costs money.
证据https://github.com/justinhe16/trade-jev/blob/main/src/trade_jev/live.py#L993
验证标签call-site-verified
jxucoder/mimicry
README 自述：and TypeSafe; requests and responses are saved locally under ignored
证据https://github.com/jxucoder/mimicry/blob/master/README.md#L37
验证标签readme-claim
kavehmz/typesafe-playground
README 自述：Small, interactive experiments with TypeSafe Jev: from understanding a support message to making decisions for a car in a 3D world.
证据https://github.com/kavehmz/typesafe-playground/blob/main/demo01/server.mjs#L67
验证标签call-site-verified
keltokhy/jgrep
README 自述：Each line becomes one yes/no question to Jev, TypeSafe's decision
证据https://github.com/keltokhy/jgrep/blob/main/README.md#L16
验证标签readme-claim
kesku/jev-freeform
README 自述：Can TypeSafe Jev Choice generate text through successive typed decisions? Well I tried it
证据https://github.com/kesku/jev-freeform/blob/main/main.mjs#L172
验证标签call-site-verified
kevinbadi/jev-voice
README 自述：Everything runs locally except one ~250 ms call to Jev (TypeSafe's System One
证据https://github.com/kevinbadi/jev-voice/blob/main/jev_voice/brain.py#L174
验证标签call-site-verified
Kevthetech143/super-jev
README 自述：One judge sits between your agent and your data: the agent asks in its own words, Jev finds the file, checks the claim, permits the action, and remembers what you approved. Your agent stays responsible for the answer.
证据https://github.com/Kevthetech143/super-jev/blob/main/README.md#L9
验证标签readme-claim
kieranklaassen/ruby_llm-typesafe
README 自述：Both register a provider named :typesafe, so don't load both. When you upgrade, remove ruby_llm-typesafe from your Gemfile and use RubyLLM's built-in provider.
证据https://github.com/kieranklaassen/ruby_llm-typesafe/blob/main/README.md#L11
验证标签readme-claim
KiritoKing/midscene-jev-runner
README 自述：The default is TypeSafe System One: POST https://api.typesafe.ai/v1/systemone
证据https://github.com/KiritoKing/midscene-jev-runner/blob/main/README.md#L66
验证标签readme-claim
klauswg/jev-guard
README 自述：Real-time risk triage gateway for crypto exchange deposits & withdrawals — Jev (TypeSafe System One) as the triage layer, deterministic code as the judge.
证据https://github.com/klauswg/jev-guard/blob/master/src/main/java/com/jevguard/client/JevClient.java#L13
验证标签call-site-verified
klauswg/jev-suite
README 自述：Four decision-quality tools built on Jev (TypeSafe System One), one shared kernel — Jev answers structured questions; deterministic code keeps the final say.
证据https://github.com/klauswg/jev-suite/blob/master/jev-fidelity/src/main/java/com/jevsuite/fidelity/eval/EvalRunner.java#L52
验证标签call-site-verified
knowlet/jevlens
README 自述：Manifest V3 Chrome extension for annotating articles, X/Twitter posts, and Threads posts with typed Jev / System One decisions.
证据https://github.com/knowlet/jevlens/blob/master/extension/lib/jev.js#L4
验证标签call-site-verified
kouhxp/gutsy
README 自述：network calls, no per-call cost. The API follows the Jev-style Decisions format, so existing clients
证据https://github.com/kouhxp/gutsy/blob/main/gutsy-inference/gutsy_inference/server.py#L21
验证标签call-site-verified
kraayenjon/awesome-jev
README 自述：
证据https://github.com/kraayenjon/awesome-jev/blob/main/README.md#L2
验证标签readme-claim
kshetrajna12/reflex
README 自述：the "System One" model TypeSafe released in September 2026, built on top of a normal
证据https://github.com/kshetrajna12/reflex/blob/main/scripts/bench_latency.py#L6
验证标签call-site-verified
kyegomez/open-jev
README 自述：System One Models and Jev
证据https://github.com/kyegomez/open-jev/blob/main/README.md#L161
验证标签readme-claim
kyu1204/jgrep
README 自述：jgrep runs on Jev, through TypeSafe or
证据https://github.com/kyu1204/jgrep/blob/main/README.md#L41
验证标签readme-claim
kyu1204/oh-my-harness
README 自述：--preset always wins and never calls Jev. No key and no LLM provider? omh init "description" falls back to safe and says so.
证据https://github.com/kyu1204/oh-my-harness/blob/main/src/catalog/blocks/semantic-rule-guard.ts#L68
验证标签call-site-verified
lakeday-org/perch
README 自述：export PERCH_BASE_URL=https://api.typesafe.ai/v1/systemone
证据https://github.com/lakeday-org/perch/blob/main/.benchmarks/reviewbench/run.mjs#L15
验证标签call-site-verified
lhotwll217/jev-cli
README 自述：JSON in, typed decisions out. A thin CLI over the TypeSafe System One API
证据https://github.com/lhotwll217/jev-cli/blob/main/src/client.ts#L25
验证标签call-site-verified
Li-Evan/awesome-jev
README 自述：http4k TypeSafe connector - TypeSafe connector for the http4k Kotlin toolkit with a typed System One client and a fake, exposing Jev Noul, Choice and Score questions as http4k actions.
证据https://github.com/Li-Evan/awesome-jev/blob/main/README.md#L468
验证标签readme-claim
libingzheren/Jev-Mem
README 自述：Jev-Mem separates the frequent decisions of memory management from the deeper
证据https://github.com/libingzheren/Jev-Mem/blob/main/memory/jev_client.py#L23
验证标签call-site-verified
liu-x27/XavierJev
README 自述：TypeSafe: nothing here calls the Jev API, and nothing was trained on its output.
证据https://github.com/liu-x27/XavierJev/blob/main/eval/decision-models/run.ts#L177
验证标签call-site-verified
liuyanghejerry/Clausura
README 自述：model: jev-latest # Default. Decision model id.
证据https://github.com/liuyanghejerry/Clausura/blob/main/README.md#L330
验证标签readme-claim
luisleo526/doc2mark
README 自述：TypeSafe's Jev can be asked about the cases they leave open (is this
证据https://github.com/luisleo526/doc2mark/blob/main/README.md#L23
验证标签readme-claim
lykycy123/RoboJEV
README 自述：Two-stage JEV control of a Franka Panda in MuJoCo. Measured simulator state → task intent → Cartesian motion and gripper commands.
证据https://github.com/lykycy123/RoboJEV/blob/main/src/jev_vla_sim/config.py#L27
验证标签call-site-verified
lyramakesmusic/jevbot
README 自述：Completeness judge: A separate noul question asks "is the reply complete?" — jev stops when it thinks it's done
证据https://github.com/lyramakesmusic/jevbot/blob/master/README.md#L11
验证标签readme-claim
maayanlevy/mysql-ailike
README 自述：Requires a TypeSafe API key. Note: evaluated values and prompts are
证据https://github.com/maayanlevy/mysql-ailike/blob/main/src/typesafe.h#L16
验证标签call-site-verified
madisonrickert/jev-permission-gate
README 自述：A shell-aware blocklist sits in front of Jev. A call it flags can still be denied by Jev but is never allowed, and a call that carries a secret never leaves your machine.
证据https://github.com/madisonrickert/jev-permission-gate/blob/main/hooks/register.ts#L25
验证标签call-site-verified
MahmoudAdelbghany/jev-browser
README 自述：one Jev call (parallel questions, ~300 ms, no LLM tokens)
证据https://github.com/MahmoudAdelbghany/jev-browser/blob/main/src/jev.mjs#L1
验证标签call-site-verified
mahynotch/newsscore
README 自述：News sentiment scoring for stocks, powered by TypeSafe Jev by default. Mainstream financial news APIs go in, a pluggable
证据https://github.com/mahynotch/newsscore/blob/main/src/newsscore/scoring/jev.py#L416
验证标签call-site-verified
maker-KK/todo-jev
README 自述：A task-routing experiment powered by TypeSafe Jev, structured skill profiles, and basic environment checks.
证据https://github.com/maker-KK/todo-jev/blob/main/app/classifier.py#L14
验证标签call-site-verified
malDuffin/typesafe-3d-chess
README 自述：D chess powered by TypeSafe AI (Jev). AI vs AI by default, or play either side. Multiple difficulty levels.
证据https://github.com/malDuffin/typesafe-3d-chess/blob/main/README.md#L2
验证标签readme-claim
mandu5/jevcompat
README 自述：About a hundred servers say they are "Jev-compatible". Nothing said what that means. Now something does.
证据https://github.com/mandu5/jevcompat/blob/main/README.md#L3
验证标签readme-claim
manifoldor/xtags
README 自述：页面文本会发送到所选 API 服务（默认 TypeSafe）做判断——这是它工作的必要部分。发送的是帖子正文和作者 handle；折叠长推文会发送 X 页面数据中已有的全文，即使你尚未点击展开。
证据https://github.com/manifoldor/xtags/blob/main/extension/service.js#L3
验证标签call-site-verified
marcreichel/laya-php
README 自述：curl -fsSL https://api.typesafe.ai/openapi.json -o openapi.json
证据https://github.com/marcreichel/laya-php/blob/main/src/Laya.php#L197
验证标签call-site-verified
MarissaFamularo/citation-verifier
README 自述：supports the sentence, and Jev's confidence in its own call. A probability
证据https://github.com/MarissaFamularo/citation-verifier/blob/main/README.md#L34
验证标签readme-claim
markjaquith/typesafe-ai-playground
README 自述：The experiments live in a Rust CLI called typesafe-ai, built with
证据https://github.com/markjaquith/typesafe-ai-playground/blob/main/src/typesafe.rs#L57
验证标签call-site-verified
MartinPuli/f1
README 自述：The engine and experiment scripts remain in the repository. The published interface only plays recorded races; it has no creation or admin screen. Replays use saved decisions and frames without calling TypeSafe.
证据https://github.com/MartinPuli/f1/blob/main/scripts/refine-prompts.mjs#L55
验证标签call-site-verified
mattn/go-jev
README 自述：text. Comes with jev-cli, a command-line tool built on it for UNIX pipelines.
证据https://github.com/mattn/go-jev/blob/main/jev.go#L19
验证标签call-site-verified
mejiasd3v/pi-jev-router
README 自述：
证据https://github.com/mejiasd3v/pi-jev-router/blob/main/index.ts#L407
验证标签call-site-verified
memos-im/xianyu-slang-helper
README 自述：追加 --ai-review 可用 TypeSafe System One 判别模型（Jev；自托管 DiffusionGemma 同 API）剔除上游的"模板化噪音"变体：每个候选变体问一个 Noul（是/否）问题，取概率 ≥ 阈值判为噪音，不生成文本、无解析、返回校准概率。未配置 key 时自动降级跳过，不阻断同步。
证据https://github.com/memos-im/xianyu-slang-helper/blob/main/README.md#L147
验证标签readme-claim
Micha0827/snapjudge
README 自述：The HTTP API follows the TypeSafe System One format
证据https://github.com/Micha0827/snapjudge/blob/main/README.md#L7
验证标签readme-claim
misbahsy/doc-router
README 自述：Jev (TypeSafe System One) — the page judge. Given the
证据https://github.com/misbahsy/doc-router/blob/main/README.md#L36
验证标签readme-claim
mithalouni/system-one-open
README 自述：Base URL: https://mithalouni--jev-serve-e2b-full-server-web.modal.run (scales to zero; first request after idle takes ~60 s).
证据https://github.com/mithalouni/system-one-open/blob/main/README.md#L31
验证标签readme-claim
mizchi/jev-lint
README 自述：export TYPESAFE_API_KEY=... # https://typesafe.ai
证据https://github.com/mizchi/jev-lint/blob/main/src/clef.ts#L54
验证标签call-site-verified
mizchi/jev-test-filter
README 自述：jev-test-filter reads a git diff, asks a model how much that change can
证据https://github.com/mizchi/jev-test-filter/blob/main/src/jev.ts#L26
验证标签call-site-verified
mkotlikov/jev-grug
README 自述：The demo now connects to the real Jev API when TYPESAFE_API_KEY is configured. Without a key it falls back to a deterministic mock, so contributors can still run the interface immediately.
证据https://github.com/mkotlikov/jev-grug/blob/main/app/api/chat/route.ts#L30
验证标签call-site-verified
mmastrac/djev
README 自述：POST /v1/systemone implements the Jev decision API. The body holds
证据https://github.com/mmastrac/djev/blob/main/structured_server.py#L2014
验证标签call-site-verified
mmastrac/djev-spark
README 自述：it implements Jev's POST /v1/systemone API and passes ordinary chat through
证据https://github.com/mmastrac/djev-spark/blob/main/scripts/long-context-probe.py#L35
验证标签call-site-verified
mohsen1/llm-debugger-vscode-extension
README 自述：Each step's "what next?" is answered by Jev, a
证据https://github.com/mohsen1/llm-debugger-vscode-extension/blob/main/README.md#L7
验证标签readme-claim
morganlinton/Albatross
README 自述：Jev Mode puts TypeSafe AI's decision model before the main LLM. Supported questions
证据https://github.com/morganlinton/Albatross/blob/main/src/jev.rs#L13
验证标签call-site-verified
moritzkremb/jev-voice-browser
README 自述：the server asks Jev (TypeSafe's System One model, jev-1.13.0) one request with a dozen typed
证据https://github.com/moritzkremb/jev-voice-browser/blob/main/src/jev.js#L7
验证标签call-site-verified
mossyfield/ST-jeved
README 自述：Jeved is a SillyTavern extension. It sends chat messages to Jev, a small decision model, with
证据https://github.com/mossyfield/ST-jeved/blob/main/src/classifier.js#L22
验证标签call-site-verified
mrbanana55/Decision_Blackjack
README 自述：Jev (Remote Cloud AI via TypeSafe System One)
证据https://github.com/mrbanana55/Decision_Blackjack/blob/main/README.md#L8
验证标签readme-claim
mrmps/classifier-dev
README 自述：Both tiers answer from TypeSafe's Jev, a decision
证据https://github.com/mrmps/classifier-dev/blob/main/README.md#L128
验证标签readme-claim
n42blockchain/N42-rs
README 自述：用户在页面确认费用并提交，页面展示 requestId。运营方建立 job.json，包含下方字段，运行 TYPESAFE_API_KEY=... N42_DECISION_SIGNING_KEY=... cargo run -p n42-decision-relay -- evaluate job.json ./archive > result.json。TYPESAFE_API_URL 默认官方 https://api.typesa
证据https://github.com/n42blockchain/N42-rs/blob/main/ddn/docs/decision/README.md#L44
验证标签readme-claim
Nainish-Rai/jev-frontend-qa
README 自述：Bounded Jev decisions: typed operation/target choices over observed controls; exact fixture values and correctness checks remain in code.
