[Plugin] | LazyMind | Visualize and Control Long-Running Agent Workflows

[ADAM-CHENG666]

English

Hi everyone! 👋

I've recently developed a plugin for DSH that integrates [LazyMind](https://github.com/LazyAGI/LazyMind), an open-source project that brings visual workflow execution and greater control to complex, long-running agent tasks.

Why did I build this?

Today's agents are increasingly capable of planning and executing complex tasks. However, as tasks become longer and involve more steps, it becomes harder for users to understand and control what's happening during execution.

For example:

• Limited visibility: Which step is currently running? What intermediate results have been produced?
• Difficult intervention: When a step doesn't produce the expected result, users often have to repeatedly guide the agent through conversation.
• Costly recovery: A failure midway through a long-running task may require rerunning work that has already been completed.
• Hard-to-track outputs: After multiple revisions and retries, managing different versions of generated artifacts can become challenging.

DSH_discussion.mov

My goal is to preserve the existing DSH agent experience while making complex workflows more transparent and controllable.

What does LazyMind bring to DSH?

① Workflow Visualization

• Clear execution progress: Visualize workflow structure, progress, and execution status.
• Inspect intermediate results: View step-level outputs and generated artifacts as the workflow runs.

② Execution Control

• Human-in-the-loop intervention: Inspect and edit intermediate artifacts to guide workflow execution.
• Artifact versioning: Keep track of previous versions for comparison, editing, and rollback.

The core idea: Turn complex agent tasks from black-box executions into workflows that users can observe, intervene in, and recover.

How to integrate with DSH?

LazyMind is a standalone, open-source application. The integration combines the DSH plugin system + MCP to bring LazyMind's workflow capabilities into DSH, without replacing the existing DSH agent.

Getting started is simple:

1. Install LazyMind and configure your preferred model API.
2. Open the Assistant Settings in LazyMind.
3. Enable DSH integration with one click, and you're ready to use LazyMind's workflow capabilities from DSH!

Project

GitHub: [LazyAGI/LazyMind](https://github.com/LazyAGI/LazyMind)

Installation, configuration, and usage instructions are available in the repository.

Note: LazyMind is an independent open-source project. This DSH integration plugin is community-developed and maintained, and is not an official DSH feature.

Feedback Welcome!

This integration is still evolving, and I'd love to hear your thoughts!

When working with complex, long-running agent tasks, have you encountered challenges with limited visibility, difficulty intervening, or having to restart after failures?

If you have similar use cases, feature suggestions, or ideas about the DSH integration, feel free to share them in the comments.

Thanks for checking it out!

---

中文版

大家好！👋

最近，我为 DSH 开发了一个插件，可以将开源项目 [LazyMind](https://github.com/LazyAGI/LazyMind) 集成到 DSH 中，为复杂、长时间运行的 Agent 任务提供可视化的工作流执行与控制能力。

为什么需要它？

如今 Agent 已经能够自主规划并执行相当复杂的任务，但当任务执行时间变长、步骤增多时，用户往往很难方便地掌控整个执行过程。

例如：

• 过程不透明：任务执行到了哪一步？中间产生了什么结果？
• 干预不方便：某个步骤的结果不符合预期，往往需要通过对话反复调整。
• 恢复成本高：任务中途失败，可能需要重新执行大量已经完成的工作。
• 结果难追踪：多次修改和重试后，不容易管理不同版本的产物。

我希望在保留 DSH 原有 Agent 交互体验的同时，让复杂任务的执行过程更加直观、可控。

LazyMind 带来了什么？

DSH_discussion.mov

① 工作流可视化

• 执行进度一目了然：展示工作流的执行结构、进度与状态。
• 中间结果随时查看：实时查看各步骤的执行结果与中间产物。

② 执行过程可控

• 人工干预：直接查看、编辑中间产物，调整工作流执行。
• 产物版本管理：保留产物历史版本，方便比较、修改和回退。

核心目标：让复杂 Agent 任务不再只是一次性的黑盒执行，而是一个可观察、可干预、可恢复的过程。

如何接入 DSH？

LazyMind 是独立运行的开源应用，通过 DSH 插件系统 + MCP 将工作流能力接入 DSH，无需替换现有的 DSH Agent。

接入非常简单：

1. 安装 LazyMind，并配置适合自己的模型 API。
2. 打开 LazyMind 的助理配置界面。
3. 一键启用 DSH 集成，即可在 DSH 中使用 LazyMind 的工作流能力。

项目地址

GitHub：[LazyAGI/LazyMind](https://github.com/LazyAGI/LazyMind)

安装、配置和使用说明可以在项目仓库中查看。

说明：LazyMind 是独立开源项目，本次分享的 DSH 集成插件由社区独立开发与维护，并非 DSH 官方提供。

欢迎反馈！

这次集成仍在持续完善，我也很希望听听大家的建议。

对于复杂、长时间运行的 Agent 任务，你是否也遇到过过程不透明、难以干预，或者失败后需要重新执行的问题？

如果你对可视化工作流、执行控制或 DSH 集成方式有任何建议，都欢迎在评论区交流！

感谢大家！

[rzsitrighthere]

Dsh 做 PPT 没有局部编辑功能哎，后期优化没法圈住某些东西进行优化，还挺麻烦的。而且产出的是html，还要转成ppt。加了 LazyMind 这个插件真的好用了不少，可以局部编辑，也可以直接导出原生可编辑PPT，真的很不错。

[xiaocaie]

dsh跑长任务不知道卡在哪，接了LazyMind可以看workflow，确实可以

[nora-harper]

感觉dsh接上lazymind插件，做ppt真的好用了不少哎。

[ADAM-CHENG666]

目前有用户反馈windows环境下运行异常，大家有遇到类似情况吗？

[TangyRyan]

可以可视化workflow诶，感觉可以