BetterZflyee/dsh-jev-adapter
Use the Jev (System One) decision-model paradigm with any OpenAI-compatible LLM — no TypeSafe key required. A jev_decide tool for DeepSeek Harness (dsh).
Project Overview项目介绍
dsh-jev-adapter is a native plugin for DeepSeek Harness (DSH) that registers a single jev_decide tool on the host, packaging the Jev (System One) decision-model paradigm so any OpenAI-compatible chat model can produce typed, probability-distributed answers. It is installed via dsh plugin --profile <profile> add github:BetterZflyee/dsh-jev-adapter from GitHub, or as dsh-jev-adapter from npm once published, and configured either through a plugin row in cordis.patch.yml or through environment variables such as JEV_OPENAI_API_KEY, JEV_BASE_URL, and JEV_MODEL, after which DSH is restarted to load the adapter.
The intended user is anyone who needs to turn ambiguous judgements into structured numbers — ticket triage, urgency scoring, risk classification, A/B picks, and similar routing tasks. The agent bundles every judgement into one call: choice over up to 255 options, score over 2–10 ordered levels, or boolean statements, and receives the picked option, fractional score, or P(true) together with full per-level probabilities and a confidence score. Callers then route on confidence — ≥0.85 automatic, 0.5–0.85 drafted for human confirmation, below 0.5 escalated — with thresholds tuned to the action's risk profile.
Dependencies and caveats are explicit. The default openai channel works with DeepSeek's own API, Ollama, vLLM, OpenRouter, SiliconFlow, LM Studio, or an air-gapped internal gateway; the typesafe channel needs JEV_TYPESAFE_API_KEY plus reachability to api.typesafe.ai. On the openai channel the probabilities are the model's self-reported estimates, not mathematically calibrated, and every result carries this caveat; retries handle 429/5xx with exponential backoff (default 3), renormalisation fixes non-summing distributions, thinking models that answer in reasoning_content are supported, and the project is released under MIT.
dsh-jev-adapter 是面向 DeepSeek Harness (DSH) 的原生插件,把 Jev(System One)决策模型范式以 jev_decide 工具注册到 Agent 中,可由 DSH 直接加载调用。安装方式为 dsh plugin --profile <profile> add github:BetterZflyee/dsh-jev-adapter,发布后也可通过 npm 名 dsh-jev-adapter 安装,配置写入 cordis.patch.yml 的插件条目或通过 JEV_OPENAI_API_KEY 等环境变量注入,重启 DSH 后生效。
该插件适合需要把模糊判断转成结构化概率分布的用户,典型场景是工单分类、风险分级、A/B 抉择等。Agent 把所有判断合并成一次类型化问题调用(choice ≤255 项 / score 2–10 级 / boolean),返回完整概率与置信度,用户据此路由:≥0.85 自动执行,0.5–0.85 草稿待人确认,<0.5 转人工。它与 zhangxaochen/dsh-jev 互补:后者走官方 TypeSafe API 并带安全护栏,本插件走任意 OpenAI 兼容端点。
依赖与限制方面,openai 通道支持 DeepSeek、Ollama、vLLM、OpenRouter、硅基流动、LM Studio 或内网网关;typesafe 通道需要 JEV_TYPESAFE_API_KEY 与 api.typesafe.ai 连通。openai 通道的概率为模型自述估计而非数学校准,每个结果都带提示;429/5xx 走指数退避,默认 3 次;思考模型 reasoning_content 已处理;许可证 MIT。
请帮我安装这个 DSH 插件。安装前先完成【兼容性检查 + 安全性检查】,检查通过再动手。
插件:dsh-jev-adapter(BetterZflyee/dsh-jev-adapter)
仓库:https://github.com/BetterZflyee/dsh-jev-adapter
本站详情页:https://www.yhbd.top/plugins/betterzflyee-dsh-jev-adapter/
本站登记:类型 plugin · 归类 原生 DSH 插件 · 许可证 MIT · ⭐ 4 · 最近提交 2026-09-21 · 主语言 JavaScript
按下面顺序执行,每步先把结论告诉我,再进入下一步:
【1 兼容性检查】
① 我这边:DSH 版本、Node 版本、操作系统、当前 profile(web / desktop)。
② 读它的 README、package.json、插件 manifest,列出它要求的 DSH 版本 / Node 版本 / 操作系统 / 外部依赖 / 需要另外先装的运行时。
③ 逐条比对,结论只写「满足 / 不满足 / 未知」三种;不满足的给出可行替代方案。
④ 检查是否和我已装的插件冲突:命令名重复、skill / tool 重名、端口占用、重复注册的 MCP server。
【2 安全性检查】
① 仓库可信度:和上面「本站登记」是否一致;star / fork 数、创建时间、最近提交,是否归档或长期停更。
② 安装脚本:逐行看 package.json 的 preinstall / install / postinstall,以及 install.sh、setup.ps1 之类脚本。出现 curl|bash、下载后直接执行、混淆代码、访问与插件功能无关的域名,立刻停下来告诉我,不要继续装。
③ 依赖:列出新增依赖,标出无人维护、或与知名包拼写近似的可疑包(typosquatting)。
④ 权限与副作用:它会读写哪些目录、访问哪些域名、需要哪些 DSH 权限(filesystem / network / shell / clipboard 等),以及怎么卸载和回滚。
⑤ 如果它要求 sudo / 管理员权限,或权限明显超出功能所需,先停下来问我。
【3 安装】
上面两步没有「不满足」和「高危项」时才执行;用官方推荐方式安装,不要自行提权。
【4 汇报】
用表格输出:检查项 / 结论 / 依据 / 是否需要我决策。拿不准的一律写「未知」并说明要我怎么确认——不要猜,也不要替我决定。
