Introduction

The common practice for e-commerce visual tasks in DeepSeek Harness (DSH) involves keeping conversations within the Harness while generating images, changing backgrounds, or performing virtual try-ons in separate tools. The results must then be downloaded and uploaded back into the conversation. Additionally, prompts often struggle to precisely target “this product image on the canvas” or “this reference image.”

WeShop for DeepSeek Harness embeds WeShop’s e-commerce visual workspace directly into the Harness: conversations and an infinite canvas are displayed on the same screen. After selecting assets, you can reference them using @, and generated results automatically appear on the canvas. This article introduces the positioning, capabilities, installation, and typical usage of this plugin.

What It Is

weshop-dsh-plugin is a native DSH Cordis plugin (client-side category) released by the maintainer weshopai. The current version is 1.0.2, under the MIT license. It targets DeepSeek Harness v0.1.0-rc.8 and later versions. It is installed via the Harness native dsh plugin workflow, automatically creates a Web profile, and enables the plugin bundle.

In one sentence: It brings WeShop’s e-commerce visual workspace into Harness, allowing you to complete product selection, image generation, editing, and batch processing synchronously with conversations on an infinite canvas.

Core Features

The following capabilities are derived from the official README and package.json and are available immediately after installation.

Native Plugin Architecture

The plugin runs as a Cordis Host and browser-side component, without relying on MCP subprocesses. The package includes the Host entry (lib/index.js), canvas UI (lib/client.js), presets (presets/weshop-canvas/), and bundled Skills (skills/).

Infinite Canvas

Supports box selection, multi-select, group movement, undo, download, deletion, zooming, and panning. Generated images, videos, audio, and text are automatically published to the canvas, eliminating the need for manual download and re-upload.

Synchronized Conversations and Canvas

The Harness conversation and canvas state remain synchronized. After selecting a product, model, or reference image on the canvas, you can use @ in the chat to reference the selected items, bringing spatial selections into natural language instructions.

Presets and Skills

Upon first activation, the WeShop Canvas preset is automatically installed, along with the weshop-canvas agent preset and WeShop OpenAPI-related Skills, ready to use out of the box.

API Key and Security

You can configure the API key by selecting Configure API Key in the canvas top bar, or inject it via the environment variable WESHOP_API_KEY before startup. The key is stored locally by the Harness Host with restricted file permissions and is not returned to the browser, canvas state, or model.

Interface and Ecosystem

The interface supports both Chinese and English, with automatic language detection and manual switching options. The plugin aligns with the sidebar and skin conventions of dsh-web-ui; it can also be installed via the community plugin marketplace dsh-market under Settings → Plugin Market.

Installation and Activation

Before installation, please ensure Node.js and DeepSeek Harness are installed. WeShop is a Harness plugin, not a standalone application.

First-Time Installation

Close Harness and execute:

npx @deepseek-ai/dsh plugin --profile web add weshop-dsh-plugin

This command installs WeShop, creates the Harness Web profile if needed, and automatically enables the bundle. No GitHub account, access token, or manual configuration file changes are required.

Then restart Harness:

npx @deepseek-ai/dsh web

Create or open a task, and select the WeShop Canvas preset to enter the canvas.

Installation via dsh-market

If you are already using the community plugin marketplace, first install the marketplace plugin:

dsh plugin --profile web add dshmarket

After restarting Harness, open Settings → Plugin Market, search for WeShop, and install it.

Updating

Close Harness and execute:

npx @deepseek-ai/dsh plugin --profile web update weshop-dsh-plugin

If installed via npm to the Web profile, you can also execute pnpm update weshop-dsh-plugin in the ~/.dsh/profiles/web directory.

When upgrading from an older version of WeShop or Harness, you can first start npx @deepseek-ai/dsh web once and close it, then run npx weshop-dsh-plugin setup to clean up any residual .tgz or old package name installations before executing the above update command.

Configuring WeShop OpenAPI

Select Configure API Key in the canvas top bar and enter the key obtained from WeShop OpenAPI.

Alternatively, pass it via an environment variable before starting Harness:

export WESHOP_API_KEY="your-key"
npx @deepseek-ai/dsh web

Typical Use Cases

The workflows listed in the official README include:

  1. Converting product images into clean main images or lifestyle scene images.
  2. Using virtual try-on to dress models in clothing and accessories.
  3. Completing background replacement, expansion, or cleanup within the conversation.
  4. Using a single natural language instruction to batch complete photography, editing, and short video production.

Operation path: Open the WeShop Canvas preset → Select the product, model, or reference image on the canvas → Describe the desired result in the conversation (using @ to reference selected items) → Wait for the generated result to automatically appear alongside the canvas.

Applicable Scenarios and Notes

Who is it for: Developers and operators working on e-commerce main images, scene images, virtual try-ons, background processing, or batch visual output in DSH who wish to reduce tool switching and file transfers.

Running Permissions: The plugin runs with the current DSH process permissions, with access to local files and network. Before installation, review the weshopai/weshop-dsh-plugin source code and MIT license to ensure compliance with your security requirements.

Other Notes:

  • Requires DeepSeek Harness v0.1.0-rc.8 or later.
  • Canvas data is saved in the local browser and is not included in the installation package; API keys and generated assets need to be configured separately on each machine.
  • The plugin is published to npm as a .tgz file (weshop-dsh-plugin) and can be installed on any machine with a configured Harness Web profile.

Links