DeepSeek Harness 插件

dsh-pgragh

Live parallel execution graph for DeepSeek Harness(英文原文)

跳到安装方式

来源信息

GitHub 仓库
dangoZhang/dsh-pgragh
最近更新
2026年8月15日
分类
工具与能力
GitHub stars
1
载体类型
plugin
目录证据
上游声明已找到 dsh.bundle
证据路径
package.json#dsh.bundle
核对版本
0.1.0-rc.8
上游核对日期
2026-08-20

该证据由上游目录提供。本站没有安装、运行或安全审核这个插件。

安装

默认先复制一段 Prompt,让 Agent 读 GitHub 仓库和源码;需要自己装时再切到命令。

复制这段 Prompt,发给 DSH、Codex 或其他 Agent,让它先读 GitHub 仓库和源码。

请先不要安装或执行任何命令。阅读这个插件的 GitHub 仓库、README 和关键源码,然后用清楚、直接的方式回答以下问题,帮助我判断它是否适合我的需求:

1. 这个插件是什么,解决什么问题;
2. 适合哪些用户和典型使用场景;
3. 安装后如何使用,并给出一个最小使用示例;
4. 有哪些已知限制,以及隐私、安全、兼容性或维护风险;
5. 给出“推荐 / 有条件推荐 / 不推荐”的明确建议和理由。

请区分仓库明确说明、根据源码推断和未知信息。证据不足时请明确说明,不要猜测或照抄 README。

GitHub:https://github.com/dangoZhang/dsh-pgragh
插件名:dsh-pgragh
作者:dangoZhang

检查来源文件

安装前先看这个插件目录里的 README 和其他文件。

文件资源管理器3 个文件
README.md来源说明 · 只读预览

<h1 align="center">dsh-pgragh</h1>

<p align="center"><strong>A live parallel execution line for DeepSeek Harness Web.</strong></p>

<p align="center"> <a href="README.zh-CN.md">简体中文</a> · <a href="https://github.com/dangoZhang/dsh-pgragh/actions/workflows/ci.yml"><img alt="CI" src="https://img.shields.io/github/actions/workflow/status/dangoZhang/dsh-pgragh/ci.yml?style=flat-square&label=tests"></a> <a href="https://www.npmjs.com/package/dsh-pgragh"><img alt="npm" src="https://img.shields.io/npm/v/dsh-pgragh?style=flat-square"></a> <a href="https://github.com/deepseek-ai/deepseek-harness"><img alt="DeepSeek Harness" src="https://img.shields.io/badge/DeepSeek_Harness-plugin-475569?style=flat-square"></a> <a href="LICENSE"><img alt="MIT" src="https://img.shields.io/github/license/dangoZhang/dsh-pgragh?style=flat-square"></a> </p>

dsh-pgragh adds one compact white line above the DeepSeek Harness composer. Every point comes from a real tool dispatch, physical completion, and ordered session commit, so you can see whether parallel work happened, where it waited, and how much wall time it saved.

model emits:    A  B  C  D
completion:     B  A  C  D
ordered commit: A  B  C  D

![Live parallel execution line in DeepSeek Harness Web](docs/demo.png)

Install

Requires DeepSeek Harness and Node.js 22.19+ or 24+.

npx --yes @deepseek-ai/dsh plugin --profile web add dsh-pgragh@0.1.0
npx --yes @deepseek-ai/dsh web

Open a conversation and run more than one tool in a step. The line appears automatically and stays hidden when there are no calls. A source install is also supported:

npx --yes @deepseek-ai/dsh plugin --profile web add github:dangoZhang/dsh-pgragh

Read the line

  • Position and state: point position is physical completion time; a hollow pulse is running, a double outline is completed but waiting for ordered commit, and a solid point is committed. Hover for the tool label, duration, and commit wait.
  • Call type: an outer ring marks a subagent, a small point marks a code-mode internal call, and a diamond marks an exclusive barrier. T17 · S4 identifies the active turn and step.
  • Savings: Parallelism = serial leaf-work estimate / wall time; Saved = serial estimate - wall time. Parent run_code calls are excluded when their internal calls are visible, preventing double counting.

Ground truth and limits

The host observes the official tools/execute around-dispatch path for start and completion timestamps. Persisted tool/result and tool/code-dispatch session events determine when results become model-visible. The observer is read-only: it does not change arguments, results, scheduler behavior, or commit order.

The SSE feed contains only session id, tool name, optional description, status, and timestamps. The UI shows the latest observed step; its in-memory timeline starts again when the DSH host restarts.

Verification

pnpm check runs strict type checking, both host and browser builds, state-machine and metrics tests, plus a real DSH integration test using the official ToolRuntime, SessionStore, and WebServer.

pnpm install
pnpm check

Focused issues and pull requests are welcome. Keep the observer read-only and include behavioral tests for timeline changes.

License

[MIT](LICENSE) © dsh-pgragh contributors