DeepSeek Harness Plugin Hub

发布与管理完整 Harness Profiles,发现适合你的插件。

探索

插件目录环境预设文档中心动态

社区

发布插件联系我们报告问题

相关链接

Plugin Hub GitHubDeepSeek Harness 官方项目系统状态隐私说明
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

独立、非官方社区项目,与 DeepSeek 官方无隶属、授权或背书关系。

Call Session — DeepSeek Harness 插件(DSH Plugin)
DeepSeek Harness Plugin Hub
ProfilesPlugins分类动态文档登录管理 Profiles
ProfilesPlugins分类动态文档登录
← Plugins
C

dsh-call-session

Call Session

DSH 原生跨会话协作、进程内会话调用和公共黑板插件

插件会安装到这里;不确定时保持 web。

npx -y @deepseek-ai/dsh plugin --profile web add github:popu2do/dsh-call-session#6f01c72886f82e5de4a8ed2b9ff8fe91820e2095
README兼容性版本
dsh-call-session canvas

兼容性与来源证明

Call Session 以 dsh-call-session 发布,当前版本为 0.1.6。Plugin Hub 会校验它的 manifest,并保存精确安装来源,便于复现安装结果。

DSH 兼容范围
*
运行环境
web
发布来源
github
Registry 更新时间
2026/9/23

版本

0.1.6stable
2026/9/23
0.1.5stable
2026/9/22
0.1.4stable
2026/9/21

相关插件

正在加载相关插件…

最新版
0.1.6
DSH
*
HMR
重启进程
Tree shaking
未声明可安全裁剪
解包体积
未提供
文件数
未提供
Surface
web
许可证
MIT
发布源
github
GitHub
★ 1
周下载
0
最近提交
2026/9/23
查看源码 ↗
README Badge

点击下方 Badge 复制 Markdown,粘贴到 README 即可。

这是你的 Plugin?认领权益 · 优先安全扫描

验证 package.json 声明的 GitHub 仓库,即可管理这个公开页面。认领后,Hub 会优先安排当前版本的安全扫描,并在通过后公开展示结果。

认领这个 Plugin →
报告问题

相关插件

继续浏览 integrations-communication 分类下经过校验的插件。

Im@xmanrui/dsh-im将十二种 IM 渠道和公网 AI Office 接入本机 DeepSeek Harness。Pocketdsh-pocket把 DeepSeek Harness 装进你的口袋:一个包、一个设置页,手机扫码即同步访问电脑上的 DSH(局域网 + 公网,实时同屏)。Im Connect@michengai/dsh-im-connectDSH IM Connect — 将主流即时通讯平台接入本机 DeepSeek Harness 智能体DSCODE@toddzheng024/dscode-bundle完整的 DeepSeek 编码代理,支持持久化 shell、Ultra 协作和自动权限审查。

README

dsh-call-session hero

dsh-call-session

In-process cross-session communication and workspace shared blackboard for DeepSeek Harness (DSH)

English | 简体中文


Overview

DeepSeek Harness sessions are isolated by default, and child tasks exit upon completion.

dsh-call-session adds in-process cross-session communication, peer session creation, and shared blackboard capabilities:

  • Inter-session unicast: send directives or hand off tasks between independent long-running sessions.
  • Shared blackboard: publish bulky artifacts to a shared board for on-demand retrieval, avoiding context bloat.
  • Visual canvas: observe session states, blackboard items, and call trajectories in the DSH Web UI.

Install

# Add to target profile web
dsh plugin --profile web add dsh-call-session

# List installed plugins
dsh plugin --profile web list

# Remove plugin
dsh plugin --profile web remove dsh-call-session

Note: The --profile option must follow the plugin subcommand, e.g. dsh plugin --profile web ....


Scenarios

Instruct the agent using plain natural language in the chat; the agent invokes the tools automatically:

Task Handoff

User prompt:

Hand off this task summary to the "audit-session" session.

The agent calls session_call to dispatch a unicast message. If the target session is running, it steers execution; if idle, it wakes a new turn.

Post Results

User prompt:

Publish the review conclusions to the board and notify relevant sessions to pick up tasks.

The agent calls board_post to store the detailed report on the blackboard and obtain a postId. Other sessions retrieve the content on demand via board_list, preventing context window explosion.

Create Session

User prompt:

Create a new session and hand off tasks.

The agent calls session_create to spawn a long-running peer session in the current workspace, automatically inheriting model specifications and presets.


Canvas

The DSH Web session view exposes a Canvas tab to observe multi-agent collaboration across workspaces:

dsh-call-session canvas

  • Topology: workspace columns, running/idle session nodes, blackboard items with TTL countdowns, and directional call traces.
  • Interaction: hovering highlights the 1-hop connected subgraph; double-clicking opens a read-only inspector drawer.
  • Read-only boundary: the canvas offers no control inputs; all state changes originate from the agents themselves.

Tools

ToolTypeModeCommon Use Case
session_callToolPush1:1 unicast call for task dispatch, progress reporting, and handoffs
session_createToolNativeCreate long-running peer sessions for parallel multi-role collaboration
session_queryToolRead-onlyList sessions and check active/idle statuses in or across workspaces
board_postToolPullPublish blackboard entries for large artifacts and shared milestones
board_listToolPullQuery blackboard records by ID, topic, or digest mode
board_clearToolManageDismiss or purge completed and expired blackboard entries

Comparison

MechanismTypeModelWhen to Use
subagentBuilt-inHierarchical delegation, ephemeral child task, exits on completionScoped exploration, code search, one-off script execution.
dsh-call-sessionPluginPeer-to-peer collaboration, in-process unicast messaging and shared boardDirectives and status sync between independent top-level sessions, or sharing large outputs.

Selection guidelines:

  • Prefer built-in: for isolated, scoped tasks, use DSH built-in subagent directly.
  • Use on demand: when multiple independent top-level sessions need to coordinate or share large context via a blackboard, use dsh-call-session.

Config

The plugin runs with built-in defaults out of the box.

To customize, add property overrides to ~/.dsh/profiles/web/cordis.patch.yml:

- id: dsh-call-session
  config:
    debounceMs: 500
    maxCapacity: 500

Options:

  • enabled: boolean, default true. Enable or disable plugin tools.
  • debounceMs: number, default 300. Atomic disk write debounce delay in milliseconds.
  • maxCapacity: number, default 200. Maximum board entries.
  • telemetryCapacity: number, default 200. Maximum call traces retained in canvas data buffer.
  • locale: string, default 'auto'. Interaction language preference ('auto', 'zh', 'en').

Architecture

Technical decisions are recorded as Architecture Decision Records (ADRs):

  • Decision catalog: see docs/adrs/README.md.
  • Architecture topology: see docs/architecture/.

Compatibility

EnvironmentSupportedVerified
DeepSeek Harness>=0.1.1-rc.10.1.2-rc.1, 0.1.5-rc.1, 0.1.5-rc.2
Cordis^4.0.24.0.2+
Node.js>=20.0.020.x, 22.x

License

This project is licensed under the MIT License.