证据https://github.com/Nainish-Rai/jev-frontend-qa/blob/main/src/jev_frontend_qa/core/model_client.py#L3
验证标签call-site-verified
Nancy-Chauhan/hearth-jev-rental-search
README 自述：A TypeSafe API key. A search costs a fraction of a cent
证据https://github.com/Nancy-Chauhan/hearth-jev-rental-search/blob/main/README.md#L18
验证标签readme-claim
NandhaKishorM/laya
README 自述：wire protocol as TypeSafe's hosted Jev API. Laya's answer payload is already
证据https://github.com/NandhaKishorM/laya/blob/main/laya-dotnet/tools/bench_python.py#L191
验证标签call-site-verified
Nasrallah-AL/jev-cli
README 自述：Fast, typed AI judgments from the command line, powered by TypeSafe's Jev model.
证据https://github.com/Nasrallah-AL/jev-cli/blob/main/README.md#L8
验证标签readme-claim
nassim-arifette/jevgrep
README 自述：authorized repository, asks Jev to score all eligible fragments, then returns the
证据https://github.com/nassim-arifette/jevgrep/blob/main/experiments/jev-contract/offline-worker.mjs#L4
验证标签call-site-verified
ndolinschi/harnessjudge
README 自述：After an agent step (paste a trace) → ok / retry / escalate / stop — for agent builders, powered by TypeSafe Jev.
证据https://github.com/ndolinschi/harnessjudge/blob/main/src/app/api/decide/route.ts#L7
验证标签call-site-verified
ndolinschi/pulselane
README 自述：Clinic triage decisions powered by TypeSafe Jev (System One).
证据https://github.com/ndolinschi/pulselane/blob/main/src/app/api/decide/route.ts#L9
验证标签call-site-verified
ndolinschi/toolgate
README 自述：Agent tool/MCP call gate — paste a planned tool call + context → allow / ask_human / deny with risk scores via TypeSafe Jev.
证据https://github.com/ndolinschi/toolgate/blob/main/src/app/api/decide/route.ts#L9
验证标签call-site-verified
newfull5/malkuth
README 自述：suites and about 129,000 questions. Every model answered the same items. Jev was measured through its API.
证据https://github.com/newfull5/malkuth/blob/main/tools/laya_bench.py#L27
验证标签call-site-verified
NiazMorshed2007/jev-review
README 自述：Your API key stays on your machine. Jev Review has no hosted backend, database, telemetry service, or author-operated proxy. The only remote request is sent directly to the configured Jev API.
证据https://github.com/NiazMorshed2007/jev-review/blob/main/src/jev/client.ts#L4
验证标签call-site-verified
nican2018/shade-arena-jev-monitor
README 自述：Strong per-action gate. Scoring each tool call before it runs, Jev ranked sabotage above almost all benign actions. Gemini Flash's scores piled up at 0 and 100.
证据https://github.com/nican2018/shade-arena-jev-monitor/blob/main/llms/typesafe_llm.py#L20
验证标签call-site-verified
nicobailon/pi-powerline-footer
README 自述：When the agent is working, Enter sends your message as steering and Alt+Enter queues it as a follow-up. Set autoFollowUp under powerline to let TypeSafe's Jev make that call for plain Enter:
证据https://github.com/nicobailon/pi-powerline-footer/blob/main/README.md#L329
验证标签readme-claim
nicobailon/surf-cli
README 自述：--allow-semantic - Opt in to bounded model decisions for a semantic workflow (model from SURF_SEMANTIC_MODEL, default jev-1.13.0; extract steps always use clairic with Cloudflare credentials)
证据https://github.com/nicobailon/surf-cli/blob/main/native/semantic-cli.cjs#L632
验证标签call-site-verified
nidhi-singh02/agent-router
README 自述：shared accounts), asks TypeSafe to rank what is left and choose an effort, starts the chosen
证据https://github.com/nidhi-singh02/agent-router/blob/main/packages/router/src/commands/effort.ts#L264
验证标签call-site-verified
Nisaka520/JevIntent
README 自述：POST https://api.typesafe.ai/v1/systemone：
证据https://github.com/Nisaka520/JevIntent/blob/main/README.md#L232
验证标签readme-claim
NobleSpartan6/otto
README 自述：In Hybrid mode, GPT-6 Astra provides a plan and generated text when needed. Jev chooses concrete actions between planning calls.
证据https://github.com/NobleSpartan6/otto/blob/main/core/typesafe.ts#L4
验证标签call-site-verified
noetion/dsh-jev
README 自述：dsh-jev is a DeepSeek Harness bundle that registers the jev_ask tool. A DSH agent on a profile that loads it can send typed noul, choice, and score questions to TypeSafe Jev.
证据https://github.com/noetion/dsh-jev/blob/main/src/client.ts#L10
验证标签call-site-verified
NorbertBodziony/guard-jev
README 自述：Text moderation demo: one TypeSafe systemOne call screens 7 Noul hazards + 1 severity Score in parallel. Verdict computed in code via policy thresholds.
证据https://github.com/NorbertBodziony/guard-jev/blob/main/app/api/moderate/route.ts#L2
验证标签call-site-verified
notque/vexjoy-agent
README 自述：/d — Jev-Powered Router
证据https://github.com/notque/vexjoy-agent/blob/main/README.md#L44
验证标签readme-claim
ntedvs/commentcop
README 自述：Use --dry-run to preview what it finds without calling Jev. Add exclusions with --ignore or a .commentcopignore file. --help has the rest.
证据https://github.com/ntedvs/commentcop/blob/main/src/judge.ts#L1
验证标签call-site-verified
Nyarlathoteppppp/pi-heed
README 自述：messages, 9,090 real tool calls) through pi-heed in enforce mode with Jev, without executing anything
证据https://github.com/Nyarlathoteppppp/pi-heed/blob/main/src/judge.ts#L53
验证标签call-site-verified
obie/ruby_decision_model
README 自述：Requests go to https://api.typesafe.ai/v1/systemone. The default model is
证据https://github.com/obie/ruby_decision_model/blob/main/README.md#L85
验证标签readme-claim
oficcejo/aiagents-stock
README 自述：Key 申请：https://console.typesafe.ai ｜完整说明：TypeSafe Jev决策引擎接入说明
证据https://github.com/oficcejo/aiagents-stock/blob/main/README.md#L170
验证标签readme-claim
ojusave/beat-jev
README 自述：A penalty shootout against TypeSafe Jev. Take turns shooting and keeping goal. One Render Workflows task owns the match, with scores saved in Render Postgres.
证据https://github.com/ojusave/beat-jev/blob/main/python/app/keeper.py#L2
验证标签call-site-verified
Olli0103/openclaw-typesafe-ai
README 自述：Jev returns decisions over predefined answer spaces. It does not generate conversation, prose or code. This package therefore registers a tool and no model provider, model picker entry, hook or background service.
证据https://github.com/Olli0103/openclaw-typesafe-ai/blob/main/src/client.ts#L4
验证标签call-site-verified
olsenbrands/fable-foreman
README 自述：Jev, from TypeSafe AI, is a decision model rather than a chatbot. It answers narrow yes/no, pick-one and score questions in under a second, for about 0.02 cents a call. Fable Foreman uses it to point the lead's attention
证据https://github.com/olsenbrands/fable-foreman/blob/main/README.md#L11
验证标签readme-claim
OmniJev/OneJev
README 自述：
证据https://github.com/OmniJev/OneJev/blob/main/qev/cli.py#L143
验证标签call-site-verified
ondata/vd-jev
README 自述：Ask three questions about the message column, one jev-column each:
证据https://github.com/ondata/vd-jev/blob/main/README.md#L55
验证标签readme-claim
onlyoneaman/jev-eval
README 自述：Jev is a model that does not write text. You give it some text and a question with fixed answers, and it gives back a probability for each answer. One call looks like this:
证据https://github.com/onlyoneaman/jev-eval/blob/main/jev_eval/backends.py#L16
验证标签call-site-verified
opencx-labs/zevals
README 自述：Jev is reachable through several gateways, and each names the same yes/no question differently. Three clients are bundled, all of them plain fetch with no extra dependencies:
证据https://github.com/opencx-labs/zevals/blob/main/README.md#L273
验证标签readme-claim
openlayer-ai/jevals
README 自述：export AI_GATEWAY_API_KEY=... # Jev through Vercel AI Gateway. TYPESAFE_API_KEY and OPENROUTER_API_KEY also work.
