好工具，依然都在这里。
保留 168 个资源与原始来源。按领域、关键词和类型查找，支持本地收藏、列表视图与 Markdown 导出。
Documentation
docs.typesafe.ai
概念、问题原语、API、设计模式及 SDK 使用指南。
Concepts, primitives, API, patterns, and SDK guides.
Introducing System One Models and Jev
typesafe.ai
System One 与 Jev 的发布文章，介绍模型接口、RLCD、官方公布的性能主张、演示和注意事项。
Launch post covering the model interface, RLCD, published performance claims, demos, and caveats.
typesafe-sdk-js
typesafe-ai/typesafe-sdk-js
官方 TypeScript / JavaScript SDK，支持自动推导答案类型。
Official TypeScript and JavaScript SDK with inferred answer types.
typesafe-sdk-python
typesafe-ai/typesafe-sdk-python
官方 Python SDK，同时提供同步与异步调用。
Official synchronous and asynchronous Python SDK.
system-one-adapter-python
typesafe-ai/system-one-adapter-python
可直接接入的适配器，用于对比 System One 接口与其他大语言模型服务商。
Drop-in adapter for comparing the System One interface with LLM providers.
skills
typesafe-ai/skills
用于构建和评估 System One 工作流的官方 Agent Skills。
Official agent skills for building and evaluating System One workflows.
Models and aliases
docs.typesafe.ai
模型版本标识、当前限制、价格，以及稳定版与预览版别名的区别。
Version IDs, current limits, pricing, and the distinction between stable and preview aliases.
Confidence
docs.typesafe.ai
区分 Choice / Score 的置信度与答案概率；Noul 不提供独立的置信度字段。
Distinguishes Choice/Score confidence from answer probability; Noul has no separate confidence field.
Jev 1.13 jaggedness
docs.typesafe.ai
官方能力局限说明：算术、日期、干扰信息、对抗性状态输入，以及输出结构上的不一致。
First-party limitations covering arithmetic, dates, distractors, adversarial state, and structural inconsistencies.
Cloudflare AI
developers.cloudflare.com
由服务商维护的 `typesafe/jev` 接入，接受状态信息和带类型的问题。
Provider-maintained `typesafe/jev` integration accepting state and typed questions.
OpenRouter
openrouter.ai
提供 `typesafe/jev-1.13`，同时提供指向版本会变化的 `typesafe/jev-latest` 别名。
Provider listing for `typesafe/jev-1.13`, alongside the moving `typesafe/jev-latest` alias.
Vercel AI Gateway
vercel.com
通过 AI SDK 实验性的 `evaluate` 接口调用 `typesafe-ai/jev`；其 Boolean 原语对应 TypeSafe 的 Noul。
`typesafe-ai/jev` through AI SDK's experimental `evaluate` interface; its Boolean primitive corresponds to TypeSafe's Noul.
advocaat
pithings/advocaat
轻量、类型安全的客户端，用来向 Jev 提出有关数据集的问题。
Small type-safe client for asking Jev questions about datasets.
jev
dannote/jev
围绕 GenServer 回复与模式匹配设计的 Elixir / OTP 客户端。
Elixir/OTP client designed around GenServer replies and pattern matching.
jev-axi
shiftynick/jev-axi
在命令行中完成选择、评级、检查、排序、分流与防护的 CLI 工具。
CLI for picking, rating, checking, ranking, triaging, and guarding from the shell.
jev-dsl
inanna-malick/jev-dsl
处于早期 Alpha 阶段的 Haskell DSL，用于编码带类型的问题包和解码答案；HTTP 传输需由调用方实现。
Early-alpha Haskell DSL that encodes typed question packets and decodes answers; HTTP transport is left to the caller.
jev-mcp
blakestone-x/jev-mcp
MCP 服务器，提供分类、评分、检查、匹配与筛选工具。
MCP server exposing classify, score, check, match, and screen tools.
jev-mcp
BYK/jev-mcp
以评估为先的 Jev MCP 服务器，返回带概率的类型化判断（noul、choice、score），而不是生成文本。
An eval-first MCP server for Jev, that returns typed judgments (noul, choice, score) with probabilities instead of generated text.
jev-shell-history
mrnugget/jev-shell-history
对已有 zsh 历史命令排序，提供行内补全；接受建议不会直接执行命令。
Ranks existing zsh history entries for inline completion; accepting a suggestion does not execute it.
Jevbridge
gamesonrblx/Jevbridge
ACP / MCP 适配器，让 Jev 与编程模型、对话模型协同使用。
ACP/MCP adapter for using Jev alongside coding and chat models.
jevclient
AboveColin/jevclient
异步 Python 客户端，用于提交带类型的 Jev 问题并获取概率。
Async Python client for typed Jev questions and probabilities.
jevr
simxnherrera/jevr
原生 R 客户端，经 TypeSafe 或 OpenRouter 提交带类型的问题，并获得不依赖具体服务商的统一答案。
Native R client for typed questions and provider-independent answers through TypeSafe or OpenRouter.
laravel-typesafe-jev
Butochnikov/laravel-typesafe-jev
Laravel 集成，支持类型化响应、异步请求和用于测试的模拟实现。
Laravel integration with typed responses, async requests, and testing fakes.
