dsh-perf-lens：按插件归因的 DSH 资源开销面板
DSH 宿主插件共用一个 Node 进程，系统级监控难以按插件归因。dsh-perf-lens（Nwflower，客户端插件）在进程内采样并将 CPU、采样堆与磁盘相关指标记到触发栈上的插件，在 dsh web 侧栏 Perf Lens 中展示进程总览、开销构成、趋势与累计排行；浏览器侧支持长帧脚本按插件归因及 DSH Desktop 下的 jank probe。要求 DSH ≥0.1.7-alpha.1、Node ≥22.13。尚未上 npm，需本地 build 后以 link 注册到 web profile 并重启 dsh web。MIT 许可。
Read MoreTencent Cloud APM：在 Agent 里用 MCP 做调用链与性能分析
Tencent Cloud APM（@tencent-adm/apm-performance-analysis，v1.0.3）由腾讯云 APM 团队发布，通过 ApmClient.SendMCPMessage 连接 APM MCP Server，配合内嵌 SDK 与 apm_mcp_client.py 完成 ping、列工具与 call-tool。文章说明 Skill 定位、核心能力、SkillHub 安装路径、TENCENTCLOUD_SECRET_ID/KEY 配置，以及 list-tools 与 DescribeApmOverview 示例调用流程，并提示 requires_api_key 与交互确认规范。
Read Moredsh-system-monitor: Real-Time Hardware Monitoring in the DSH Sidebar
The `dsh-system-monitor` plugin provides real-time hardware monitoring (CPU, GPU, memory, and disk) in the DSH sidebar. Core features include ring gauge visualization, application-level drill-down analysis, slope-based capacity forecasting, threshold alerts, and data export in multiple formats. The plugin supports preset and custom time ranges for viewing, with history records persisted to disk using a rolling write strategy. Full functionality is supported only on Windows 10/11; Linux and macOS support basic readings only. Install it with `dsh plugin install dsh-system-monitor`. Enable the sidebar with `dsh web`. It runs without administrator privileges and binds to localhost only.
Read Moredsh-token-speed: Real-Time Speed Monitoring Plugin for DSH Clients
The DeepSeek Harness (DSH) plugin `dsh-token-speed` visualizes real-time inference speed using a circular dial. It offers real-time speed display, per-step precise calibration based on `usage.outputTokens` and `timing`, a details panel, status indicators, and a draggable memory feature. Built on a browser-side architecture, it does not modify host state and only reads conversation data through `uiConversation`, providing developers with a lightweight performance monitoring solution. Installation requires adding a local link dependency to `package.json` in the profile and running the installation command.
Read MoreDeepSeek Harness Plugin: vLLM-Ascend Profiler Analyzer
The DeepSeek Harness plugin is designed to address the performance profiling challenges encountered when deploying vLLM on Ascend NPUs. By parsing profiling artifacts generated by CANN and torch_npu, it supports streaming analysis and visualization, automatically distinguishes between the Prefill and Decode stages, and generates optimization recommendations and reports. The plugin has zero runtime dependencies and can be installed via pnpm or a local link.
Read More