证据https://github.com/openlayer-ai/jevals/blob/main/src/jevals/backends/_base.py#L65
验证标签call-site-verified
OskarLebuda/precog
README 自述：Asks TypeSafe Jev which link a visitor is about to click, then warms
证据https://github.com/OskarLebuda/precog/blob/main/README.md#L19
验证标签readme-claim
oso95/x-scanner
README 自述：Behavioral labels on every post you scroll past on X, judged by Jev,
证据https://github.com/oso95/x-scanner/blob/main/README.md#L3
验证标签readme-claim
PanAchy/jevvy
README 自述：Custom endpoints use Jevvy's default approval policy unless you replace it. Its calibration evidence applies only when the route is verified to serve the calibrated Jev model.
证据https://github.com/PanAchy/jevvy/blob/main/packages/core/src/system-one.ts#L12
验证标签call-site-verified
pavy23/web-doom
README 自述：Tactical policy (TypeSafe System One, Jev). At each step with an enemy
证据https://github.com/pavy23/web-doom/blob/main/README.md#L288
验证标签readme-claim
pengchujin/ad-radar
README 自述：智能识别：用你自己的 Jev API key，识别广告（硬广 / 软广 / 引流）和你选择的话题（AI、科技、军事、政治、性别对立、色情低俗、明星八卦、炒股理财、体育、游戏，也可以自己添加），标在内容旁边或折叠起来；还能估计内容由 AI 生成的比例。
证据https://github.com/pengchujin/ad-radar/blob/main/README.md#L6
验证标签readme-claim
PerryLink/jevcore
README 自述：TypeSafe Jev for DeepSeek Harness
证据https://github.com/PerryLink/jevcore/blob/main/README.md#L8
验证标签readme-claim
peterfriese/system-one-foundation-models
README 自述：Jev Cloud: TypeSafe AI hosted service on https://api.typesafe.ai.
证据https://github.com/peterfriese/system-one-foundation-models/blob/main/README.md#L364
验证标签readme-claim
philippdubach/pi-jev-router
README 自述：pi-jev-router is a pi extension. It classifies each task with Jev (TypeSafe
证据https://github.com/philippdubach/pi-jev-router/blob/main/src/classifier.ts#L13
验证标签call-site-verified
picaye/jev-compaction
README 自述：Jev judges to be stale, and it shows Jev the whole conversation while deciding.
证据https://github.com/picaye/jev-compaction/blob/main/README.md#L19
验证标签readme-claim
PistachioAIHQ/jev-synergy-screening
README 自述：export TYPESAFE_API_KEY="$(tr -d ' ' < ~/.config/typesafe/api_key)"
证据https://github.com/PistachioAIHQ/jev-synergy-screening/blob/main/src/config.py#L43
验证标签call-site-verified
pithings/advocaat
README 自述：Requests use https://ai-gateway.vercel.sh/v4/ai and model typesafe-ai/jev.
证据https://github.com/pithings/advocaat/blob/main/src/api.ts#L127
验证标签call-site-verified
PixelML/av
README 自述：Configure a TypeSafe System One key to make Jev refinement automatic for av ask:
证据https://github.com/PixelML/av/blob/main/README.md#L35
验证标签readme-claim
pnthn-ai/polar_llama
README 自述：TypeSafe System One: A native Rust inference layer for the TypeSafe API — typed, calibrated yes/no, choice and score questions answered per row, or one Pydantic contract applied to every line of a document in a single re
证据https://github.com/pnthn-ai/polar_llama/blob/main/README.md#L22
验证标签readme-claim
PromptEngineer48/laya-vs-jev-arena
README 自述：docker run -d --name arena --restart unless-stopped -p 80:8732 --cpus 1 --memory 512m -e TYPESAFE_API_KEY=your_key_here ghcr.io/promptengineer48/laya-vs-jev-arena:latest
证据https://github.com/PromptEngineer48/laya-vs-jev-arena/blob/main/server.py#L29
验证标签call-site-verified
psyb0t/aigate
README 自述：| decidealot _(optional, DECIDEALOT=1 and/or DECIDEALOT_CUDA=1)_ | Typed decisions with local Laya, Von, CLM, and hosted TypeSafe Jev over the TypeSafe System One API. POST /v1/systemone takes a model selector, a state,
证据https://github.com/psyb0t/aigate/blob/main/README.md#L179
验证标签readme-claim
PyModel/jev-judge-mcp
README 自述：uvx --from 'jev-judge-mcp[typesafe]' jev-judge-mcp setup
证据https://github.com/PyModel/jev-judge-mcp/blob/main/evals/confinement/broker.py#L16
验证标签call-site-verified
qddegtya/qualm
README 自述：Jev is the model. This is what qualm adds on top of calling its HTTP API yourself.
证据https://github.com/qddegtya/qualm/blob/main/README.md#L25
验证标签readme-claim
qkal/Canny
README 自述：Jev
证据https://github.com/qkal/Canny/blob/main/README.md#L155
验证标签readme-claim
QuentinDanblon/pi-fast-jev-compaction
README 自述：anything. It asks Jev, over the whole conversation, whether each old tool call and each
证据https://github.com/QuentinDanblon/pi-fast-jev-compaction/blob/main/index.ts#L206
验证标签call-site-verified
Query-farm/vgi-typesafe
README 自述：
证据https://github.com/Query-farm/vgi-typesafe/blob/main/vgi_typesafe/mock_server.py#L39
验证标签call-site-verified
r-ms/mini-jev
README 自述：What a Jev-style interface looks like on a frozen model: classify against a schema that arrives with the request by reading next-token logits, not by generating JSON. Measured on Qwen3-4B.
证据https://github.com/r-ms/mini-jev/blob/main/README.md#L3
验证标签readme-claim
RafalWilinski/vibecheck
README 自述：A Chrome and Firefox extension that runs your draft post through TypeSafe's Jev before you hit Post. It renders a scorecard under the composer with virality, funniness, informativeness, clarity, insult level, ragebait, c
证据https://github.com/RafalWilinski/vibecheck/blob/main/README.md#L3
验证标签readme-claim
rajivkuriakose/typesafe-jev-examples
README 自述：TypeSafe's direct API is behind an early-access waitlist. OpenRouter serves the
证据https://github.com/rajivkuriakose/typesafe-jev-examples/blob/main/src/jevx/client.py#L17
验证标签call-site-verified
ranjan2829/AskJev
README 自述：Talk to Claude in plain English. AskJev drives Brave/Chrome. TypeSafe Jev decides on-page. Guard freezes irreversible clicks.