ruby_decision_model
obie/ruby_decision_model
Ruby 客户端，使用标准库传输层接入 TypeSafe 与 OpenRouter 的决策端点。
Ruby client with standard-library transport for TypeSafe and OpenRouter decision endpoints.
semdecide
sharziki/semdecide
为 Unix 管道与持续集成（CI）提供类型化语义决策。
Typed semantic decisions for Unix pipelines and CI.
typesafe-go
zhirschtritt/typesafe-go
符合 Go 语言惯用方式的 TypeSafe API SDK。
Idiomatic Go SDK for the TypeSafe API.
typesafe-mcp
itsmostafa/typesafe-mcp
MCP 连接器，让 Agent 可以调用 Jev 的决策能力。
MCP connector that gives agents access to Jev decisions.
typesafe-sdk-java
Premo-Cloud/typesafe-sdk-java
社区维护的 Java 17 客户端，支持 Choice、Score、Noul，并提供可选的 Spring Boot starter。
Community Java 17 client for Choice, Score, and Noul, with an optional Spring Boot starter.
zod-jev
jomatsu/zod-jev
将本地 Zod 结构校验与 Jev 语义校验组合使用。
Pairs local Zod shape validation with Jev semantic validation.
agent-router
nidhi-singh02/agent-router
尚未正式发布的 Herdr 集成：先按额度和策略筛选可用的编程模型，再由 Jev 排序。
Pre-release Herdr integration that filters eligible coding models by quota and policy before Jev ranks them.
blink
ellipsis-dev/blink
借助 Jev 引导的遍历器浏览文件名与目录名，为自然语言查询定位代码库路径。
Navigates file and directory names with Jev-guided walkers to find codebase paths for a natural-language query.
Canny
qkal/Canny
为编程 Agent 建立证据台账，质疑缺少依据的“已完成”声明。
Evidence ledger that challenges unsupported "done" claims from coding agents.
commit-miner
devanshbatham/commit-miner
将 Git diff 与提交信息归类为变更类型及候选安全修复 / CWE 标签，供进一步检查。
Classifies Git diffs and commit messages into change types and candidate security-fix/CWE labels for inspection.
foreman
thruwire/foreman
软件工厂的监督工具，利用 Jev 让编程 Agent 持续围绕任务执行。
Software-factory supervisor that uses Jev to keep coding agents on task.
is-malicious
luantak/is-malicious
在代码运行前，用 Jev 扫描源码、配置、构建及 CI 文件，报告可疑行为与涉及的代码行。
Scans source, configuration, build, and CI files with Jev, then reports suspicious behavior and implicated lines before the code is run.
jev-codex-router
0xNatoshi/jev-codex-router
逐轮为 Codex 选择模型、推理配置与速度模式。
Per-turn Codex model, reasoning, and speed-mode routing.
jev-guard
leepokai/jev-guard
对不同 Agent 的工具调用进行风险评分，输出允许、询问或拒绝的结果。
Cross-agent tool-call risk scoring with allow, ask, and deny outcomes.
jev-pref
doeixd/jev-pref
让 Jev 按 `jev-pref.json` 中的项目偏好检查代码变更，并将发现的问题反馈给编程 Agent。
Linter that has Jev check code changes against project preferences from `jev-pref.json` and feeds findings back to coding agents.
jev-review
devagrawal09/jev-review
分阶段的代码审查工作流，配有本地仪表盘。
Staged code-review workflow with a local dashboard.
jev-review MCP plugin
NiazMorshed2007/jev-review
面向编程 Agent、优先在本地运行的持续软件质量审查 MCP 插件。
Local-first continuous software-quality review for coding agents.
jev-router
gargpratyush/jev-router
封装现有 CLI，在 Claude Code 或 Codex 每个新轮次开始时选择模型。
Chooses a model for each fresh Claude Code or Codex turn while wrapping the existing CLI.
jevwire
Brainwires/jevwire
提供 MCP 工具、可嵌入的决策库，以及建议型或限制型 Claude Code hooks；模型判断不会授予宿主原生权限。
MCP tools, an embeddable decision library, and advisory or restrictive Claude Code hooks; judgments do not grant native permissions.
opencode-jev-orchestrator
aaronshaf/opencode-jev-orchestrator
保持 OpenCode 的父模型不变，由 Jev 判断任务难度，将较难的轮次委派给临时子 Agent。
Keeps an OpenCode parent model fixed and uses Jev difficulty judgments to delegate harder turns to temporary subagents.
perch
lakeday-org/perch
语义代码检查器，使用可配置的 Jev 问题评估代码单元。
Semantic code linter that evaluates code units against configurable Jev questions.
pi-jev
y0usaf/pi-jev
为 Pi 编程 Agent 提供经过测量的工具调用关卡和通用类型化决策层。
Measured tool-call gate and general typed decision layer for the Pi coding agent.
pi-warden
DevMortimer/pi-warden
Pi 扩展，用于判断规则遵守情况、风险操作、卡住的循环与完成声明；具体执行约束取决于 hook 和策略。
Pi extension that judges rule compliance, risky actions, stuck loops, and completion claims; enforcement depends on the hook and policy.
