dsh-perf-hud：在 Harness 界面内查看 CPU / GPU / 内存与温度
dsh-perf-hud 是 ChendarCe 为 DeepSeek Harness 编写的客户端插件，在 Web 界面内叠加透明可拖动的三行读数，展示 CPU、GPU、内存占用及可选 CPU/GPU 温度。数据经本机 loopback 采集，拒绝外网访问。支持负载配色、主题令牌与窗口角锚点。安装命令为 dsh plugin add ChendarCe/dsh-perf-hud，仅支持 Windows 10/11；温度需可选运行 helper\setup.cmd。适合本地智能体调试时免切换任务管理器。
Read Moredsh-perf-hud：DSH 客户端性能 HUD 与守夜事故包
dsh-perf-hud 是 yanzhaohui1999 维护的 DSH 纯客户端性能 HUD，右下角浮层展示 FPS、Longtask 四条腿归因、各插件回调与 DOM 写入、JS 堆及宿主 event-loop 与链路 RTT。默认开启守夜层，用帧环与 LoAF 在卡顿超 100ms 时自动切片并留存最差现场，支持 __perfHud 控制台 API 与一键导出 JSON。Web 安装：dsh plugin --profile web add dsh-perf-hud。与 DSH 开发者工具分工：显微镜断点/网络 vs 行车记录仪复盘。
Read MoreDSH 右侧栏性能监控：@xmwengxing/dsh-client-ui-sidebar-perfmon
介绍社区插件 @xmwengxing/dsh-client-ui-sidebar-perfmon：在 DeepSeek Harness web 右侧栏展示主机 CPU、内存、交换区、GPU、温度与可排序进程列表，支持 Linux/macOS/Windows 及远程保活。涵盖安装命令、打开方式、cordis.patch.yml 配置项、平台差异与安全注意。需 DSH 0.2.0-rc.1 及以上与 web profile。
Read Moredsh-wsl-perf: Monitoring Windows Host Performance from WSL
dsh-wsl-perf is a DeepSeek Harness plugin designed to collect Windows host performance data from WSL. It retrieves CPU, memory, disk, and process information via WMI/CIM, addressing cross-environment monitoring challenges. It supports customizing the number of displayed processes, making it suitable for developers who need to tune host resources in WSL. The plugin is open source under the MIT License, easy to install, and includes comprehensive unit tests.
Read MoreDeepSeek Harness Web Plugin: Automotive-Style Real-Time Performance Strip
`dsh-live-perf-gauges` is a client-side plugin for DeepSeek Harness (DSH) Web that provides real-time monitoring of decoding throughput, time to first token (TTFT), and other performance metrics. It does not rely on the backend and renders data by projecting the session event stream, while preserving privacy by not storing sensitive information. After installation, restart the service and refresh the browser to use it.
Read Moredsh-telemetry Local Telemetry Plugin
`dsh-telemetry` is a local-first plugin for DeepSeek Harness (DSH) that provides privacy-friendly runtime observability. It does not collect conversation content, and only records lifecycle metrics for requests, models, and tools, such as latency, token counts, cost, error rates, and cache status. It supports status queries via the command line or a local read-only Web UI, helping developers optimize performance and control costs. After installation, users can query data such as P95 latency, time to first token, and estimated cost.
Read Moredsh-response-meta: A Real-Time Response Metadata Plugin for DeepSeek Harness
`dsh-response-meta` is a web plugin for DeepSeek Harness (DSH) designed to improve the observability of streaming output. It displays metrics such as model name, inference progress, throughput, and time to first token in real time within the response stream. Installation requires DSH 0.1.5-rc.1, takes effect after a service restart, supports manual interruption scenarios, and automatically hides duplicate native timestamps.
Read Moredsh-api-budget: API Latency Budget
`dsh-api-budget` is a pure Node.js plugin designed to address the difficulty of precisely calculating interface latency by parameters in DeepSeek Harness (DSH). It provides an `api_budget` tool that enables detailed latency budget calculations locally. The plugin requires no external network dependencies, is easy to install, is released under the MIT License, and is suitable for developers who need fine-grained control over calculation logic.
Read Moredsh-local-diagnostics: A Privacy-First Local Diagnostics Plugin for DeepSeek Harness
`dsh-local-diagnostics` is a local memory diagnostics plugin for DeepSeek Harness (DSH). It leverages Cordis extension points to focus on collecting performance metrics (such as dispatch latency) and structured error codes without modifying the Agent Loop. It strictly follows a privacy-first principle and does not collect prompts, sensitive information, or make network requests. The plugin provides sanitized aggregate reports through the `diagnostics_report` interface. Installation requires Node.js and other environment prerequisites, and, after configuration, it can quickly troubleshoot tool-calling issues locally. The current version does not include a UI, and statistics are limited to the plugin process lifecycle.
Read More