DeepSeek Harness Plugin Hub

Publish and manage complete Harness Profiles. Discover Plugins for your next setup.

Explore

PluginsPresetsDocsNews

Community

Publish a pluginContactReport an issue

Resources

Plugin Hub on GitHubDeepSeek HarnessSystem statusPrivacy notice
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

Independent and unofficial. Not affiliated with, authorized by, or endorsed by DeepSeek.

DeepSeek Harness Plugin Hub
ProfilesPluginsCategoriesNewsDocsSign inManage Profiles
ProfilesPluginsCategoriesNewsDocsSign in
← Plugins
C

@lsfcy/dsh-composer-predictions

Composer Predictions

Codex-style composer predictions for DeepSeek Harness: after a turn settles, the empty composer shows a dimmed suggestion of your next message; Tab accepts it.

The plugin will be installed here. Keep web if you are unsure.

npx -y @deepseek-ai/dsh plugin --profile web add github:lsfcy/dsh-composer-predictions#74a97a0a69453305e07739c889a686d9f7774187
READMECompatibilityVersions

Compatibility and provenance

Composer Predictions is published as @lsfcy/dsh-composer-predictions and currently resolves to version 0.1.3. The Hub verifies its manifest and preserves the exact installation source for reproducible installs.

DSH compatibility
*
Runtime surfaces
web
Release source
github
Registry updated
10/29/2026

Versions

0.1.3stable
10/29/2026
0.1.2stable
10/29/2026
0.1.0stable
10/28/2026

Related plugins

Loading related plugins…

Latest
0.1.3
DSH
*
HMR
Process restart
Tree shaking
Safe tree shaking not declared
Unpacked size
Unavailable
Files
Unavailable
Surface
web
License
MIT
Source
github
GitHub
★ 1
Weekly downloads
0
Last push
11/2/2026
View source ↗
README badge

Click the badge to copy Markdown for your README.

Do you maintain this Plugin?Claim benefit · Priority security scan

Verify the GitHub repository declared in package.json to manage this listing. After you claim it, Hub will prioritize a security scan of the current version and publish the result when it passes.

Claim this Plugin →
Report an issue

Related plugins

More verified plugins in ui-customization.

Client Ui Skill Explorer@linxin666/dsh-client-ui-skill-explorerDSH skill center: browse loaded skills by source (bundled / project / user / custom / runtime), enable or disable, create and delete, in a web GUI panel.Client Ui Task Board@linxin666/dsh-client-ui-task-boardHost-authoritative task board for the DSH Web GUI with real session execution, Host cron scheduling, and optional cross-platform idle-sleep protection; mounted without DSH source changes.Client Ui Git Graph@linxin666/dsh-client-ui-git-graphExternal dsh web GUI plugin: a blank-session git branch selector + Git graph, with real host-side git operations and guards, as a dsh profile bundleClient Ui Market@linxin666/dsh-client-ui-marketWorkshop browser card: browse skins, pets and community plugins from dsh-market.com and install them in the GUI (assets into the DSH home directories, plugins through the optional plugin manager).

README

dsh-composer-predictions

给 DeepSeek Harness 的输入框补上“下一句预测”:一轮对话结束后,在空的输入框里用灰字显示 你接下来可能发送的消息。Tab 填入草稿,Esc 或开始打字即忽略。不会自动发送。

English | 中文 · 在线试玩

示例

助手刚说完「测试全部通过」,空输入框里出现灰字:

顺手把这次改动提交了吗?提交信息我一起写好

按 Tab,它变成可编辑的草稿;按 Esc 或直接打字,它消失。

灰字预测:出现、Tab 接受

安装

仅适用于 DSH 的 Web / 桌面端(0.2.x),不能用于其他 agent/harness。

方式一(推荐,终端):

dsh plugin --profile desktop add @lsfcy/dsh-composer-predictions

装完不用重启:运行中的 DSH 会监听 profile 的 bundle 清单,装好即加载,已经打开的页面也会直接挂上新插件。桌面端 profile 是 desktop; 其他入口用 ~/.dsh/profiles 里对应的名字,如 web。不想走 npm 时,也可以用固定版本: dsh plugin --profile desktop add git+https://github.com/lsfcy/dsh-composer-predictions.git#v0.1.3。

灰字一直不出现就刷新一次页面;万一安装时宿主报了模块加载失败(pnpm 正在改写 node_modules,宿主读到了半写入的文件),重启一次 DSH 即可。

提示 command not found: dsh 说明终端里没有 CLI,先补上再装插件:

  • 桌面端:菜单栏 DeepSeek Harness → 管理 dsh 命令…,按提示把 dsh 链接到终端。
  • 或者 npm 全局安装(需要先有 Node.js):npm install -g @deepseek-ai/dsh。

方式二(备选,在 DSH 会话里):让 agent 执行

plugin_manager  action: install_bundle
                target: @lsfcy/dsh-composer-predictions

这条路依赖 agent 工具和当前 profile 状态,失败率更高;装不上就用方式一,效果一样。

配置

设置 → 输入框预测:

配置默认说明
enabledtrue总开关:关闭后不再请求,也不显示灰字。
provider自动留空用你当前选中的 provider。
model自动留空用该 provider 下最便宜的型号。
reasoningEffort自动留空用模型支持的最省推理档。

这些键同样可以写在 profile patch 的 bundle row config 里。

原理

  1. 宿主读取会话里真人的 user/assistant 消息,只取最近 12 条,压到 6000 字符。
  2. 只在一轮成功结束、且最后一条是 assistant 回复时预测。
  3. 调一次便宜模型,取一行(≤160 字符),按会话内容缓存 30 分钟。
  4. 客户端把它画成灰字;Tab 写入输入框。

成本

单次生成约 2–3k token(system 提示词 + 转录 + 输出上限 512),走最便宜型号和最低推理档。 命中缓存不产生费用;被中止的轮次不会调用模型。

限制

  • 它只是猜测,没有记忆:不做检索,不读长期画像,只看最近这段转录。
  • 只在成功轮次后出现,且永不自动发送。
  • 便宜模型可能平淡或猜错;在意质量可在配置里固定更强的模型。
  • 只给一行,不会给多步计划。