skillbox
kitze/skillbox
可自托管的 Skills 库；可选择使用 Jev，在已授权的目录中推荐相关技能。
Self-hosted skill library with optional Jev relevance recommendations over an authorized catalog.
skillranker
Dicklesworthstone/skillranker
Rust CLI，根据当前会话上下文为 Agent Skills 排序，也可以放弃作出选择。
Rust CLI that ranks agent skills against live session context and can abstain.
supercov
supercorp-ai/supercov
对源文件评分，帮助编程 Agent 确定代码质量改进的优先级。
Scores source files so coding agents can prioritize code-quality work.
wakegate
shitianfang/wakegate
实验性的唤醒关卡：由 Jev 判断定时器或传入事件是否值得恢复休眠 Agent 的大语言模型。代码仅在 Jev 有把握时跳过；用户消息、错误或达到跳过次数上限时始终唤醒。
Experimental gate where Jev decides whether a timer or incoming event is worth resuming a sleeping agent's LLM; code skips only when Jev is confident and always wakes on user messages, errors, and a skip limit.
fast-dev-compaction
leonaaardob/fast-dev-compaction
面向 Codex 的移植实现，在原生会话压缩前后恢复由 Jev 选中的逐字历史记录。
Codex port that restores Jev-selected verbatim history around native session compaction.
fast-jev-compaction
tamaratran/fast-jev-compaction
Claude Code 插件与库，对工具调用 / 结果对评分以决定删除或截断，同时原样保留选中的文本。
Claude Code plugin and library that score tool-call/result pairs for deletion or truncation while retaining selected text verbatim.
pi-fast-jev-compaction
joelhooks/pi-fast-jev-compaction
Pi 扩展，裁剪过时的工具历史；裁剪不足时，由 Pi 完成摘要式压缩。
Pi extension that prunes stale tool history and leaves summary compaction to Pi when pruning is insufficient.
pi-jev-compact
ilkerulusoy/pi-jev-compact
使用 Jev 为 Pi 进行选择性的上下文压缩，原样保留选中的文本。
Selective, verbatim context compaction for Pi using Jev model.
pi-jev-context
kevinpita/pi-jev-context
需主动启用的 Pi 扩展，从模型请求中筛除较旧的消息，同时保留原始会话历史。
Opt-in Pi extension that filters older messages from model requests while preserving the original session history.
winnow
GhalebDweikat/winnow
在工具结果进入 Claude Code 上下文前，先进行判断筛选。
Judges tool results before admitting them into Claude Code context.
yoshi
compozy/yoshi
实验性的 Claude Code / Codex 代理，用 Jev 裁剪请求上下文，同时保持工具调用协议的结构。
Experimental Claude Code/Codex proxy that uses Jev to prune request context while preserving tool-call protocol structure.
jev-browser-use
wy-coliney/jev-browser-use
Codex 浏览器 Skill：Jev 负责导航与目标选择，Codex 负责文字输入与结果验证。
Codex browser skill that uses Jev for navigation and target selection while Codex handles text entry and outcome verification.
jev-ultrafast
browser-use/jev-ultrafast
浏览器 Agent：由 Jev 选择操作及匹配的 DOM 目标，另一个大语言模型负责提供需要输入的文本。
Browser agent where Jev selects an operation and compatible DOM target, and a separate LLM supplies typed text.
jev-voice-browser
moritzkremb/jev-voice-browser
将部分语音转写映射为浏览器意图和已观察到的目标，再由代码决定执行、等待或询问。
Maps partial speech transcripts to browser intents and observed targets, with code deciding whether to act, wait, or ask.
mobile-jev
droidrun/mobile-jev
使用 Mobilerun 观察结果与受限 Jev 动作集的 Android Agent。执行模式会控制真实设备；公开的 Uber 演示在下单前停止。
Android agent using Mobilerun observations and bounded Jev actions; execute mode controls a real device, while the published Uber demo stops before booking.
typesafe-computer-use
awlevin/typesafe-computer-use
macOS 电脑操作实验，结合 OCR 与受限的 Jev 动作选择。
macOS computer-use experiment using OCR plus bounded Jev action selection.
duckdb-jev
colliber/duckdb-jev
DuckDB 扩展，将 Jev 判断作为 SQL 值返回；返回类型由声明的判断标准决定。
DuckDB extension that exposes Jev judgments as SQL values with return types derived from the declared criteria.
HA-Jev
AboveColin/HA-Jev
Home Assistant 集成，将类型化答案提供为传感器、自动化动作和 Assist 对话 Agent。
Home Assistant integration exposing typed answers as sensors, automation actions, and an Assist conversation agent.
jev-curate
AkashPriyadarshii/jev-curate
为 Parquet 与 JSONL 数据集提供流式筛选与评分。
Streaming filter and scorer for Parquet and JSONL datasets.
jev-reviewer
choxos/jev-reviewer
研究文档提取辅助工具：Jev 选择并验证原始行，提取逐字引文。结果需要人工复核，不构成临床决策。
Research-document extraction aid where Jev selects and verifies source lines for verbatim quotes; findings require human review and are not clinical decisions.
jev-search
superagents-lab/jev-search
使用 Jev 选择搜索来源并对 Search1API 结果排序，返回来源链接与摘要片段。
Uses Jev to select search sources and rank Search1API results, returning source links and snippets.
