把 DeepSeek Harness 的会话、消息、工具活动与每轮结果实时推送到 JARVIS Mission Control 看板,附任务状态与人工复核。
安装
# npm 包(预构建)
dsh plugin --profile web add dsh-plugin-mission-control
# GitHub 源码(首次需按提示配置 allowBuilds 构建授权后重试)
dsh plugin --profile web add github:Asif2BD/JARVIS-Mission-Control-DeepSeek#path:/integrations/deepseek-harness/dsh-plugin-mission-control
装任何插件都等于在你的机器上跑第三方代码,权限和你本人一样大——能读你的文件、用你的凭据、访问网络,工具审批管不到它。GitHub 来源的插件还会在安装时执行构建脚本——pnpm 默认拦截,所以安装可能停在 ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED 或 ERR_PNPM_IGNORED_BUILDS;dsh 会打印出需要添加的确切键名,把它加进该 profile 的 pnpm-workspace.yaml 的 allowBuilds 下,重跑一次即可装上。放行构建本身就是一次信任判断:请只安装可信来源,并尽量锁定 commit(github:owner/repo#sha)。
README
该插件的 README 只有英文版本。
Mirror DeepSeek-Harness (dsh) agent activity onto a live JARVIS Mission Control board: every dsh session becomes a task, messages and tool calls stream into the activity log, and finished turns land in REVIEW for a human to approve.
Validated against @deepseek-ai/dsh@0.1.0-rc.7 — both the shipped type definitions and a live headless run. dsh is a developer preview; event names are config-overridable so upstream renames are a config fix.
Install
# Into a dsh profile — the plugin self-activates via its dsh.bundle declaration.
# Pin the currently validated dsh preview, or use `dsh plugin ...` if already installed.
npx @deepseek-ai/dsh@0.1.0-rc.7 plugin --profile web add dsh-plugin-mission-control
# Point it at your board (defaults shown)
export MISSION_CONTROL_URL=http://localhost:3000
export MC_AGENT_ID=agent-dsh
export MC_AGENT_TOKEN=... # only if your board enforces auth
npx @deepseek-ai/dsh web
Need a board? git clone https://github.com/Asif2BD/JARVIS-Mission-Control-DeepSeek && cd JARVIS-Mission-Control-DeepSeek/server && npm install && npm start → http://localhost:3000
What it does
| dsh event | Board effect |
|---|---|
session/created |
Task created (IN_PROGRESS, assigned to the harness agent) |
user/message, assistant/message |
Activity log excerpts |
tool/call, tool/result |
Activity log entries (tool name; error code on failure) |
turn/end → completed |
Task → REVIEW (humans approve to DONE) |
turn/end → blocked / failed |
Task → BLOCKED, structured error logged |
turn/end → aborted / interrupted |
Task → ASSIGNED |
session/flush |
Awaited bounded flush of the plugin's HTTP queue |
All Mission Control calls go through a serialized retry queue — an unreachable board never blocks the dsh agent loop.
Configuration
Override the bundle defaults per profile by addressing row id mission-control in the profile's cordis.patch.yml:
- id: mission-control
config:
missionControlUrl: http://localhost:3000
agentId: agent-dsh # must match ^[a-zA-Z0-9-_]+$
agentName: DeepSeek Harness
authToken: null # or set env MC_AGENT_TOKEN
maxExcerpt: 280 # max chars of content forwarded to logs
flushTimeoutMs: 10000 # bound for the awaited session/flush drain
events: {} # override dsh event names after upstream renames
Troubleshooting
- Nothing on the board — confirm the server is up (
curl $MISSION_CONTROL_URL/api/metrics) and check dsh logs for[mission-control]lines. SetdryRun: truein the row config to print payloads instead of sending. - Unknown event warnings after a dsh upgrade — map the renamed events via the
eventsconfig block. - Auth failures (401/403) — set the same token in the server env (
MC_AGENT_TOKEN) and the plugin (authTokenor env).
License
MIT — part of JARVIS-Mission-Control-DeepSeek.
链接
同类插件
loopx-project/loopx#dsh-loopx-plugin★ 6242
LoopX——面向长周期 Agent 的提供商中立、本地优先状态内核与控制平面:在 DeepSeek Harness 执行层之上持久化 Goal、Todo、门禁、证据、配额、恢复与交接状态;插件负责引导安装 CLI 与技能、准入有界的同会话续跑,并为精确绑定的工作循环提供本地 GoalBar。
Q00/ouroboros#integrations/dsh-plugin★ 6203
通过 DSH MCP 客户端挂载 Ouroboros 的纯配置包,在 DSH 中提供 36 个涵盖需求访谈、Seed、执行、评估与演化流程的工具。
chuspeeism/dashi-taskboard#deepseek-harness★ 3333
把当前已安装并运行中的 Codex Taskboard 嵌入 DeepSeek Harness 侧边栏,并通过 Launcher 运行时描述文件连接,而不是使用固定端口。
NanmiCoder/dsh-agent-teams★ 2035
AgentTeams 多智能体团队。
EthanYoQ/AI-Novel-Writer#dsh-ai-novel-writer★ 1411
安装专用 AI 小说创作预设与工作台:提供带修订号的本地项目资产、紧凑侧边工作台,以及需要原生审批的逐文件变更。
tong-io/tongflow#dsh-tongflow★ 1045
基于 TongFlow 的“片场”插件,用于图片、配音、音乐与视频制作:agent 为每个资产生成 TongFlow 工作流文件(.tongflow.json)并通过 TongFlow 插件执行,内嵌工作流画布,按镜头/角色/take 组织项目,附漫剧模板;以 @tongflow 开头的会话进入 Studio 界面。
社区评论
评论公开保存在 GitHub Discussions。加载评论会连接 GitHub 和 Giscus;发表内容需要 GitHub 账号。