When handling academic literature, it is common to switch between multiple databases, manually download PDFs, and verify citations. litkit-dsh provides a CLI tool that uniformly handles literature retrieval, downloading, citation auditing, and export from 25+ sources.

What It Is

litkit-search is a command-line tool for academic research. It supports searching across 25+ sources (such as arXiv, PubMed, Semantic Scholar, and more), with batch retrieval and deduplication. As a DeepSeek Harness (DSH) plugin, it can also audit citations in manuscripts and export results in BibTeX, RIS, or JSON formats.

Core Features

  • Multi-source search: Covers more than 25 academic sources, including arXiv, PubMed, Crossref, Semantic Scholar, OpenAlex, DBLP, IEEE Xplore, Scopus, and Web of Science.
  • Batch processing: Supports batch retrieval and deduplication, and returns ranked results.
  • Citation auditing: Uses litkit-dsh verify to check whether references in a document exist in connected sources.
  • PDF download chain: Prefers open-access sources, then attempts institutional access (requires configuration), and finally uses shadow libraries as an optional fallback.
  • Format export: Supports exporting search results to BibTeX, RIS, or JSON.
  • Workflow support: Provides preset workflows such as topic-search, deep-search, and bulk-review.
  • Chinese literature: Provides the zh-search command for searching Chinese literature.
  • Agent-ready: Provides MCP (Model Context Protocol) server support and a DeepSeek Harness (DSH) plugin adapter.

Installation and Enablement

This plugin is distributed via GitHub. Before installing, make sure your environment meets the following requirements:

  • Node.js >= 20
  • peerDependencies: @deepseek-ai/cordis and @deepseek-ai/dsh-tools

Run the following command to install:

dsh plugin --profile web add github:bpc-oss/litkit-search

Typical Usage

After installation, you can operate through the CLI.

  1. Search literature and export:
    Search for a specific topic and export the results in RIS format.
    litkit-dsh search "deep learning drug discovery" --limit 20 --export ris
  1. Citation auditing:
    Audit citations in a local document (such as docx or pdf) and output the report to a specified directory.
    litkit-dsh verify paper.docx -o report/
  1. Environment check:
    Check whether the current environment configuration is correct.
    litkit-dsh doctor

Configuration and Notes

  • API keys: Most sources work without API keys. Some sources (such as Scopus and Web of Science) require interactive configuration via litkit-dsh sync-keys or key setup in a .env file to improve rate limits and metadata quality.
  • Institutional access: To access institutionally subscribed PDFs, set INSTITUTIONAL_PROXY or INSTITUTIONAL_DIRECT in the .env file.
  • Chinese literature search: The zh-search feature requires institutional library access to be configured (for example, Shenzhen University Library).
  • Shadow libraries: The project includes shadow libraries such as Sci-Hub, LibGen, and Anna’s Archive as optional fallback download paths. These paths are optional; use them at your own risk and only access content you are authorized to use.

Summary

litkit-dsh suits researchers or developers who need to quickly retrieve and organize large volumes of academic literature locally. It provides a complete closed loop from search to verification via CLI and is deeply integrated into the DeepSeek Harness ecosystem, making it easy to call in Agent workflows.