jev-trader
jarrodwatts/jev-trader
Kuru / Monad 交易实验，可选用 Jev 作出买卖决策。默认使用模拟模型，无私钥时进行试运行；配置实际执行后可以下真实订单。
Kuru/Monad trading experiment with optional Jev buy/sell decisions; defaults to a mock model and dry-runs without a private key, but configured execution can place real orders.
jev-tree
reachjalil/jev-tree
在超过 Jev 单次直接选项上限的分类体系中，递归完成选择。
Recursive choice over taxonomies larger than Jev's direct option limit.
jevlogs
reachjalil/jevlogs
在调用更昂贵的分析之前，先对 OpenTelemetry 日志进行分流。
OpenTelemetry log triage before more expensive analysis.
jevql
kylemclaren/jevql
面向原生 Postgres 的类 psql 客户端及 Go / TypeScript / Python SDK。普通 SQL 先在服务器执行，再由 Jev 对各行作 Noul、Choice、Score 判断，客户端据此筛选、排序或分组。
psql-shaped client and Go/TypeScript/Python SDKs for vanilla Postgres where Jev makes Noul, Choice, and Score judgements about individual table rows after the plain SQL has run on the server, and the client applies the resulting filter, sort, or group.
jevsql
EugeneBoondock/jevsql
使用自然语言判断条件，以类似 SQL 的方式进行筛选、排序、分类与评分。
SQL-like filtering, ranking, classification, and scoring with natural-language predicates.
llama-index-jev
WiktorB2004/llama-index-jev
基于 Jev Score / Choice 的 LlamaIndex 重排序器与选择器，支持配置置信度处理方式。
LlamaIndex reranker and selector using Jev Score and Choice answers, with configurable confidence handling.
n8n-nodes-typesafe-jev
n3ndor/n8n-nodes-typesafe-jev
社区 n8n 节点，可针对工作流状态一次提出多个带类型的问题。
Community n8n node for asking multiple typed questions over workflow state.
pg-jev
realZachi/pg-jev
PostgreSQL 扩展，对数据表行提出语义问题。
PostgreSQL extension for semantic questions over table rows.
pg_typesafe
giuliosmall/pg_typesafe
Pre-alpha 阶段的 PostgreSQL C 扩展，通过 SQL 暴露 Choice、Noul、Score 与批量判断能力。
Pre-alpha PostgreSQL C extension exposing Choice, Noul, Score, and batched judgments from SQL.
tax-doc-classifier
kyotofin/tax-doc-classifier
带置信度关卡的文档分类器，将含文本的 PDF 页面归为候选 IRS 表单与页面类型。用于文档分流而非税务建议；扫描页需要 OCR。
Classifies text-bearing PDF pages into IRS form and page-kind candidates with a confidence gate; document triage, not tax advice, and scanned pages need OCR.
tiershift
iamvatsalpatel/tiershift
在明确的策略边界内，为 TypeScript 和 Python 应用进行模型路由。
Policy-bounded model routing for TypeScript and Python.
typesafe-jev-workflow
GiesN/typesafe-jev-workflow
基于 LangGraph 的邮件意图工作流，使用 Jev 的类型化选择。
LangGraph email-intent workflow using a typed Jev choice.
heist-one
AbdelStark/heist-one
浏览器潜行游戏：Jev 判断守卫行为，确定性代码负责管理游戏世界。
Browser stealth game where Jev judges guards while deterministic code owns the world.
jev-drone
RomanSlack/jev-drone
MuJoCo 四旋翼无人机模拟：Jev 根据处理后的相机观察作出较低频的战术判断，确定性代码负责飞行控制。
Simulated MuJoCo quadrotor with Jev making slower tactical judgments from processed camera observations; deterministic code controls flight.
jev-experiments
dabit3/jev-experiments
一组可检查实现的 Jev 演示，包括脚本化客服对话中的类型化意图识别、升级处理与建议回复决策。
Collection of inspectable Jev demos, including scripted support conversations with typed intent, escalation, and suggested-response decisions.
JevPilot
standardagents/jevpilot
Three.js 驾驶模拟：Jev 在候选路径与速度中选择，本地代码负责车辆动力学与几何计算。
Three.js driving simulation where Jev chooses among candidate paths and speeds while local code handles vehicle dynamics and geometry.
JevScape
Skyvern-AI/jevscape
基于 RuneBench 的 RuneScape 测试框架，将 Jev 选择映射为受限游戏动作集，并记录逐 tick 结果。
RuneBench-based RuneScape harness that maps Jev choices to a bounded game-action catalog and records tick-level results.
killmyidea
monteduro/killmyidea
创业点子评估器，在放弃（kill）、修改（fix）或推出（ship）之间作选择。
Startup-idea evaluator that chooses kill, fix, or ship.
tsai-sc
phyous/tsai-sc
通过 Jev 控制原版《星际争霸》共享软件版本，并记录动作概率。
Original StarCraft shareware controlled with recorded Jev action probabilities.
typesafe-mario
fhshaik/typesafe-mario
《超级马里奥兄弟》Agent，根据结构化模拟器状态选择动作。
Super Mario Bros. agent choosing actions from structured emulator state.
typesafe-snake
sorrycc/typesafe-snake
贪吃蛇自动玩家，每个 tick 作一次类型化决策，合法动作由代码生成。
Snake autoplayer with one typed decision per tick and code-generated legal moves.
jevmeter
ChetasLua/jevmeter
为视频中的每句话评分，并将结果渲染为叠加图层。
Scores every sentence in a video and renders the result as an overlay.