证据https://github.com/ranjan2829/AskJev/blob/main/mcp/src/jev-client.ts#L16
验证标签call-site-verified
RayenMlayeh/jevcrawl
README 自述：| TYPESAFE_API_KEY | Jev API key (required) | none |
证据https://github.com/RayenMlayeh/jevcrawl/blob/main/README.md#L202
验证标签readme-claim
rbalch/typesafeai-review
README 自述：slices the diff, asks Jev (TypeSafe's System One model) narrow typed questions per hunk,
证据https://github.com/rbalch/typesafeai-review/blob/master/README.md#L4
验证标签readme-claim
rchovatiya88/cyber-breach-jev
README 自述：An infinite 3D arcade cyberpunk arena shooter built with Three.js and powered by Jev—the sub-100ms "System One" decision model developed by TypeSafe AI (released September 2026).
证据https://github.com/rchovatiya88/cyber-breach-jev/blob/main/backend/bot_pilot.py#L50
验证标签call-site-verified
reachjalil/jevlogs
README 自述：
证据https://github.com/reachjalil/jevlogs/blob/main/README.md#L2
验证标签readme-claim
ReallyArtificial/stuntdouble
README 自述：Find out whether a local decision model can replace Jev on your traffic before you switch.
证据https://github.com/ReallyArtificial/stuntdouble/blob/main/sidecars/kev-space.mjs#L42
验证标签call-site-verified
realZachi/pg-jev
README 自述：
证据https://github.com/realZachi/pg-jev/blob/master/README.md#L2
验证标签readme-claim
realZachi/typesafe-adblock
README 自述：BYOK: bring your own key. There is no backend. You paste your own TypeSafe API key
证据https://github.com/realZachi/typesafe-adblock/blob/main/src/typesafe.js#L5
验证标签call-site-verified
receptron/laya
README 自述：implementation (RLAgent.system_one) and as TypeSafe Jev's system_one API, and the output
证据https://github.com/receptron/laya/blob/main/README.md#L15
验证标签readme-claim
reinhard-z/vision-jev
README 自述：Move an object to ask Jev again. Drag it off the road or click its × to remove it.
证据https://github.com/reinhard-z/vision-jev/blob/main/README.md#L24
验证标签readme-claim
RenaGao/jev-dataops
README 自述：Screenshots show the repository's public demo running locally with its bundled synthetic example. The measurements demonstrate the workflow, not JEV screening quality or LLM performance.
证据https://github.com/RenaGao/jev-dataops/blob/main/jev_dataops/jev.py#L23
验证标签call-site-verified
rohanarun/computer-use-cache
README 自述：Set your own TypeSafe API key to enable JEV in either server:
证据https://github.com/rohanarun/computer-use-cache/blob/main/server.py#L85
验证标签call-site-verified
rokbenko/quackd
README 自述：quackd run arm-grip-check --robot arm-01 --by-hand --decision-llm jev --decision-mode shadow
证据https://github.com/rokbenko/quackd/blob/main/quackd/agent/decision/systemone.py#L100
验证标签call-site-verified
RomanSlack/jev-drone
README 自述：jev-drone
证据https://github.com/RomanSlack/jev-drone/blob/main/README.md#L1
验证标签readme-claim
romeromarcelo/jev-retrieval
README 自述：
证据https://github.com/romeromarcelo/jev-retrieval/blob/main/src/jev/client.rs#L28
验证标签call-site-verified
rorshopping/jev-on-a-laptop
README 自述：TypeSafe's own workflow evals (https://evals.typesafe.ai/) put Jev at 67.8% mean accuracy (61.7–76.0% per workflow) against a frontier consensus (average of GPT-6 Astra and Fable 5.1 answering every question), at $0.0004
证据https://github.com/rorshopping/jev-on-a-laptop/blob/main/README.md#L103
验证标签readme-claim
runta-dev/jot
README 自述：![Powered by Jev](https://typesafe.ai/)
证据https://github.com/runta-dev/jot/blob/main/README.md#L19
验证标签readme-claim
rupeshpoojary9/poorjev
README 自述：The poor man's Jev. An open source, local-first "System One" decision layer for LLM apps: typed decisions with provably calibrated confidence . No API key. No waitlist.
证据https://github.com/rupeshpoojary9/poorjev/blob/main/crossbench/jev_client.py#L8
验证标签call-site-verified
ryana/jevify
README 自述：TypeSafe introduction
证据https://github.com/ryana/jevify/blob/main/README.md#L15
验证标签readme-claim
safzanpirani/pi-jev-skill-picker
README 自述：One skill_search call costs about 38,600 Jev input tokens, or $0.0016 at $42 per billion. Jev bills input only. Against gpt-6-astra that is under 1% of what a single un-stripped request wastes.
证据https://github.com/safzanpirani/pi-jev-skill-picker/blob/master/extensions/jev.ts#L12
验证标签call-site-verified
samdotmak/jev-recall
README 自述：Jev Recall gives you LLM re-ranker quality retreival at semantic search prices and speed
证据https://github.com/samdotmak/jev-recall/blob/main/web/lib/systems.ts#L146
验证标签call-site-verified
samtay32/jev-system-architect
README 自述：A system-architecture skill for TypeSafe AI's Jev / System One. Continuously asks: *Where does this system contain fuzzy semantic judgment that should become a small Jev decision primitive?*
证据https://github.com/samtay32/jev-system-architect/blob/main/README.md#L3
验证标签readme-claim
santos-sanz/jev-audio-beeper
README 自述：Private TypeScript proof of concept: identify Spanish profanity with Jev's typed probabilistic decisions and replace the matching word intervals with a beep without changing the audio duration.
证据https://github.com/santos-sanz/jev-audio-beeper/blob/main/index.ts#L1
验证标签call-site-verified
savka777/jev-use
README 自述：
证据https://github.com/savka777/jev-use/blob/main/Sources/JevCore/Decision.swift#L177
验证标签call-site-verified
schwarzschlyle/immune
README 自述：export TYPESAFE_API_KEY=... # TypeSafe Jev, which Immune asks about each call
证据https://github.com/schwarzschlyle/immune/blob/main/src/immune/sensing/jev.py#L15
验证标签call-site-verified
scienthoon/jev-ood-calibration
README 自述：Jev's probabilities are quantised to 0.01 and Choice/Score use the endpoints freely, so a correct answer sometimes receives exactly 0. A temperature cannot
证据https://github.com/scienthoon/jev-ood-calibration/blob/main/README.md#L39
验证标签readme-claim
sdras/system-one-explainer
README 自述：TypeSafe’s System One terminology and Jev quickstart are discussed in the final case study. Choice, Score, Noul, and distribution-derived confidence are specific API details.
证据https://github.com/sdras/system-one-explainer/blob/main/README.md#L31
验证标签readme-claim
SeeAPI/awesome-jev-use-cases
README 自述：
证据https://github.com/SeeAPI/awesome-jev-use-cases/blob/main/recipes/model-routing/run.py#L14
验证标签call-site-verified
sgzmd/rss-morning
README 自述：Jev classification on cheap RSS metadata (TypeSafe System One)
证据https://github.com/sgzmd/rss-morning/blob/main/README.md#L16
验证标签readme-claim
shaharia-lab/jev-cli
README 自述：
证据https://github.com/shaharia-lab/jev-cli/blob/main/crates/jev-cli/src/help/docs.rs#L813
验证标签call-site-verified
shamazharikh/qwen-rlcd
README 自述：Training uses RLCD (Reinforcement Learning for Calibrated Decisions). The objective is decisions plus calibrated probabilities. TypeSafe argues that RLHF hurts calibration through sycophancy and mode dropping.
证据https://github.com/shamazharikh/qwen-rlcd/blob/main/README.md#L160
验证标签readme-claim
shantanugoel/ask-jev-skill
README 自述：Hermes skill: call TypeSafe Jev (jev-latest) as a typed tiebreaker. Jev is not a chat model. It returns Choice / Score / Noul answers with probabilities so an agent can act or escalate.
