DeepSeek Harness (DSH) adopts an “everything is a plugin” architecture. When developers use models, they often need to retrieve preprints or papers in specific domains. Manually browsing web pages is inefficient. The following section introduces a tool plugin that can directly call the arXiv API within a conversation.

What Is This

This is a web tool plugin designed for DeepSeek Harness (DSH). It registers a model-available tool named arxiv_search, retrieves paper metadata through the public arXiv API, and returns structured results. The plugin is maintained by lixvn888 and does not require API keys, a Python environment, or client build packages.

Core Features

  • Registers a tool arxiv_search for the model.
  • Queries papers through the public arXiv API.
  • Returns structured metadata including title, authors, abstract, date, categories, abs URL, and PDF link.
  • No API key, Python dependencies, or client packages are required.

Installation and Enablement

Use the following command to install the plugin:

dsh plugin --profile web add https://github.com/<your-account>/dsh-arxiv-search/archive/refs/tags/v0.1.0.tar.gz

After installation, the web server (dsh web) needs to be restarted so the host can load the new plugin. In a new conversation session, the model can directly call the arxiv_search tool.

Typical Usage

Simply ask in a natural language prompt, for example:

  • “Search for uplift modeling papers from 2024 onwards”
  • “Find recent papers about diffusion models, 5 results”
  • “Search for protein folding sorted by submission time”

The tool accepts the following parameters:

Parameter Type Description
query string (required) Search keywords, such as "uplift modeling"
maxPapers number Maximum number of results (default 10, cap 200)
sortBy string Sorting method: relevance | submittedDate | updatedDate (default relevance)

Use Cases and Notes

This plugin is suitable for developers and researchers who need real-time retrieval of academic literature. The plugin runs with the permissions of the current DSH process. Before installation, you should review the source code and license (MIT). The original skill content is adapted from a skill in the aiskillstore marketplace.

After completing the above steps, arXiv search capability can be integrated into DSH. For source code and more details, refer to the GitHub repository or community directory.