Jevthoven
cocktailpeanut/jevthoven
符号音乐工作室：Jev 选择创作方案、乐器和小节模式，代码生成可编辑音乐与 MIDI。
Symbolic-music studio where Jev chooses plans, instruments, and bar patterns, and code renders editable music and MIDI.
SlidePilot
harshil1712/slidepilot
实验性的 Slidev 控制器，根据语音转写判断当前幻灯片是否讲完，并保留确定性检查与手动翻页。
Experimental Slidev controller that judges speech transcripts for slide completion, with deterministic checks and manual navigation.
Sponsor Skip
trungdq88/youtube-sponsor-detection
从 YouTube 字幕或音频转写中识别赞助口播；时间戳和播放跳过由代码负责。
Finds sponsor reads in YouTube transcripts or transcribed audio while code owns timestamps and playback skipping.
unclutter
kitze/unclutter
浏览器扩展，使用 Jev 识别页面杂乱内容，保存可复用、可撤销的隐藏规则。
Browser extension that uses Jev to identify page clutter and saves reusable, reversible hiding rules.
Vibe Check for X
RafalWilinski/vibecheck
Chrome 扩展，在发布前为推文草稿和回复上下文评分；可选的媒体描述由独立视觉模型提供。
Chrome extension that scores draft posts and reply context before posting; optional media descriptions come from a separate vision model.
Jev Visual
hr98w/jev-visual
面向教学的 MLX / Qwen 视觉语言实验，在候选评分问题间共享图像上下文；输出概率不是经过校准的正确率估计。
Educational MLX/Qwen vision-language experiment sharing image context across candidate-scoring questions; its probabilities are not calibrated correctness estimates.
Jevlike
vinnylarouge/jevlike
为可变候选集合提供可训练的编码器与选项注意力头，另有独立的视觉游戏实验。
Trainable encoder and option-attention head for variable candidate sets, with separate visual game experiments.
LitJev
zhengxuyu/litjev
项目自述为 Jev 复现：无需训练、不生成答案文本，将 Qwen 模型转为快速决策模型，并提供相同的 `/v1/systemone` 接口结构（Choice、Score、Noul）。
Reproduction of Jev that turns any Qwen model into a fast decision model, serving the same `/v1/systemone` schema (Choice, Score, Noul) with no training and no generated answer text.
jevmlx
bnsd55/jevmlx
为 Apple Silicon 上的 MLX 模型提供 Jev 风格的并行受限决策。
Jev-style parallel constrained decisions for MLX models on Apple Silicon.
kev
jaredpalmer/kev
基于 Qwen2.5-0.5B 的适配器与决策头，提供训练代码、已发布权重以及并行的类型化问题推理。
Qwen2.5-0.5B adapter and decision head with training code, released weights, and parallel typed-question inference.
NanoJev
TianyuCodings/NanoJev
小型并行决策模型，支持动态候选项、训练流水线；已记录的游戏对比包含共享的代码规划逻辑。
Small parallel-decision model with dynamic candidates, a training pipeline, and recorded game comparisons that include shared code planning.
openjev
zhihz/openjev
根据上下文、问题与候选答案，在本地进行双语概率决策。
Local bilingual probability decisions from context, questions, and candidate answers.
OpenJev (DiffusionGemma)
razorback16/openjev
基于 DiffusionGemma / vLLM 的独立 Jev 兼容服务器；文档中的部署方案需要自定义 vLLM 补丁。
Independent Jev-compatible server over DiffusionGemma/vLLM; the documented setup requires custom vLLM patches.
openjev-sglang
ekzhang/openjev-sglang
由开放模型与仅预填充（prefill-only）推理支持的 Jev 兼容 API 端点。
Jev-compatible API endpoint backed by open models and prefill-only inference.
openvons
genai-craft/openvons
面向有限选项的开放决策层，支持文本、图像与日语语音指令。
Open decision layer for finite options across text, images, and Japanese voice commands.
parallelConstraintDecoding
stephanj/parallelConstraintDecoding
使用 Java 与 llama.cpp 探索并行受限解码的实验。
Java and llama.cpp experiments in parallel constrained decoding.
reflex
kshetrajna12/reflex
开放模型决策引擎，采用共享状态推理、隔离的问题分支，并提供 WebGPU 演示；浏览器与 Python 配置不同。
Open-model decision engine with shared-state inference, isolated question branches, and a WebGPU demo; browser and Python configurations differ.
SemIf
TheoLeeCJ/SemIf
原名 OpenJev：研究如何从冻结的开放模型中读取类型化选项，包含共享前缀实验及 WebGPU 演示。
Formerly OpenJev: an independent study of typed option readout from frozen open models, with shared-prefix experiments and a WebGPU demo.