证据https://github.com/shantanugoel/ask-jev-skill/blob/main/scripts/askjev.py#L15
验证标签call-site-verified
shantanugoel/mario-jev
README 自述：Jev receives structured RAM observations and answers focused questions about movement, starting a jump, and sustaining
证据https://github.com/shantanugoel/mario-jev/blob/main/src/mario_jev/policy.py#L3
验证标签call-site-verified
sharziki/semdecide
README 自述：Typed semantic decisions for Unix pipelines and CI, powered by TypeSafe AI Jev.
证据https://github.com/sharziki/semdecide/blob/main/src/reflex_guard/cli.py#L180
验证标签call-site-verified
shhivv/arc-cua
README 自述：not use decision models, JEV or TypeSafe; it is a standalone driver.
证据https://github.com/shhivv/arc-cua/blob/master/src/arc_cua/policies/typesafe.py#L26
验证标签call-site-verified
shinshin86/aituber-onair
README 自述：| typesafe | TypeSafe AI | jev-latest | https://api.typesafe.ai/v1/systemone |
证据https://github.com/shinshin86/aituber-onair/blob/main/packages/comment-intelligence/README.md#L291
验证标签readme-claim
shitianfang/jev-use
README 自述：hand the tasks that need no text output to Jev — faster steps, fewer
证据https://github.com/shitianfang/jev-use/blob/main/README.md#L7
验证标签readme-claim
shivendrasoni/vector-cache
README 自述：The default judge is TypeSafe Jev (vendor-quoted ~100 ms per call).
证据https://github.com/shivendrasoni/vector-cache/blob/main/README.md#L98
验证标签readme-claim
ShuhanSun/jev-oas-sentinel
README 自述：jev-oas-sentinel compares two OpenAPI documents in two layers:
证据https://github.com/ShuhanSun/jev-oas-sentinel/blob/main/src/jev_oas_sentinel/jev.py#L19
验证标签call-site-verified
sileod/tasksource
README 自述：(pick one option), score (an ordered scale) and noul (yes/no or a probability). This is the Jev / System One request format.
证据https://github.com/sileod/tasksource/blob/main/README.md#L76
验证标签readme-claim
socai-io/jev-social
README 自述：
证据https://github.com/socai-io/jev-social/blob/main/src/classifier.js#L222
验证标签call-site-verified
sonnylazuardi/superterminal
README 自述：The ranking comes from Jev,
证据https://github.com/sonnylazuardi/superterminal/blob/main/README.md#L58
验证标签readme-claim
sorrycc/typesafe-snake
README 自述：TypeSafe's Jev (a System One decision model, not an LLM) plays Snake in real time.
证据https://github.com/sorrycc/typesafe-snake/blob/master/server/index.ts#L2
验证标签call-site-verified
sosopop/jev_stock
README 自述：My current view is that JEV is more useful here as a disciplined reasoning layer than as a proven quantitative trading model.
证据https://github.com/sosopop/jev_stock/blob/main/jev_hk_predict.py#L29
验证标签call-site-verified
SoundBlaster/Jev4Mellea
README 自述：A small Python adapter that brings TypeSafe Jev's semantic checks into
证据https://github.com/SoundBlaster/Jev4Mellea/blob/main/src/mellea_jev/providers/categorical.py#L45
验证标签call-site-verified
SREGym/SREGym
README 自述：Jev is disabled by default. To enable it for Codex, set TYPESAFE_API_KEY and run:
证据https://github.com/SREGym/SREGym/blob/main/clients/jev/client.py#L14
验证标签call-site-verified
sriganesh/jevibe-check
README 自述：Usage shows request counts and estimated cost. Set token prices under Cost estimate, or use Clear stats to reset the counters. Your key and settings are kept. Check TypeSafe for actual charges.
证据https://github.com/sriganesh/jevibe-check/blob/main/extension/background.js#L133
验证标签call-site-verified
steven-shoemaker/hunch
README 自述：To reach Jev through a gateway your company already uses, pass gateway="openrouter" (reads OPENROUTER_API_KEY) or gateway="vercel" (reads AI_GATEWAY_API_KEY) to configure. Everything else works the same.
证据https://github.com/steven-shoemaker/hunch/blob/main/src/hunch/client.py#L37
验证标签call-site-verified
steven-shoemaker/hunch-js
README 自述：To reach Jev through a gateway your company already uses, pass gateway: "openrouter" (reads OPENROUTER_API_KEY) or gateway: "vercel" (reads AI_GATEWAY_API_KEY) to configure. Everything else works the same.
证据https://github.com/steven-shoemaker/hunch-js/blob/main/src/client.ts#L1
验证标签call-site-verified
STRML/omp-jevens-classifier
README 自述：Jev-judged permission checks for OMP's bash tool, and for eval payloads that spawn
证据https://github.com/STRML/omp-jevens-classifier/blob/main/jev.ts#L64
验证标签call-site-verified
sufianetaouil/every
README 自述：every needs a Jev API key. If you don't have one, join the hosted-tier waitlist:
证据https://github.com/sufianetaouil/every/blob/main/every/judge.py#L16
验证标签call-site-verified
sumleo/prompt2jev
README 自述：Natural language, an LLM prompt, or the code that runs one in. A Jev decision out: typed questions and a runnable script.
证据https://github.com/sumleo/prompt2jev/blob/main/skills/prompt2jev/scripts/prompt2jev.py#L27
验证标签call-site-verified
superagents-lab/jev-search
README 自述：Search the web in plain language. A decision model such as TypeSafe's Jev, Cloudflare's Clef or OpenAI's GPT-6 Luna chooses sources, time ranges and search terms, then ranks the results returned through Search1API. You g
证据https://github.com/superagents-lab/jev-search/blob/main/README.md#L5
验证标签readme-claim
supercorp-ai/supercov
README 自述：export TYPESAFE_API_KEY=... # get one at https://typesafe.ai
证据https://github.com/supercorp-ai/supercov/blob/main/crates/supercov-cli/src/quality/tests.rs#L2768
验证标签call-site-verified
szczyglis-dev/py-gpt
README 自述：Jev / System One (inline) - integrates TypeSafe AI Jev for fast, typed semantic decisions over structured state, including classification, routing, selection, verification, and scoring.
证据https://github.com/szczyglis-dev/py-gpt/blob/master/src/pygpt_net/plugin/jev/worker.py#L106
验证标签call-site-verified
tacticocc/Jevbridge
README 自述：ACP and MCP adapter that bridges TypeSafe Jev with any LLM.
证据https://github.com/tacticocc/Jevbridge/blob/main/src/types.ts#L111
验证标签call-site-verified
tamnd/kime
README 自述：The base URL defaults to http://127.0.0.1:8000 in place of https://api.typesafe.ai.
证据https://github.com/tamnd/kime/blob/main/ts/README.md#L22
验证标签readme-claim
tanghaojin/x-smart-tags
README 自述：扩展通过 Background Service Worker 直接调用 https://api.typesafe.ai/v1/systemone。
证据https://github.com/tanghaojin/x-smart-tags/blob/main/README.md#L42
验证标签readme-claim
tanxarx/awesome-jev
README 自述：A curated list of resources, open-source clones, integrations, and engineering playbooks for Jev — TypeSafe AI's non-generative "System 1" decision model.
