dsh-performance-optimizer: A DSH plugin for Host Performance Observability in DeepSeek Harness
weidows/dsh-performance-optimizer
A DSH plugin that samples host-wide CPU, memory and IO metrics, exposing them via a Settings tab, floating HUD and native agent tool for diagnosing lag and memory drift.
dsh-performance-optimizer is a performance observability DSH plugin for DeepSeek Harness that turns host process CPU, memory, and IO metrics into data both users and agents can read directly. It targets hard-to-pin-down issues such as scroll lag, memory drift, and tool loops, so diagnosis no longer relies on guesswork. The plugin samples Node process usage once per second and keeps 60 samples, exposing them through a live Settings table and chart, an optional floating HUD, and a native performance_report tool, maintained by Weidows under the MIT license.
How to Install
dsh plugin --profile web add github:weidows/dsh-performance-optimizer- Category
- Development & Operations
- Platform
- DSH Plugin
- Author
- weidows
- Distribution
- Plugin
dsh-performance-optimizer Key Features
dsh-performance-optimizer Repo Summary
What Does It Do?
It is a DSH plugin for DeepSeek Harness that turns host-wide performance metrics into data both users and agents can read directly. It solves the problem of hard-to-pin-down scroll lag, memory drift, and tool loops, so performance diagnosis no longer relies on guesswork. The plugin samples Node process memory and CPU usage once per second, keeps a rolling buffer of 60 samples, and exposes them through three surfaces: a Settings tab, an optional floating HUD, and a native agent tool. The bundle is maintained by Weidows under the MIT license and written in JavaScript.
Core Features
- Live metrics panel: mounts a live table and a Canvas2D chart into the per-plugin detail page, with CPU, RSS, Heap, and Wall tabs.
- Floating HUD: when hudVisible is true, mounts an overlay into the document body so metrics stay visible while you browse.
- Native agent tool: registers a performance_report tool that returns a fresh JSON report on demand inside an agent session.
- Rolling sample buffer: samples process.memoryUsage() and process.cpuUsage() once per second and keeps only the latest 60 samples, deliberately staying lightweight.
- Detailed sample fields: separates rss, heapUsed, heapTotal, and external, and records both wall-clock and monotonic time so you can tell whether the process grew or the V8 heap expanded.
How to Use This Plugin?
After enabling this DSH plugin in the Web profile and restarting, open the plugin detail page under Settings to see the live metrics table and chart. To watch metrics while working, set hudVisible to true and the floating HUD appears; to let the agent help analyze, have it call the performance_report tool for a JSON report. When chasing scroll lag or memory drift, compare the RSS and heapUsed curves to locate the source.
How to Troubleshoot This Plugin?
If the metrics panel shows no data, confirm the Web profile was restarted and that the plugin loaded successfully. If the floating HUD does not appear, hudVisible is usually not enabled. If the agent gets no report, the performance_report tool was not registered; fix it per the official instructions and retry.
This page is an independent rewrite of the plugin's official README — for authoritative documentation and the latest changes, refer to the source: weidows/dsh-performance-optimizer. The plugin is third-party code that runs on your machine once installed; inclusion does not imply endorsement — please review the source before installing.
DSH Plugin FAQ