Simple Jev
featherless-ai/simple-jev
将兼容开放模型的 logits 转为类型化决策，无需单独训练分类头；模型兼容性有限制。
Transforms compatible open-model logits into typed decisions without a separately trained classifier head; model compatibility is constrained.
Verdict-open-jev
Heman10x-NGU/Verdict-open-jev
基于 ModernBERT 的决策引擎，提供经过校准的不确定性与 WebGPU 体验环境。
ModernBERT decision engine with calibrated uncertainty and a WebGPU playground.
Janus
FirasSX914/Janus
测量何时应使用 Jev、何时应使用其他模型，并据此进行路由。
Measures when to use Jev versus other models and routes accordingly.
jev-behavior-study
RINNECODER/jev-behavior-study
对 Jev 1.13.0 问题表述敏感性与失败模式的独立合成任务研究，附原始响应和离线报告检查。
Independent synthetic-task study of Jev 1.13.0 framing sensitivity and failures, with raw responses and offline report checks.
jev-benchmarks
AbdelStark/jev-benchmarks
用于校准、选择性风险与延迟的可复现评估。
Reproducible evaluation for calibration, selective risk, and latency.
jev-eval
4esv/jev-eval
在三个带标签的分类任务上，独立对比 Jev 与 GPT-5.6 Terra，报告准确率、校准、延迟和成本。
Independent Jev versus GPT-5.6 Terra comparison on three labeled classification tasks, reporting accuracy, calibration, latency, and cost.
jev-eval-agent
vinilana/jev-eval-agent
在包含 100 个模拟工具的个人助理测试框架中，对比大语言模型工具选择与 Jev 路由。
Compares LLM tool selection with Jev routing in a personal-assistant harness containing 100 mocked tools.
jev-korean-benchmark
mahlernim/jev-korean-benchmark
韩文 / 英文小样本研究，保留模型响应，包含医学文本问题；不构成临床验证。
Small Korean/English sample study with recorded responses, including medical-text questions; not a clinical validation.
jev-search-rerank-eval
zhuyansen/jev-search-rerank-eval
中英文检索评估，将 Jev 重排序与词法、向量嵌入及融合基线对比，并分析评判者循环论证问题。
Chinese/English retrieval evaluation comparing Jev reranking with lexical, embedding, and fusion baselines, including judge-circularity analysis.
jevcal
abhixhek/jevcal
基于带标签的数据拟合置信度阈值，并检查其漂移。
Fits and drift-checks confidence thresholds against labeled data.
typesafe-ai-benchmark
iammrduncan/typesafe-ai-benchmark
模拟 System One 输出结构的大语言模型网关，用于对比研究。
LLM gateway that mimics the System One output shape for comparison work.
Building with Jev
dbreunig/building-with-jev-skill
社区 Agent Skill，介绍问题设计、状态准备、置信度阈值和决策调试。
Community agent skill covering question design, state preparation, confidence thresholds, and debugging decisions.
Classifying RAG passages
docs.typesafe.ai
官方示例：在将检索片段交给回答模型之前，先判断片段内容。
Official example of judging retrieved passages before passing them to an answering model.
Date extraction
docs.typesafe.ai
官方模式：将类型化提取与代码中的日期校验、日期运算分开。
Official pattern separating typed extraction from date validation and arithmetic in code.
Double-checking citations
docs.typesafe.ai
官方示例：检查来源上下文是否支持某项陈述。
Official example of checking whether source context supports a claim.
Parallel questions
docs.typesafe.ai
官方完整示例：在一次请求中，针对共享状态并行评估多个问题。
Official worked example of evaluating many questions over shared state in one request.
Skill suggestion
docs.typesafe.ai
官方两阶段工作流：先选择 Skill，也可以拒绝整个候选列表。
Official two-stage workflow that selects a skill and can reject the shortlist.
awesome-jev-by-typesafe
Anil-matcha/awesome-jev-by-typesafe
附有证据的用例、模式、提示词与起步代码。
Evidence-backed use cases, patterns, prompts, and starter code.
awesome-jev
hellogumbo/awesome-jev
大型社区导航目录，并提供可搜索的配套网站。
Large community directory with a searchable companion site.
yibie/awesome-jev
yibie/awesome-jev
按决策领域组织的高信息密度实践指南。
High-signal field guide organized by decision domain.
awesome-typesafe
AbdelStark/awesome-typesafe
覆盖更广泛 TypeSafe 与 System One 生态的资源清单。
Broader TypeSafe and System One ecosystem list.
