dsh-cmdwatch
在终端中运行以下命令:
dsh plugin install GavinQiEr/dsh-cmdwatch
将以下提示词粘贴到 DeepSeek Harness 对话框中:
在 DeepSeek Harness 中运行 dsh plugin install GavinQiEr/dsh-cmdwatch 即可安装该插件,源码地址为 https://github.com/GavinQiEr/dsh-cmdwatch
插件介绍
Running a pytest suite or a build script through DSH has one annoying gap between pressing enter and getting results: you have to pause the conversation and manually scroll output to see how far it has progressed. dsh-cmdwatch turns that into a live, line-by-line feed in a dedicated panel — foreground commands and background jobs alike — without ever interrupting the chat.
The heart of the plugin is a semantics-preserving real-time stream. When a command contains pipeline stages that would block progressive output (Select-Object -Last, tail, Out-File, shell redirections), the plugin automatically inserts Tee-Object or tee before the earliest such stage, writes the stream to a temporary log, and the panel polls the delta every 500 ms. The result DSH ultimately receives is byte-for-byte identical to the unwrapped command: -Last 3 still returns exactly three lines, Out-File still writes the file. Wrapped entries carry a modified badge with the original command on hover, and pipelines the plugin cannot safely wrap surface a yellow warning instead of being silently altered. Panels are isolated per session, the output area auto-scrolls to the latest line, and long commands collapse to a single line with a full-text tooltip.
It is built for developers who routinely hand long-running jobs to DSH — test runs, dependency installs, compile builds, data pipelines — and want progress visible without their eyes leaving the window. Toggle the real-time stream off and the output delta is handed back to the DSH job_output tool on demand, so both monitoring modes coexist without conflict.
使用场景
- 运行 pytest 或编译构建时实时逐行查看命令进度,无需暂停对话
- 后台任务执行中在面板里滚动跟踪输出,前台命令同样实时显示
- 命令含 Select-Object -Last 或重定向等管道时自动插入 Tee 落盘,面板实时显示全量输出且原管道语义不变
适合人员
- 经常通过 DSH 委托长时任务(测试、构建、安装)的开发者
- 需要在保持对话上下文的同时监控命令执行进度的用户
- 使用 DSH web profile 进行脚本自动化与 CI 类工作流的工程师