证据https://github.com/tanxarx/awesome-jev/blob/main/README.md#L3
验证标签readme-claim
TauricResearch/TradingAgents
README 自述：export TYPESAFE_API_KEY=... # Jev social-post screening (optional)
证据https://github.com/TauricResearch/TradingAgents/blob/main/tradingagents/agents/post_screen.py#L68
验证标签call-site-verified
tbdavid2019/ai-hedge-fund-API
README 自述：Fallback Tier-3: TypeSafe Official Service (https://api.typesafe.ai/v1/systemone)
证据https://github.com/tbdavid2019/ai-hedge-fund-API/blob/main/README.md#L57
验证标签readme-claim
The-Vibe-Company/quivr
README 自述：That is far under Jev's budget of 32k tokens for the state plus the longest
证据https://github.com/The-Vibe-Company/quivr/blob/main/plugins/alerts/README.md#L292
验证标签readme-claim
TholeG/typesafe-chess
README 自述：Two virtual chess players, both powered by TypeSafe's System One model Jev.
证据https://github.com/TholeG/typesafe-chess/blob/main/README.md#L3
验证标签readme-claim
TianyuCodings/JevHarness
README 自述：| docs/ | Task authoring, a recorded Jev call, comparison video, and website screenshots |
证据https://github.com/TianyuCodings/JevHarness/blob/main/README.md#L212
验证标签readme-claim
timpratim/macbrow
README 自述：LiveKit Agents for the voice loop, TypeSafe's Jev
证据https://github.com/timpratim/macbrow/blob/main/README.md#L26
验证标签readme-claim
tinyhumansai/opencompany
README 自述：asked of Jev — TypeSafe's System One routing model — through the TinyHumans
证据https://github.com/tinyhumansai/opencompany/blob/main/README.md#L114
验证标签readme-claim
TKY-27/JevSlop
README 自述：Eight atomic detail Score questions plus a Jev overall Score and choice in one systemOne request.
证据https://github.com/TKY-27/JevSlop/blob/main/functions/api/evaluate.ts#L1
验证标签call-site-verified
togethercomputer/tev1
README 自述：Jev-inspired implementation and does not use Jev's answers as training labels.
证据https://github.com/togethercomputer/tev1/blob/main/README.md#L102
验证标签readme-claim
TokenTrim/jev-routing-experiment
README 自述：Jev is not a text model. It returns typed decisions (Choice / Score / Noul) with calibrated
证据https://github.com/TokenTrim/jev-routing-experiment/blob/main/jev_router/jev.py#L17
验证标签call-site-verified
tonone-ai/tonone
README 自述：| JEV_API_KEY or TYPESAFE_API_KEY | https://api.typesafe.ai/v1/systemone | jev-latest | jev |
证据https://github.com/tonone-ai/tonone/blob/main/lib/jev/README.md#L220
验证标签readme-claim
tonsky/tgadmin
README 自述：With Jev
证据https://github.com/tonsky/tgadmin/blob/main/README.md#L8
验证标签readme-claim
tpaulshippy/shady-town
README 自述：Decisions and thresholds stay in code. Built on TypeSafe's Jev via
证据https://github.com/tpaulshippy/shady-town/blob/main/README.md#L88
验证标签readme-claim
trancong12102/agentskills
README 自述：| answer audit (optional) | JEV_API_KEY | Create a key at console.typesafe.ai/keys, then export JEV_API_KEY=… (TYPESAFE_API_KEY also works). Turns on Jev's two extra checks in scripts/audit-answers.py. |
证据https://github.com/trancong12102/agentskills/blob/main/README.md#L38
验证标签readme-claim
trpc-group/trpc-agent-go
README 自述：For Jev, set JEV_BASE_URL=https://api.typesafe.ai, configure JEV_API_KEY
证据https://github.com/trpc-group/trpc-agent-go/blob/main/internal/systemone/README.md#L63
验证标签readme-claim
truespar/sentio
README 自述：provider = "jev"
证据https://github.com/truespar/sentio/blob/main/README.md#L998
验证标签readme-claim
turenlabs/turenos
README 自述：The detector uses one request with Noul questions for danger and confirmation, a Score for risk, and a Choice for category. Set TYPESAFE_ENDPOINT only if using a compatible HTTPS endpoint; plain HTTP is accepted only for
证据https://github.com/turenlabs/turenos/blob/main/command-guard/README.md#L41
验证标签readme-claim
tyler-dot-earth/patdown
README 自述：The judge is swappable. The default backend uses Effect Decision / DecisionModel with TypeSafe; rules and CLI commands use a provider-neutral interface.
证据https://github.com/tyler-dot-earth/patdown/blob/main/scripts/probe-jev-input-size.mjs#L25
验证标签call-site-verified
tylergibbs1/sift
README 自述：Load dist/ as an unpacked extension in chrome://extensions, then enter your TypeSafe API key on
证据https://github.com/tylergibbs1/sift/blob/main/scripts/smoke.ts#L74
验证标签call-site-verified
tylerjharden/ailerix
README 自述：POST /api/v1/systemone — Jev-compatible Choice / Score / Noul API
证据https://github.com/tylerjharden/ailerix/blob/main/src/components/playground-client.tsx#L45
验证标签call-site-verified
typakon4/jev-layer
README 自述：
证据https://github.com/typakon4/jev-layer/blob/main/src/providers/demo.mjs#L36
验证标签call-site-verified
TypeLLM/TypeLLM
README 自述：TypeLLM brings type-safe generation to existing autoregressive LLMs without changing their architecture or weights. Inspired by TypeSafe AI's Jev, it lets models retain their native thinking and free-form generation whil
证据https://github.com/TypeLLM/TypeLLM/blob/main/README.md#L48
验证标签readme-claim
TypeSafeAI/typesafe-playground
README 自述：Small experiments. Clear decisions. This is an independent community project, not an official TypeSafe AI product, production agent harness, or claim that a typed answer is correct.
证据https://github.com/TypeSafeAI/typesafe-playground/blob/main/lib/serverJev.ts#L52
验证标签call-site-verified
UditAkhourii/quicksilver
README 自述：Jev, TypeSafe's System One model. Jev returns typed
证据https://github.com/UditAkhourii/quicksilver/blob/main/README.md#L21
验证标签readme-claim
uehaj/sys1grep
README 自述：sys1grep: endpoint api.typesafe.ai/v1/systemone (default), model jev-latest (default)
证据https://github.com/uehaj/sys1grep/blob/main/README.md#L229
验证标签readme-claim
uezo/aiavatarkit
README 自述：Gates — Smart Turn, Namo Turn, Jev, OpenAI Decisions, LLM turn gate, filler-only, session hold, custom gates
证据https://github.com/uezo/aiavatarkit/blob/main/aiavatar/sts/vad/turn_end_gates/jev.py#L172
验证标签call-site-verified
unownone/jevsume
README 自述：ATS-friendly resume review powered by Jev (TypeSafe System One). The frontend extracts resume text the way a parser would, then a Cloudflare Worker runs typed JEV questions and composes a JevScore.
证据https://github.com/unownone/jevsume/blob/main/readme.md#L3
验证标签readme-claim
uu889/laya-opencv
README 自述：| TypeSafe 官方 | https://api.typesafe.ai | TypeSafe 控制台的密钥 |
证据https://github.com/uu889/laya-opencv/blob/main/README.md#L386
验证标签readme-claim
v-modal/awesome-jev-tools
README 自述：A curated list of public projects and developer patterns built on Jev, TypeSafe AI's System One model for typed decisions.
