Repolode
Submit
PC2005-cloud

PC2005-cloud/dsh-pet

A transparent animated desktop pet that runs inside the DeepSeek Harness web UI or as a separate always-on-top desktop window.

View on GitHubWebsite
Stars
2,562
Forks
168
Open issues
11
Last commit
14 hours ago

What is it?

dsh-pet is a desktop pet plugin for DeepSeek Harness (DSH). It shows a transparent animated character that idles, plays random actions, walks around the screen, reacts to clicks and drags, and can be thrown and bounce off screen edges. Right-click menus let you pick any animation directly.

Beyond decoration, the pet can reflect DSH session activity by switching animations and bubbles for thinking, working, waiting, success and error states. It can also show account balance or usage for supported providers, chat with you, and send system notifications when the window is unfocused. Multiple pets can run at once, each with its own size and position.

The pet can run in the browser, as a transparent always-on-top desktop window through Electron, or standalone without opening DSH. Developers can add new pet types with their own animation sets, and the repository also includes a pipeline that turns AI-generated green-screen videos into transparent animations.

Key features

  • 01Idle breathing, random actions, screen roaming and click or drag reactions
  • 02Balance or usage display with bubble text for supported providers
  • 03Work-status animations that follow DSH session events
  • 04Multiple pets on screen with independent size and position
  • 05Desktop mode via a transparent always-on-top Electron window
  • 06Custom pet packs and custom animations in VP9-alpha webm

Use cases

  • Adding a lively companion to the DeepSeek Harness interface
  • Keeping a small pet visible on the desktop while working in other apps
  • Creating a custom desktop mascot from your own AI-generated animations

Quick start

  1. 1

    Confirm Node.js is installed

    node -v
  2. 2

    Install the DSH launcher and pnpm

    npm install -g @deepseek-ai/dsh pnpm
  3. 3

    Install this plugin for web mode

    dsh plugin --profile web add dsh-pet
  4. 4

    Restart the DSH desktop app or dsh web so the plugin takes effect

Things to know

  • Requires the DeepSeek Harness (dsh) environment; it is a plugin, not standalone software.
  • Safari is not supported because it does not render WebM alpha transparency; macOS Safari needs extra .mov assets.
  • Balance display needs a provider API key, and unsupported providers only show a text notice.
  • Desktop mode downloads Electron in the background on first run, which can take minutes on slow networks.

Topics

#desktop-pet#electron#animation#deepseek-harness#transparent-window#webm#dsh#dsh-plugin

FAQ

What is dsh-pet?
A transparent animated desktop pet that runs inside the DeepSeek Harness web UI or as a separate always-on-top desktop window. dsh-pet is a desktop pet plugin for DeepSeek Harness (DSH). It shows a transparent animated character that idles, plays random actions, walks around the screen, reacts to clicks and drags, and can be thrown and bounce off screen edges. Right-click menus let you pick any animation directly.
Is dsh-pet free to use? What license does it have?
dsh-pet is open source under the MIT license. Check the license terms before using it in commercial products.
Is dsh-pet still maintained?
Actively maintained. The last commit to dsh-pet was on Oct 19, 2026.
What can I use dsh-pet for?
Adding a lively companion to the DeepSeek Harness interface Keeping a small pet visible on the desktop while working in other apps Creating a custom desktop mascot from your own AI-generated animations
How do I install dsh-pet?
Confirm Node.js is installed: node -v Install the DSH launcher and pnpm: npm install -g @deepseek-ai/dsh pnpm Install this plugin for web mode: dsh plugin --profile web add dsh-pet Restart the DSH desktop app or dsh web so the plugin takes effect
Where can I get dsh-pet?
The source code and installation instructions are on GitHub: https://github.com/PC2005-cloud/dsh-pet

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects