In the DeepSeek Harness (DSH) ecosystem, extending functionality through plugins is common. The standard DeepSeek model does not have vision capabilities and cannot directly process images. dsh-mindsee is a community plugin that acts as an intermediary layer. It uses MindSee services to pass image descriptions to DeepSeek, enabling the model to “see” screenshots or local images without switching to a dedicated vision model.

Installation and Enablement

Before installing, make sure DeepSeek Harness is installed and the web profile is activated. Run the following command to install the plugin:

dsh plugin --profile web add dsh-mindsee

After installation, restart the dsh web process to load the new plugin.

Configure an Access Token

To use this plugin, you first need to obtain an access token for MindSee. Follow these steps:

  1. Create an access token on mindsee.app.
  2. In DeepSeek Harness, open Settings → Plugins → MindSee.
  3. Paste the token and save.

Core Features

The plugin supports JPEG, PNG, and WebP formats, with an image size limit of 20 MB. In conversations, DeepSeek does not need to switch to a dedicated vision model. Instead, it uses MindSee to describe images and answers based on those descriptions.

Typical Usage

There are two ways to have DeepSeek process images:

  1. Paste an image: Paste a screenshot directly into the chat input box and enter your question in the same message.
  2. Enter a path: Enter the file path of a local image directly (for example, /path/to/image.png). The system will read it automatically.

Notes

This plugin runs with the permissions of the current DSH process. Before installing community plugins, it is recommended to check their source code and license (MIT).