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。

Locale Pack — DeepSeek Harness 插件(DSH Plugin)
← Plugins
L

dsh-locale-pack

Locale Pack

DeepSeek Harness 桌面应用的本地化语言包:注册 29 种语言,并为每种语言提供一个词典。

插件会安装到这里;不确定时保持 web。

npx -y @deepseek-ai/dsh plugin --profile web add github:sayho-pm/dsh-locale-pack#3c062212a484aa21dfd33bd8f42c1ced88c3d174
README兼容性版本

兼容性与来源证明

Locale Pack 以 dsh-locale-pack 发布,当前版本为 0.3.1。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
web
发布来源
github
Registry 更新时间
2026/10/14

版本

0.3.1stable
2026/10/14
0.3.0stable
2026/10/12
0.2.0stable
2026/10/3

相关插件

正在加载相关插件…

最新版
0.3.1
DSH
*
HMR
重启进程
Tree shaking
未声明可安全裁剪
解包体积
未提供
文件数
未提供
Surface
web
许可证
MIT
发布源
github
GitHub
★ 10
周下载
0
最近提交
2026/10/14
查看源码 ↗
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
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

相关插件

继续浏览 ui-customization 分类下经过校验的插件。

Web App@deepseek-ai/dsh-web-appdsh 浏览器界面捆绑包:位于 dsh-base 之上的 Web 补丁层,加上运行时粘合插件(提供前端 dist、Web 界面提示符、bash 运行时变量和 URL 行)Experimental Agent Team Web Profile@deepseek-ai/dsh-experimental-agent-team-web-profile用于 Agent Teams Remote 和 UI 插件的实验性 Web 配置层Ab Chartdsh-ab-chart带有绘制其调用的浏览器卡片的图表渲染工具,封装在一个可发布的插件包中Whale Widgetdsh-whale-widgetDSH Web 界面右下角的 DeepSeek 余额小鲸鱼挂件:余额/今日已用/峰谷定价、自定义泡泡点击序列(文本/余额/今日/峰谷/图片/随机语句与并列加权选择)、逐行样式与字体、悬浮快捷编辑、音效与每轮消耗、自定义角色/动图/音效、吸附与翻转自定义

README

dsh-locale-pack

English | 한국어

DeepSeek Harness Multilingual Pack, the dsh-locale-pack plugin: 29 languages, 73,312 entries, 31 selectable

The 29 languages and their codes

A locale pack for the DeepSeek Harness desktop app. It registers 29 languages in the language picker and ships a dictionary for each one. Strings that are not translated yet fall back to English, so the screen stays readable while coverage grows. DeepSeek Harness ships English and Simplified Chinese by default, which makes 31 languages selectable after install.

Why this exists

DeepSeek Harness is a coding tool: you hand it a task and the agent reads files, edits them, and runs terminal commands. Its desktop interface is built around chat, settings, plugin management, sub-agents, task history, approval prompts, voice input, and scheduled jobs.

That interface was available in English and Simplified Chinese. Developers who prefer to work in another language had to read every label in a second language. This project fills that gap by adding the dictionaries, and by keeping the number of languages wide rather than deep: a language is usable once the picker offers it and the common screens are readable.

Supported languages

LanguageCodeLanguageCode
한국어koPolskipl
日本語jaРусскийru
繁體中文 (台灣)zh-twУкраїнськаuk
繁體中文 (香港)zh-hkSvenskasv
DeutschdeTürkçetr
FrançaisfrBahasa Indonesiaid
EspañolesTiếng Việtvi
Português (Brasil)pt-brไทยth
Italianoitहिन्दीhi
Nederlandsnlالعربيةar
FilipinotlRomânăro
বাংলাbnاردوur
עבריתheفارسیfa
Português (Portugal)ptKiswahilisw
Yorùbáyo

29 languages × 2,528 keys = 73,312 translated entries. The dictionaries are built against DeepSeek Harness 0.2.0-rc.2.