证据https://github.com/v-modal/awesome-jev-tools/blob/main/README.md#L7
验证标签readme-claim
valentynkit/awesome-jev-typesafe
README 自述：
证据https://github.com/valentynkit/awesome-jev-typesafe/blob/main/scripts/jev-client.mjs#L5
验证标签call-site-verified
valentynkit/jev-commit
README 自述：that the two agree. This does, in one call to Jev, before the commit lands.
证据https://github.com/valentynkit/jev-commit/blob/main/jev_commit/jev.py#L18
验证标签call-site-verified
valentynkit/jev-skip
README 自述：Gateway shim rather than the direct API. Re-measure on api.typesafe.ai before pinning
证据https://github.com/valentynkit/jev-skip/blob/main/README.md#L19
验证标签readme-claim
vinilana/live-jev
README 自述：TypeSafe's Jev (a "System One" decision model)
证据https://github.com/vinilana/live-jev/blob/main/README.md#L4
验证标签readme-claim
vnmoorthy/siege
README 自述：Built at CoreWeave Hacks: Agent Loops (Weights & Biases, AGI House, TypeSafe AI), San Francisco, September 13 to 14, 2026.
证据https://github.com/vnmoorthy/siege/blob/main/README.md#L248
验证标签readme-claim
w3cj/jev-chat
README 自述：Jev does not produce text. Every text argument, such as a search query, a task title or a city, is
证据https://github.com/w3cj/jev-chat/blob/main/apps/server/src/jev/client.ts#L1
验证标签call-site-verified
walidboulanouar/awesome-jev-use-cases
README 自述：
证据https://github.com/walidboulanouar/awesome-jev-use-cases/blob/main/README.md#L2
验证标签readme-claim
waynesutton/ask-jev-ai
README 自述：| Judge | Jev, TypeSafe System One, through the Convex AI Gateway Decisions endpoint (typesafe/jev-1.13) on a deployment token. No key. TypeSafe direct is the fallback when a key is set. | Decisions, docs.typesafe.ai |
证据https://github.com/waynesutton/ask-jev-ai/blob/main/README.md#L40
验证标签readme-claim
WeSecureYou/Jev-test
README 自述：The project works immediately without credentials. When JEV_API_KEY is absent, it logs a warning and uses deterministic, occupation-aware mock data with the same validated shape as a live Jev response.
证据https://github.com/WeSecureYou/Jev-test/blob/main/src/services/jevClient.ts#L9
验证标签call-site-verified
wfzyx/von
README 自述：von serve exposes /v1/systemone, byte-compatible with the TypeSafe specification. Anything that talks to Jev talks to Von by changing the base URL; the SDKs do the same via VON_BASE_URL.
证据https://github.com/wfzyx/von/blob/master/js/src/client.ts#L10
验证标签call-site-verified
whyashthakker/awesome-jev-use-cases
README 自述：git clone https://github.com/whyashthakker/awesome-jev-use-cases.git
证据https://github.com/whyashthakker/awesome-jev-use-cases/blob/main/shared/providers.js#L16
验证标签call-site-verified
Wizhill05/typesafe-image-diffusion
README 自述：Jev (via TypeSafe System One) is a judgment model. You give it state, ask multiple choice questions, and it picks answers with confidence scores. It classifies things. It does not draw things.
证据https://github.com/Wizhill05/typesafe-image-diffusion/blob/main/fix_synthesizer.py#L3
验证标签call-site-verified
wobsoriano/is-jeven
README 自述：Uses TypeSafe AI's Jev model under the hood to determine if a number is even.
证据https://github.com/wobsoriano/is-jeven/blob/main/index.js#L19
验证标签call-site-verified
wustep/jev-playground
README 自述：npm test # vitest: renderer invariants, Jev request shapes, proxy, MIDI round-trip, inbox/match stubs
证据https://github.com/wustep/jev-playground/blob/main/server/jevHandler.ts#L20
验证标签call-site-verified
Xiaooolong/vev
README 自述：the official Jev API (shape reference, costs a few cents)
证据https://github.com/Xiaooolong/vev/blob/main/conformance/README.md#L17
验证标签readme-claim
xinyao27/jevonian
README 自述：Jev picks the route for jevonian/auto — model and thinking depth in a single call.
证据https://github.com/xinyao27/jevonian/blob/main/internal/brain/types.go#L45
验证标签call-site-verified
xiwan/acp-bridge
README 自述：| Auto agent routing | Send a task to the virtual agent auto and let Jev (TypeSafe System One) pick the best-fit agent with a calibrated confidence score; falls back safely |
证据https://github.com/xiwan/acp-bridge/blob/main/README.md#L48
验证标签readme-claim
y0usaf/pi-jev
README 自述：The gate sends the working directory, the tool name, the last user message (first 1200 characters), and the tool's arguments to api.typesafe.ai. For write and edit those arguments contain file content. The output judge s
证据https://github.com/y0usaf/pi-jev/blob/main/README.md#L145
验证标签readme-claim
yairshy/decido
README 自述：ordinary Python. Start with Jev or supply your own decision provider.
证据https://github.com/yairshy/decido/blob/main/src/decido/providers/typesafe.py#L111
验证标签call-site-verified
yan-labs/yan-skills
README 自述：| jev | Typesafe(JEV / System One) | https://api.typesafe.ai/v1/systemone,Authorization: Bearer 鉴权(protocol: "typesafe-systemone",不是 Anthropic Messages/OpenAI 协议) | ⚠️ 不支持 run/run-many——它是结构化决策 API,不生成文本、不支持多轮工具调用,只能用下面「
证据https://github.com/yan-labs/yan-skills/blob/main/agent-fleet/README.md#L123
验证标签readme-claim
Yifan-Lan/awesome-jev-robustness
README 自述：entries · jev-1.13.0 in 111, jev-1.14.0 in 1, other decision models (Laya, Kev, Clef, Decision-1, OpenAI Decisions, open Jev-style checkpoints) in 31 · snapshot 2026-10-25 · numbers are the authors' own
证据https://github.com/Yifan-Lan/awesome-jev-robustness/blob/main/README.md#L5
验证标签readme-claim
Yinsongxu/LLM2Jev
README 自述：Turn your LLM / MLLM into a Jev-style decision model. Zero training required.
证据https://github.com/Yinsongxu/LLM2Jev/blob/main/demos/web/server.py#L19
验证标签call-site-verified
ZeroGold/call-coach-ai
README 自述：You need a TypeSafe API key. There are three ways to run Call Coach:
证据https://github.com/ZeroGold/call-coach-ai/blob/main/lib/coach.mjs#L13
验证标签call-site-verified
ZeroTang05/VideoAdGuard
README 自述：| TypeSafe | https://api.typesafe.ai/v1 | jev-latest | 官方接口 |
证据https://github.com/ZeroTang05/VideoAdGuard/blob/main/README.md#L138
验证标签readme-claim
zhengxuyu/litjev
README 自述：A reproduction of Jev: turn any Qwen model into a fast decision model.
证据https://github.com/zhengxuyu/litjev/blob/main/scripts/dump_prompt.py#L10
验证标签call-site-verified
zhuyansen/jev-search-rerank-eval
README 自述：Does a TypeSafe Jev rerank beat embedding search? Measured, with the judge-bias removed.
证据https://github.com/zhuyansen/jev-search-rerank-eval/blob/main/README.md#L3
验证标签readme-claim
zsavage8/padflow-jev-evals
README 自述：them. It exists so a decision model (TypeSafe Jev, or anything else) can be
证据https://github.com/zsavage8/padflow-jev-evals/blob/main/scripts/run_baseline.py#L7
验证标签call-site-verified

接着看这些页面

试用、文档、案例和价格是连着的。从这一页可以直接走到下一页。