AllDSH
中文

DSH Dafeiyu

QCYTSN · QCYTSN/dsh-dafeiyu

Always-on-top desktop BigFish companion driven by real DSH session events, showing project, phase and true todo progress.

DSH Dafeiyu is a DSH plugin by QCYTSN. Always-on-top desktop BigFish companion driven by real DSH session events, showing project, phase and true todo progress. It is filed under Entertainment and UI & Experience, for Notifications. Verification level L2 (Structured), an automated security scan has not run yet, and it declares no specific compatibility target; the listing was last checked on 2026-09-18 and is published under the MIT licence. Install it with dsh plugin --profile web add dsh-dafeiyu

Listing checked by Chiyu , data updated · How to cite

L2active
DSH Dafeiyu verification panel: level L2 of 5 (Structured) — Manifest valid, key fields complete. Security scan: unverified — no scan has run. Compatibility target: DeepSeek Harness not declared. Licence: MIT. GitHub stars: 332. Listing last checked 2026-09-18.
Verification panel for DSH Dafeiyu. The listing holds level L2 of 5 — Structured: Manifest valid, key fields complete. Security scan: unverified — no scan has run. Compatibility target: DeepSeek Harness not declared. Licence: MIT. Listing last checked 2026-09-18. How verification works

In use

Screenshots taken by this directory of the plugin running in a throwaway DeepSeek Harness on a generated demo project — not of your own machine, and not supplied by the author. Click any capture for the full-size version.

DSH Dafeiyu — Installed and enabled, as the expanded row in Settings → Plugins → Plugin list: full package name, configuration state and runtime state.
Installed and enabled, as the expanded row in Settings → Plugins → Plugin list: full package name, configuration state and runtime state.

Captured by AllDSH on 2026-09-19 against DSH 0.1.5-rc.1 at a 1920×1080 viewport, package dsh-dafeiyu@^0.1.14.

Install command

dsh plugin --profile web add dsh-dafeiyu

Requires a running harness: start it withnpx @deepseek-ai/dsh webthen run the command above.New to this? See the installation guide.

What it does

DSH Dafeiyu (大肥鱼, “BigFish”) is a desktop companion that lives outside the browser: a transparent, borderless, always-on-top native window showing a whale character whose animation and status card follow your real DSH session events — thinking, working, waiting for confirmation, done, error.

Two design decisions define it. First, it is driven by agent events only: it does not read your screen, does not watch other applications, and never mistakes your work in VS Code or a browser for DSH activity. Second, it shows only what DSH actually provides — the project directory name, the current phase (analysis / implementation / verification), the in-progress todo, real progress like “3/5 steps”, and the actual reasoning effort when DSH reports it. If DSH writes no structured todo list, the card shows phases only; the README is explicit that it will not fabricate a percentage.

With several sessions running, it surfaces the one that most needs attention — waiting-for-confirmation outranks error, then working, thinking, idle — and the status bubble can list multiple active tasks at once. It is bound to the DSH Host lifecycle: DSH launches it, it stays up when you close the browser tab, and it exits when the Host truly exits.

How it runs: bundled native helpers

This is the part to understand before installing. The plugin spawns bundled native helper executables at runtime, and the author discloses this openly:

  • Windows 10/11 x64 and Linux x64 helpers are prebuilt with PyInstaller and ship inside the npm package (runtime/bin/). No Python or PySide6 install is needed, and there is no second app to launch.
  • WSL2 runs the Windows helper through Windows interop (cmd.exe) — install the plugin inside WSL, and the desktop window appears on the Windows side.
  • macOS 12.0+ gets an experimental native helper written in Swift/AppKit (Universal binary), which the README states was AI-assisted generated and merged after human review, replacing an unstable Qt/PyObjC prototype. The macOS binary is ad-hoc signed — not Developer ID signed, not notarized.

Interaction is direct: drag to move (position persists), click or double-click for short reactions, right-click for a menu with size, bubble and motion options plus “hide this session” and “close this session”. Settings live under Settings → Plugins → 插件配置 → 大肥鱼桌面伴侣 — enable toggle, character size (55–140%), bubble size and visibility, idle activity level, reduced motion, completion/error sounds, and whether sub-agent states participate (default off). Settings are stored by DSH and survive plugin updates. Note that the settings and desktop copy are currently Simplified Chinese only.

Requirements

  • A supported desktop: Windows 10/11 x64, WSL2, Linux x64 with glibc ≥ 2.35 (the binaries are built on Ubuntu 22.04; a graphical session with DISPLAY or WAYLAND_DISPLAY is required, X11/XWayland preferred, and Debian/Ubuntu desktops usually need libxcb-cursor0), or macOS 12.0+ (experimental).
  • A working DeepSeek Harness WebUI and a dsh CLI with plugin --profile web.
  • Node >=22.19, per engines.
  • No harness version is declared anywhere — no engines.dsh, no dsh.compatibility block, no README range. That absence is why this listing is verification L2: the package, manifest and docs are coherent and the plugin installs through the official profile mechanism, but with no declared compatibility range there is nothing to check your harness release against. Treat any harness upgrade as unverified until the author declares one.
  • No install-time lifecycle scripts — package.json carries test and helper-build scripts only. The native helpers ship prebuilt; nothing compiles at install time.

Privacy and boundaries

The README’s own list: it does not read or store model API keys; it does not take screenshots or read other windows; it sends no telemetry; it does not monitor keyboard input or other applications; it opens no new network ports — the settings card reuses DSH’s local web service. By default it follows only the most recently active top-level session. As with any listing here, that is author disclosure, not an audit.

Licensing: MIT code, separate asset terms

The split is easy to miss. The code is MIT, but the package.json license field literally reads SEE LICENSE IN ASSET_LICENSE.md. The current 16 character animations were selected and re-encoded from the community project dsh-pet (Copyright © 2026 PC2005-cloud) under its MIT terms, with the upstream license text shipped in the package (assets/dsh-pet-LICENSE.txt). The early BigFish frames are archived out of the package and their original restricted terms continue to apply. ASSET_LICENSE.md holds the full provenance and usage boundaries — read it before reusing the artwork.

Known limits

  • You are running prebuilt native binaries from an npm tarball — PyInstaller executables on Windows/Linux and an ad-hoc-signed, notarized-nowhere Swift app on macOS. The author discloses this; the trust decision is yours.
  • macOS support is experimental: CI-verified for the Universal build, AppKit rendering and lifecycle, but real-machine Apple Silicon experience is still being gathered through user feedback. Finder-extracting the .tgz is the one path that triggers Gatekeeper — install the tarball directly instead.
  • Linux desktop verification is narrow: only Ubuntu 24.04 (glibc 2.39) has passed real-desktop acceptance; other distros are CI smoke-tested under Xvfb at best. ARM, headless SSH, containers and servers are explicitly not display targets.
  • Progress percentages appear only when DSH writes structured todos; otherwise you get phases.
  • UI copy is Simplified Chinese only in the current builds.
  • Sub-agent status is excluded from the priority logic unless you opt in.

When to use it — and when not to

Use it if you work on a supported desktop and want ambient, glanceable DSH status while you are in other windows — the always-on-top card with real phase and todo progress is genuinely more informative than a taskbar icon, and the event-driven design means it never lies about who is working.

Do not install it on headless servers, SSH-only boxes or containers — there is nothing for it to draw on. Skip it if your policy forbids running bundled, unnotarized native binaries, if you need an English UI today, or if you want a companion that reacts to more than one agent’s world: by design, it only follows DSH.

Install

Fully quit DSH first — the Host, not just the browser tab. Then:

dsh plugin --profile web add dsh-dafeiyu

Start the DSH WebUI as usual; the companion is launched by DSH automatically. Do not start the helper by hand. Alternatives: dsh-dafeiyu@alpha for the preview channel, or the .tgz from GitHub Releases installed unextracted (dsh plugin --profile web add <path-to.tgz>) — on macOS, do not Finder-extract it first. Updates are dsh plugin --profile web update dsh-dafeiyu with DSH fully closed.

Sources & references

The technical claims on this page rest on the primary documents below. Each link is followed so it can be checked directly.

  1. [1]QCYTSN/dsh-dafeiyu — GitHubThe plugin’s own repository and README — the primary document every claim in this listing is checked against.
  2. [2]dsh-dafeiyu on npm — npmThe published package the install command on this page resolves.
  3. [3]deepseek-ai/deepseek-harness — GitHubUpstream repository for DeepSeek Harness, the Cordis runtime and the plugin loading model.
  4. [4]@deepseek-ai/dsh on npm — npmPublished harness package, including the release tags our dshTarget field tracks.
  5. [5]dsh-plugin GitHub topic — GitHubThe discovery mechanism used to find plugins listed in this directory.

All links last verified 2026-09-11

Questions about this listing

Is DSH Dafeiyu safe to install?
DSH Dafeiyu holds verification level L2 of 5 — Manifest valid, key fields complete — and its security status is unverified — no scan has run. A verification level records how far the install path was exercised; it is not a safety guarantee. Safety depends on what the plugin can actually reach, so read its source and install script and check its file, shell, network and credential access before you run it.
Which DeepSeek Harness version does DSH Dafeiyu target?
DSH Dafeiyu declares no dshTarget, so no specific DeepSeek Harness release has been committed to. The field is optional, and its absence means this listing makes no claim about working on a particular release — test it in an isolated profile first.
How do I install DSH Dafeiyu?
Start the harness with npx @deepseek-ai/dsh web, then copy the install command shown at the top of this page. That command is generated from the package name the repository publishes — or from an install spec the listing declares — so it cannot drift from what the repository actually ships. This listing was last checked on 2026-09-18.

Similar DSH plugins

Browse all

DSH Taskboard

shengsheng90

A local SQLite project taskboard for DSH with a native Web UI and eight agent tools, where only a human can accept work as done.

L2
325 TypeScript

QQ-pet-style companion floating in the DSH Web GUI — a 15-state sprite reacting to real session and task events.

L2
325 JavaScript

A whale-girl widget for the DSH Web corner that reads DeepSeek balance from the official API and books spend from deltas, stored locally.

L2
2.7k JavaScript

Deep Whale

Small-tailqwq

A whale-girl skin pack that redresses the DSH Web interface, shipped as three separately installable subdirectory packages.

L1
2.1k TypeScript

Open Sea Skin

d-dev0101

A real-time WebGPU ocean skin for DeepSeek Harness — animated waves, sunset lighting and translucent glass UI, installable as a DSH plugin, browser extension or static frontend.

L2
379 JavaScript

Renders local Wallpaper Engine wallpapers behind the DSH web UI, with an iOS-style liquid-glass settings window.

L3
301 JavaScript
ESC

Type to search the index. Built at deploy time by Pagefind.