Skip to content

omdsh-dev/dsh-lark

231Last commit Aug 15, 2026

dsh-lark DSH plugin

dsh-lark is a plugin that integrates DeepSeek Harness with Feishu/Lark, allowing you to send tasks to your DSH agents directly from messaging. It supports persistent sessions, workspace switching, model selection, and human-in-the-loop approvals. You can also add multiple bots to a group chat for collaborative workflows.

How to install the dsh-lark DSH plugin

dsh plugin --profile web add dsh-lark-channel@latest

Copying does not run this command. Review the repository and version before installing the dsh-lark DSH plugin.

dsh-lark DSH plugin data source

dsh-lark DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-lark DSH plugin can do

  • Persistent sessions that resume after restart, with the ability to start fresh using /new.
  • Multi-workspace support: view and switch workspaces via /ws and /cd commands.
  • Human-in-the-loop cards for answering questions, approving plans, and allowing or denying tool calls.

Where the dsh-lark DSH plugin fits

  • Check build failures: Ask the agent to analyze a project and provide a plan before executing.
  • Switch between projects: Use /cd to change workspace without leaving the chat.
  • Collaborate with multiple agents: Add a reviewer bot to a group for code review tasks.

Who the dsh-lark DSH plugin is for

  • Developers who use DeepSeek Harness and want to interact via Feishu/Lark.
  • Teams that need to collaborate on agent tasks through chat.

dsh-lark DSH plugin limitations

  • Requires DeepSeek Harness 0.1.0-rc.6 or newer.
  • Native thinking process requires Feishu PC 7.70+ or mobile 7.74+; older clients need to use output: 'stream'.

dsh-lark DSH plugin: from the repository README

Quoted from the omdsh-dev/dsh-lark README, the upstream source of the dsh-lark DSH plugin. Copyright remains with the original authors.

简体中文 | [English](README.en.md) **把你正在使用的 DeepSeek Harness(DSH)接进飞书。** 直接在聊天里给 Agent 派任务、看执行过程、切换工作区和模型。遇到提问、计划确认或工具审批,也不用回到终端,直接在飞书里处理。需要时,还能把多个 Agent 放进同一个群里协作。 ## 快速开始 ```sh npm i -g dsh-lark-channel dsh-lark-channel start ``` 终端会显示二维码。用飞书扫码完成应用创建,然后私聊机器人,或在群里 @ 它即可开始。 不想装到全局也可以直接跑,只是之后每条命令都要带 `npx`: ```sh npx dsh-lark-channel@latest start ``` 如果还没有安装 DeepSeek Harness,先运行: ```sh npm i -g @deepseek-ai/dsh ``` 无需公网服务器,也无需配置回调地址。 ## 为什么值得装 - **不用守着终端**:从飞书发起任务,随时查看进度和结果。 - **不只是聊天机器人**:可以切换真实工作区和模型,执行 Harness 已有的命令与工具。 - **关键决定仍由你控制**:模型提问、计划审阅和工具审批都会回到当前聊天,按钮或文字都能作答。 - **上下文不会混在一起**:不同聊天、话题和工作区可以保留各自的会话。 - **Agent 之间也能协作**:一条命令添加更多机器人,让它们在群聊中通过 @ 交接回合,并用轮数上限防止无限对话。 ## 可以这样开始 先看看当前状态和可用工作区: ```text /status /ws /cd my-project /model ``` 然后直接派一个任务: ```text 检查这个项目为什么构建失败。先给我计划,需要操作时让我确认。 ``` Agent 的执行过程会显示在飞书中;需要你参与时,会发送提问、计划或审批卡片。渠道自带文案会按每位读者的飞书语言显示中文或英文。 ## 主要能力 | 能力 | 使用体验 | |---|---| | 持久会话 | 重启后可以恢复;后续消息继续当前上下文,`/new` 可以原地重开一个 | | 多工作区 | `/ws` 查看、`/cd` 切换;回到原工作区时继续之前的任务 | | 模型切换 | `/model` 打开模型选择卡片;切换后保留当前会话,也可随时恢复默认模型 | | 原生执行过程 | 在飞书中查看推理、工具调用和结果,最终答案单独发送 | | 人机协作卡片 | 单选、多选或文字回答问题;批准计划或提出修改意见;允许或拒绝工具调用 | | 实时状态 | `/status` 展示工作区、模型和 session;可用时还显示上下文占用与累计 tok

Read the full READMERepository license: BSD-3-Clause

dsh-lark DSH plugin questions

How to install dsh-lark?

Run `npm i -g dsh-lark-channel` to install globally. Then use `dsh-lark-channel start` to start the bot. If you prefer not to install globally, you can use `npx dsh-lark-channel@latest start` instead.

Do I need a public server to use dsh-lark?

No, you don't need a public server or to configure any callback URLs. The plugin connects to Feishu/Lark via a WebSocket, so it works behind firewalls as long as you have internet access.

How do I switch workspaces in the chat?

Use the `/ws` command to see all available workspaces. Then use `/cd <workspace-name-or-path>` to switch to the desired workspace. The current session will be maintained when you return to the original workspace.

Can I add multiple bots to a group chat for collaboration?

Yes, you can add multiple agent instances by running `dsh-lark-channel add <name>`. Each bot has its own settings and sessions. In a group, they can pass turns by @mentioning each other, with a default limit of 6 consecutive bot rounds.

How do I upgrade dsh-lark to the latest version?

Run `dsh-lark-channel upgrade` to update the CLI and restart the bot with the new version. If you use npx, you don't need to upgrade manually; just run `npx dsh-lark-channel@latest start` again to get the latest version.