OmniJev/awesome-jev
OmniJev/awesome-jev
收录论文、开放复现、独立评估与技术脉络。
Papers, open reproductions, independent evaluations, and technical lineage.
awesome-jev-typesafe
valentynkit/awesome-jev-typesafe
采用 CC0、通过 awesome-lint 检查，按准备安装的内容分类，并简要说明动手前需要了解的限制。
CC0, awesome-lint clean, sorted by what you would install, with a short know-before-you-build section on the limits.
jev-go
Gaurav-Gosain/jev-go
面向 TypeSafe System One API 与 Jev 的 Go 客户端，返回类型化判断和概率，而不是生成文本；原清单对“校准概率”的描述不代表本站验证。
Go client for TypeSafe's System One API and its model Jev: typed judgments and calibrated probabilities instead of generated text
jev-go
Stumble/jev-go
独立维护的 Go SDK，用于调用 TypeSafe AI 的 Jev / System One API。
An independent Go SDK for TypeSafe AI's Jev / System One API.
swift-typesafe
ainame/swift-typesafe
Swift 6.4 SDK，按原清单说明对齐 Python SDK 0.6.0 的 API；接入前请检查当前版本兼容性。
Swift 6.4 SDK for TypeSafe AI, following the Python SDK's 0.6.0 API.
typesafe_sdk
nshkrdotcom/typesafe_sdk
用于 TypeSafe AI 及其首个 System One 模型 Jev 的 Elixir SDK。
typesafe_sdk is the Elixir SDK for TypeSafe AI and its first System One model, Jev.
typesafe_sdk_ex
vinnie357/typesafe_sdk_ex
基于 Req 的 TypeSafe AI Elixir SDK。
Typesafe AI SDK in Elixir using Req
typesafe-ai
Twister915/typesafe-ai
将 TypeSafe System One 评估 API 接入 Rust，把小型、类型化判断嵌入常规应用代码。
typesafe-ai brings TypeSafe's System One evaluation API into Rust as small, typed judgments that fit inside ordinary application code.
typesafe-ai-rs
gilljon/typesafe-ai-rs
独立维护的 TypeSafe AI System One API Rust 客户端，仓库为 gilljon/typesafe-ai-rs。
An independent Rust client for the TypeSafe AI System One API, maintained at gilljon/typesafe-ai-rs.
typesafe-rs
AbdelStark/typesafe-rs
针对同一状态提交带名称的 Noul、Choice、Score 问题，并为每个问题返回类型化答案。
Evaluate a state against named questions (noul, choice, score) and get one typed answer per question.
typesafe-sdk
joshmn/typesafe-sdk
TypeSafe System One API 的 Ruby 客户端。
A Ruby client for the TypeSafe System One API.
typesafe-sdk-go
Tangerg/typesafe-sdk-go
TypeSafe AI API 的 Go SDK：输入类型化问题，返回概率分布。
Go SDK for the TypeSafe AI API — typed questions in, probability distributions out.
typesafe-sdk-php
Butochnikov/typesafe-sdk-php
TypeSafe AI System One API 的 PHP 客户端。
PHP client for the TypeSafe AI System One API.
typesafeai-dotnet-sdk
saibimajdi/typesafeai-dotnet-sdk
社区维护的 .NET SDK，支持类型化 Noul、Choice、Score 问题与结构化答案；并非 TypeSafe 官方项目。原清单称其提供置信信息，但 Noul 本身没有独立 confidence 字段。
Community .NET SDK for the TypeSafe AI System One API — typed noul, choice, and score questions with structured, confidence-scored answers. Not affiliated with TypeSafe AI.
zio-typesafe-ai
jamesward/zio-typesafe-ai
Scala 3 / ZIO 库，用于调用 Jev / System One API，围绕给定状态回答类型化、原子问题，而不是生成文本。
A Scala 3 / ZIO library for TypeSafe AI's Jev / System One API — a "System One model" that answers typed, atomic questions about a piece of state instead of generating text.
ha-conversation-jev
luxus/ha-conversation-jev
Home Assistant 自定义对话 Agent：用 Jev 对话语分类，再调用灯光服务（v0 快速路径），或转交 SpaceXAI Grok 对话 Agent。
Home Assistant custom conversation agent: Jev (TypeSafe System One, jev-latest) classifies an utterance, then either calls a light service (v0 fast path) or hands off to the SpaceXAI Grok conversation agent.
hermes-jev
keeltrace/hermes-jev
为路由、排序、验证、多问题评估、工具调用把关及上下文价值判断提供狭窄、类型化决策，避免让主生成模型临场编写答案。
It is built for narrow, typed judgments that should not require the main generative model to improvise an answer: routing, ranking, verification, multi-question assessment, tool gating, and context-value decisions.
Jev MCP
jkudish/jev-mcp
使用 Jev 进行主张核验、内容筛查及语义候选排序的 MCP 服务。
MCP server that uses TypeSafe Jev for claim verification, content screening, and semantic candidate ranking.
jev-browser
Ying-Kai-Liao/jev-browser
由大语言模型负责规划、Jev 负责判断的浏览器自动化项目，提供库、CLI 和 MCP 服务。
Browser automation where an LLM plans and Jev (Typesafe System One) decides. Library, CLI and MCP server.
jev-ego
romaluev/jev-ego
面向 ego lite 的 TypeScript 浏览器 Agent，采用 Jev Ultrafast 的索引动作空间，不依赖 Chrome、Playwright 或 Browser Harness。
TypeScript browser agent for ego lite — Jev Ultrafast's indexed action space, without Chrome, Playwright, or Browser Harness.
jev-starter
hamakyo/jev-starter
在 Jev 上构建类型化、策略驱动的决策流程，为 TypeScript 应用提供置信度路由、回退、评估和 RAG 模式。
Typed, policy-driven decision workflows on top of TypeSafe AI Jev: confidence routing, fallbacks, evaluation, and RAG patterns for TypeScript apps.
ruby_llm-typesafe
kieranklaassen/ruby_llm-typesafe
为 RubyLLM 2 提供 TypeSafe 结构化输出的服务商适配。
TypeSafe structured-output provider for RubyLLM 2
typesafe-ai-rails
GenieRobot/typesafe-ai-rails
基于社区 typesafe-sdk Ruby gem 构建的 Rails 集成，用于接入 TypeSafe AI System One API。
Community Rails integration for TypeSafe AI's System One API, built on the community typesafe-sdk Ruby gem.