Send this message to DSH in your current session: it verifies compatibility and security first (answering met / not met / unknown item by item) and only installs once everything checks out — it will stop and ask you if it finds a high-risk item. The box scrolls; the copy is the full prompt. CLI install commands may not be accurate across systems, so DSH is the safer route.把上面这条消息直接发给当前会话里的 DSH:它会先核对兼容性与安全性(逐条给「满足 / 不满足 / 未知」),确认没问题再安装,有高危项会停下来问你。框内可滚动,复制到的是完整提示词;安装命令不一定准确,发给 DSH 更稳。
- Only 4 stars - very few users, little community feedback星标只有 4,几乎没人在用,遇到问题缺少社区反馈
DSH walks through these 9 checksDSH 会逐条核对这 9 项
Compatibility兼容性
- DSH, Node, OS and profile requirementsDSH 版本 / Node 版本 / 操作系统 / profile 是否满足要求
- External dependencies and runtimes (Electron / Python / Docker, ...)外部依赖与运行时(Electron / Python / Docker 等)是否齐备
- Conflicts with installed plugins: command names, skill / tool names, ports, duplicate MCP registration与已装插件是否冲突:命令名、skill / tool 重名、端口占用、重复 MCP 注册
Security安全性
- Repo matches the facts registered here; archived or abandoned?仓库是否与页面登记一致,是否归档或长期停更
- Safety of preinstall / install / postinstall and install.sh / setup.ps1preinstall / install / postinstall 与 install.sh、setup.ps1 是否安全
- curl|bash, download-then-execute, obfuscation, unrelated domains → stop immediatelycurl|bash、下载即执行、混淆代码、无关域名 → 立刻停止
- Typosquatting or unmaintained packages among the new dependencies新增依赖里有没有 typosquatting 或无人维护的包
- Requested permissions vs. what the feature actually needs申请了哪些权限、是否超出功能所需(filesystem / network / shell / clipboard)
- Any sudo / admin requirement, plus uninstall and rollback是否要求 sudo / 管理员权限,以及卸载与回滚方式
Anything uncertain must be marked unknown with a note on how to confirm it. This site's signal screen is a static snapshot, not a security audit.拿不准的必须标「未知」并说明要我怎么确认。本站的信号筛查是静态快照,不能替代安全审计。
Or use CLI install (for developers)或使用命令行安装(适合开发者)
CLI Install命令行安装
dsh plugin --profile web add dsh-jev-adapter
把 BetterZflyee/dsh-jev-adapter 加入你的 DSH 配置(web profile)即可启用。
READMEREADME
dsh-jev-adapter
Use the Jev (System One) decision-model paradigm with any OpenAI-compatible LLM — no TypeSafe API key required. Registers a jev_decide tool on DeepSeek Harness (dsh).
Jev (TypeSafe AI) is a decision model: you send a state plus typed questions (choice / score / boolean), it returns calibrated, probability-distributed answers — no text generation. This plugin brings that paradigm to any chat model you already have, and can also talk to the real Jev API if you have a key.
Why
zhangxaochen/dsh-jev(worth checking out) integrates the official TypeSafe API — great if you have a key and can reachapi.typesafe.ai.- This plugin covers everyone else: any OpenAI-compatible endpoint works — DeepSeek's own API, Ollama, vLLM, OpenRouter, SiliconFlow, LM Studio, or an air-gapped internal gateway.
- Probabilities on the
openaichannel are the model's self-reported estimates (not mathematically calibrated — every result carries this caveat). On thetypesafechannel you get the real thing.
What you get
One tool, jev_decide, callable by the agent:
| Question type | You provide | You get back |
|---|---|---|
choice |
options map (≤255) | picked option + full probability distribution + confidence |
score |
ordered levels (2–10) | fractional score + per-level probabilities + confidence |
boolean |
a statement | P(true) |
Ask many questions in one call — 1 and 10 questions cost about the same.
Install
dsh plugin --profile <profile> add github:BetterZflyee/dsh-jev-adapter
# or, once published on npm:
dsh plugin --profile <profile> add dsh-jev-adapter
Then set one environment variable and restart dsh:
# openai channel (default) — point at any OpenAI-compatible endpoint
export JEV_OPENAI_API_KEY="sk-..."
# optional overrides:
# export JEV_BASE_URL="https://api.deepseek.com/v1" # or Ollama/vLLM/OpenRouter/...
# export JEV_MODEL="deepseek-chat"
Or configure in cordis.patch.yml instead of env vars (see below).
Configure
All settings live in the plugin row in cordis.patch.yml:
Showing the opening section of the README — the full document lives in the repository以上为 README 开头摘要,完整文档在仓库内 · View the full README on GitHub →在 GitHub 查看完整 README →
mafeis/dsh-net-proxy
esengine/DeepSeek-Reasonix
strukto-ai/mirage