Install

The DeepSeek Harness desktop app has to be installed first. Then clone the repository and register it as a plugin link.

macOS and Linux (bash):

# 1) Get the sources
git clone https://github.com/sayho-pm/dsh-locale-pack.git
cd dsh-locale-pack

# 2) Register the folder with the desktop profile
dsh plugin --profile desktop add link:$(pwd)

Windows (PowerShell):

# 1) Get the sources
git clone https://github.com/sayho-pm/dsh-locale-pack.git
cd dsh-locale-pack

# 2) Register the folder with the desktop profile
dsh plugin --profile desktop add link:(Get-Location).Path

Then open Settings → Language and pick the language you want.

Install with an AI assistant

If you do not have a development environment, an AI assistant can do the setup for you. Open the assistant and paste this:

Install this GitHub repository: https://github.com/sayho-pm/dsh-locale-pack
Wherever I have to type something myself, such as a token or a path,
mark it with **** and tell me exactly where it goes. Handle the rest.

The assistant runs the commands and stops only at the points that need your input.

Repository layout

locale/<language>/*.json   Dictionary sources — edit here
locale/<language>.json     Plugin display text per language (title/description)
locale/_policy/            Terms kept in English
lib/client.js              Generated bundle (from the dictionaries)
styles/locale-fonts.css    Locale-scoped font corrections (inlined at build)
tools/build-client.mjs     Dictionaries → lib/client.js
tools/smoke.mjs            Smoke test
tools/check-keep.mjs       Translation quality check
tools/translate_worker.py  Translation worker (runs outside the chat session)
tools/clean-romanized.py   Fixes transliteration slips
tools/en.baseline.json     English baseline (DeepSeek Harness 0.2.0-rc.2)

Editing a string

node tools/build-client.mjs   # rebuild lib/client.js
node tools/smoke.mjs          # self-check: loading, fallback, idempotence
node tools/check-keep.mjs     # translation quality check

Placeholders such as {name} must be kept exactly as they are. If the order or the spelling changes, the sentence breaks at runtime.

Bundling only the languages you need

Installing the whole pack is the default, and switching language afterwards is a matter of the settings screen. When the full bundle is more than you want, build a subset instead:

node tools/build-client.mjs --langs ko,ja   # Korean and Japanese only

A two-language bundle comes out at roughly 320 KB instead of 4.2 MB. The flag accepts any comma-separated list of the language codes in the table above. Omitting it rebuilds every language.

Translation quality checks

tools/check-keep.mjs compares every language against the English baseline and reports five things:

  1. Placeholders such as {name} appear the same number of times as in the source
  2. Product and protocol names (DeepSeek Harness, MCP, JSON, OpenAI Responses, and others) are left untranslated
  3. Languages that do not use Chinese characters contain none
  4. Korean sentences contain no em dashes
  5. No value is empty, and no key is present that the baseline does not define

Terms that stay in English are collected in locale/_policy/keep-english.json. Add a term there to keep it untranslated.

Re-running the translation

Translation runs in a separate process from the chat session. Nothing from the session model or the conversation history is attached to a request: only the English source strings leave the machine.

python3 tools/translate_worker.py --langs ja,de      # specific languages
python3 tools/translate_worker.py --all              # everything

Requests move through four models in order. If the first one fails, the worker retries along an equivalent route, then an earlier version, and finally a model that is not used for training. 번역_로그 records which model handled each batch.

Notes for users

  • Arabic, Urdu, Hebrew, and Persian are written right to left. The dictionaries are ready; the screen layout is handled by DeepSeek Harness itself.
  • The menu bar, the right-click menu, and the update dialog belong to the desktop shell, so this pack does not change them.
  • When DeepSeek Harness ships a new version, new strings appear in English until they are translated.
  • Korean is reviewed by hand. Other languages come from machine translation, and each language's dictionary file is the place to adjust wording.
  • Japanese and Traditional Chinese keep their own glyph shapes. The harness font stack lists Simplified-Chinese faces ahead of the other CJK faces, so Han characters shared between the standards would otherwise be drawn in their Simplified form. The pack scopes a corrected stack to ja, zh-tw, and zh-hk, and it follows the language picker.

Contributing

Requests for additional languages are welcome. Open an issue in this repository naming the language, and it gets picked up in the next round. Corrections to an existing language go straight into that language's dictionary file under locale/<language>/, and a pull request is the easiest form.

License

MIT