Introduction¶

In the DeepSeek Harness (DSH) ecosystem, agents frequently need to process web page data. Directly parsing raw HTML strings usually requires writing complex regular expressions or DOM manipulation logic. This plugin provides a set of ready-to-use parsing tools that can convert HTML into structured data for easier downstream processing.

Plugin Introduction¶

uckkk/dsh-html-parse is a DSH plugin implemented in Node.js. Its core purpose is to extract the title, Meta information, heading hierarchy, links, tables, and plain text from an HTML file or string. After installation, it registers its parsing capabilities into the DSH toolset.

Core Features¶

The plugin provides a tool named parse_html, which supports extracting the following structured data:
* Basic Information: title, Meta description.
* Heading Hierarchy: extracts the hierarchical relationships of headings.
* Links: extracts all links on the page.
* Tables: extracts table data from the page.
* Plain Text: extracts the plain text content of the page.

Installation and Configuration¶

To install the plugin, use the official installation command.

dsh plugin add github:uckkk/dsh-html-parse

After installation, the plugin will automatically generate a configuration file. You need to add "dsh-html-parse" to the dsh.profile.bundles field in your current Profile’s package.json.

{
  "dsh": {
    "profile": {
      "bundles": ["dsh-html-parse"]
    }
  }
}

Typical Usage¶

After installing and enabling the plugin, you can directly invoke the parse_html tool in a session. This tool accepts a file path as a parameter.

Parse this HTML file and extract the title and all links:
Call parse_html(file="page.html")

Applicable Scenarios and Notes¶

This plugin is suitable for scenarios where you need to quickly extract specific structured data from local HTML files.

Note: Installing it means running third-party code on your local machine. Please review the source code yourself.

Before use, make sure the plugin meets your environment dependencies (such as the required versions of @deepseek-ai/cordis and @deepseek-ai/dsh-tools) and complies with its open-source license (MIT).

References¶

Plugin directory: https://www.skillhub.cn/plugins/uckkk/dsh-html-parse
Source repository: https://github.com/uckkk/dsh-html-parse

uckkk/dsh-html-parse: Structured HTML Parsing