DeepSeek Harness Plugin Hub

发布与管理完整 Harness Profiles,发现适合你的插件。

探索

插件目录环境预设文档中心动态

社区

发布插件联系我们报告问题

相关链接

Plugin Hub GitHubDeepSeek Harness 官方项目系统状态隐私说明
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

独立、非官方社区项目,与 DeepSeek 官方无隶属、授权或背书关系。

Open Auth — DeepSeek Harness 插件(DSH Plugin)
DeepSeek Harness Plugin Hub
ProfilesPlugins分类动态文档登录管理 Profiles
ProfilesPlugins分类动态文档登录
← Plugins

dsh-open-auth

Open Auth

为每个 pi-ai 提供商提供一个 DeepSeek Harness 插件,包括订阅 OAuth

插件会安装到这里;不确定时保持 web。

npx -y @deepseek-ai/dsh plugin --profile web add dsh-open-auth@0.1.1
README兼容性版本

兼容性与来源证明

Open Auth 以 dsh-open-auth 发布,当前版本为 0.1.1。Plugin Hub 会校验它的 manifest,并保存精确安装来源,便于复现安装结果。

DSH 兼容范围
*
运行环境
any
发布来源
npm
Registry 更新时间
2026/9/20

版本

0.1.1stable
2026/8/14
0.1.0stable
2026/8/14

相关插件

正在加载相关插件…

最新版
0.1.1
DSH
*
HMR
重启进程
Tree shaking
未声明可安全裁剪
解包体积
56.1 kB
文件数
14
Surface
any
许可证
MIT
发布源
npm
GitHub
★ 0
周下载
32
最近提交
2026/8/19
查看源码 ↗项目主页 ↗
README Badge

点击下方 Badge 复制 Markdown,粘贴到 README 即可。

这是你的 Plugin?认领权益 · 优先安全扫描

验证 package.json 声明的 GitHub 仓库,即可管理这个公开页面。认领后,Hub 会优先安排当前版本的安全扫描,并在通过后公开展示结果。

认领这个 Plugin →
报告问题

相关插件

继续浏览 security-access 分类下经过校验的插件。

Doctor@linxin666/dsh-doctorDSH 配置档案的事务性救援模式,配备受监督的启动器、隔离的恢复容器、确定性修复、健康监控以及本地 Web 恢复控制台Pocketdsh-pocket把 DeepSeek Harness 装进你的口袋:一个包、一个设置页,手机扫码即同步访问电脑上的 DSH(局域网 + 公网,实时同屏)。DSCODE@toddzheng024/dscode-bundle完整的 DeepSeek 编码代理,支持持久化 shell、Ultra 协作和自动权限审查。Auto Reviewdsh-auto-review针对 DeepSeek Harness 审批请求的第二模型 AI 自动审查:只读审查子代理在审批应答链上决定允许或拒绝,并采用故障关闭回退机制和完整的会话日志审计。

README

dsh-open-auth

English | 简体中文

A DeepSeek Harness LLM plugin that connects every chat provider currently built into pi-ai and provides one OAuth/API-key login CLI. The plugin and CLI share a permission-protected credential file, allowing subscriptions such as OpenAI Codex, Kimi Code, and GitHub Copilot to authenticate DSH model requests.

This does not convert a ChatGPT subscription into an OpenAI API key. The openai-codex provider uses the OpenAI Codex OAuth flow and ChatGPT backend implemented by pi-ai. Availability, supported models, and usage limits remain subject to your OpenAI account and applicable service terms.

Installation

Node.js 22.19 or later is required.

Install the package into a DSH profile (replace web with your profile name when needed):

dsh plugin --profile web add dsh-open-auth

If you manage the DSH runtime as a regular Node.js project instead, install it in that project:

npm install dsh-open-auth

Then add the plugin to your DSH configuration as shown below. Installing through dsh plugin add activates the bundled default configuration automatically.

Authentication

When installed in a DSH profile, run the CLI through that profile. Sign in with an OpenAI Codex subscription:

dsh plugin --profile web exec dsh-open-auth login openai-codex
dsh plugin --profile web exec dsh-open-auth status openai-codex

Sign in with a Kimi Code subscription:

dsh plugin --profile web exec dsh-open-auth login kimi-coding oauth

List all providers and models currently exposed by pi-ai:

dsh plugin --profile web exec dsh-open-auth providers
dsh plugin --profile web exec dsh-open-auth models openai-codex

For a regular local npm installation, replace the prefix above with npx, for example: npx dsh-open-auth login openai-codex.

DSH configuration

The bundled configuration registers every built-in pi-ai chat provider. No extra configuration is required after installing it into a profile. Verify the composed row without starting DSH:

dsh --profile web --dump-config

To restrict the registered routes, add the following override to that profile's cordis.patch.yml:

- id: llm-open-auth
  config:
    providers:
      - openai-codex
      - kimi-coding
      - anthropic
      - openrouter
    streamIdleTimeoutMs: 300000

For a standalone Cordis configuration that does not use DSH profiles, add the package as a normal plugin row:

- id: llm-open-auth
  name: dsh-open-auth
  config: {}

Do not load another LLM plugin that owns any of the same provider routes. For example, registering anthropic from both this plugin and the official llm-pi-ai plugin will be rejected by DSH as a duplicate route.

Credential storage

Default locations:

  • macOS: ~/Library/Application Support/dsh-open-auth/auth.json
  • Linux: $XDG_CONFIG_HOME/dsh-open-auth/auth.json or ~/.config/dsh-open-auth/auth.json
  • Windows: %APPDATA%/dsh-open-auth/auth.json

Use --file PATH in the CLI, set credentialFile in the plugin configuration, or set DSH_OPEN_AUTH_FILE for both. The file is created with mode 0600, replaced atomically, and protected by a cross-process lock during writes and token refreshes. It is not an operating-system keychain and has no additional encryption; never upload, share, or commit it to Git.

API-key providers may continue to use the standard environment variables supported by pi-ai, such as OPENAI_API_KEY, ANTHROPIC_API_KEY, and DEEPSEEK_API_KEY. Running dsh-open-auth login PROVIDER api_key stores the key in the shared credential file instead. Ambient-only providers such as AWS Bedrock and Google Vertex read environment variables or local credential files according to pi-ai conventions.

Architecture

flowchart LR
  CLI["dsh-open-auth CLI"] --> Store["Shared CredentialStore"]
  DSH["DSH dsh-open-auth plugin"] --> Store
  DSH --> Models["pi-ai builtinModels"]
  CLI --> Models
  Models --> OAuth["Provider OAuth / token refresh"]
  Models --> APIs["OpenAI Codex, Kimi, Anthropic, etc."]

This project does not maintain a duplicate provider allowlist. builtinModels() is the single source of truth, so upgrading @earendil-works/pi-ai automatically exposes newly built-in providers to the CLI and the default DSH route set.

Development

npm install
npm run check
npm test
npm run build

Publishing

The package name is dsh-open-auth. The package includes a dsh.bundle manifest and cordis.patch.yml, so dsh plugin add installs and activates it as a profile layer. prepublishOnly runs the type check, tests, and production build automatically before publication.

npm login
npm publish

Use npm pack --dry-run to inspect the files that will be published without uploading anything.

The DSH message, stream-event, and replay mappings in this project are adapted from the MIT-licensed DeepSeek Harness implementation. See NOTICE.