Jev Review Action
fatwang2/jev-review-action
可配置的 GitHub Action：用 Jev 审查资源目录投稿并分类拉取请求，附带证据链接，由模板生成评论；最终合并由维护者决定。
Configurable GitHub Action using Jev to review directory submissions and classify pull requests, with evidence links and template-generated comments.
jev-code
devagrawal09/jev-code
供编程 Agent 委派大量判断类工作的命令行工具集。
jev-code is a command-line toolkit that coding agents can delegate judgment-heavy work to.
jev-git
AkashPriyadarshii/jev-git
Git pre-commit / pre-push 检查工具，用 Jev 筛查暂存差异中的密钥与破坏性命令。原清单称其为“亚秒级”，本站未测量；不能替代其他安全检查。
Sub-second Git pre-commit and pre-push reflex gate that screens staged diffs for secrets and destructive commands using TypeSafe Jev.
jev-scout
AkashPriyadarshii/jev-scout
用 Jev 评分发现开源仓库和 Rust crate。原清单使用“零幻觉”表述，本站保留原文供核对，不把这一宣传语作为可靠性保证。
Zero-hallucination open-source repo and crate scout powered by TypeSafe AI Jev System One scoring.
jev-seo
AkashPriyadarshii/jev-seo
基于 DuckDuckGo 与 Jev 的 SEO / 生成式引擎优化（GEO）搜索雷达，提供面向 Agent 的 CLI 和 MCP 服务。原清单称“零成本”，实际 API 与服务费用需另行确认。
Zero-cost, agent-first SEO & Generative Engine Optimization (GEO) search radar CLI suite and MCP server powered by DuckDuckGo and TypeSafe Jev System One.
jev-superpowers
AkashPriyadarshii/jev-superpowers
面向 AI 编程 Agent 的系统化开发框架，加入 Jev 类型化决策、软件包核验和完成检查关卡。原清单对核验使用“零幻觉”描述，本站未验证该保证。
Systematic software development framework for AI coding agents upgraded with TypeSafe Jev System One typed decisions, zero-hallucination package vetting, and completion gates.
JevSeek
morcoan/JevSeek
桌面与 CLI 编程 Agent，根据用户意图和已记录的执行结果，让 Jev 选择下一项工具。
Desktop and CLI coding agent that uses Jev to select the next tool from user intent and recorded execution results.
Leanest
baronunread/leanest
本地优先的测试选择器，用 Jev 判断一次代码变更可以跳过哪些测试，附带 GitHub Action 以接入 CI；跳过测试的判断不构成正确性保证。
Local-first test selector that uses TypeSafe Jev semantic judgments to decide which tests are safe to skip for a given code change, dropping straight into CI via its bundled GitHub Action.
TypeSafe AI Playground
markjaquith/typesafe-ai-playground
Rust 命令行实验集，展示受保护健康信息（PHI）检测、代码注释审查、语气分析及分类，不是临床验证工具。
Rust CLI with TypeSafe Jev experiments for PHI detection, code-comment review, tone analysis, and classification.
hev reranker
hev/reranker
用 Jev Noul 判断给候选文档的查询相关性评分，再排序或过滤结果的 Python 库。
Python library using Jev Noul judgments to score candidate documents for query relevance, then sort or filter the results.
neo4jev
jexp/neo4jev
Neo4j 图导航演示，在束搜索过程中让 Jev 选择关系并检查目标是否满足。
Neo4j graph navigation demo that uses TypeSafe Jev to select relationships and check goals during beam search.
Jev-Trades
zadescoxp/Jev-Trades
Next.js 仪表盘，结合实时加密货币市场数据与 TypeSafe 驱动的模拟交易（paper trading）；不是投资建议或收益承诺。
A Next.js dashboard for live crypto market data and TypeSafe-powered paper trading.
typesafe-ai-playground
BunsDev/typesafe-ai-playground
社区交互实验场：编辑分类实验、对比 A/B 输入、路由对话、提取文档字段、检查代码策略，并探索基于类型化模型输出的游戏与模拟。
A community playground for TypeSafe AI's Jev: edit classification experiments, compare A/B inputs, route conversations, extract document fields, inspect code-policy decisions, and explore games and simulations built around typed model outputs.
UI Generator Instinct Jev
joevidev/ui-generator-instinct-jev
根据用户描述的界面需求，由 Jev 回答 Choice / Noul / Score 问题，选择并配置既有 shadcn/ui 组件或页面区块；这是演示应用，不生成代码或文案。
A demo app where the user describes a UI case in free text and Jev (TypeSafe's System One model) answers typed Choice/Noul/Score questions to pick and configure a real shadcn/ui component or page block, never generating code or copy.