Academic Harness

A research workbench for DeepSeek Harness (DSH). Academic Harness adds literature discovery, evidence-linked paper memory, scientific writing and review tools, research ledgers, and 790 bundled skills to the DSH web UI.

It installs as a DSH plugin and provides an Academic Harness preset. Its tools support human-directed research; advisory reviews and audit results are not substitutes for expert judgment.

Quick start

Academic Harness requires Linux and DeepSeek Harness >=0.2.1-rc.5 <0.3.0. It is a Linux-only plugin: the npm package declares "os": ["linux"], and Windows and macOS are not supported. See installation requirements for validated versions and optional runtimes.

npm install -g @deepseek-ai/dsh@0.2.1-rc.5
dsh web

In the web UI:

1. Open Plugins → Add plugin.
2. Enter github:raktim-mondol/academic-harness and click Install.
3. Start a new chat and select the Academic Harness preset.

No API key is required to install or start; some connectors and compute services require their own credentials. For a dedicated CLI profile, troubleshooting, and setup details, see Install and run and Configuration.

What you can do

• Find and assess literature: search scholarly sources, retrieve papers, check claims against stored quotations, and triage prior art with explicit coverage limits. Search · Papers and evidence · Prior-art triage
• Investigate research questions: run deep research, systematic reviews, meta-analyses, and hypothesis tournaments; follow the approval-gated research cycle from scope and prior art to review and export. Research
• Write and revise: use the LaTeX writing oracle, citation tools, manuscript reviews, and reviewer-response ledgers. Writing · Deep review
• Keep work auditable: record scoped experiment comparisons, repeated-seed summaries, and read-only branch views; attribute components with ablation; and trace figures back to their declared generating runs. Analysis · Provenance · Workspace
• Extend your workflow: use scientific data tools, specialist agents, bundled skills, and remote compute. Scientific data · Agents · Skills · Compute

Documentation

The documentation home lists all guides and diagrams. Start with the links below:

Topic	Guides
Getting started	Install and run · Configuration · Architecture
Research workflow	Search · Papers and evidence · Research · Writing
Data and reproducibility	Analysis · Provenance · Scientific data · Workspace
More capabilities	Discovery and reference checks · Identity and audit tools · Compute · Agents · Skills · Local self-improvement
Reference	Tool index · Development · Changelog · Engine notes · Human-directed research cycle · ScientistTwo roadmap

License

MIT — see LICENSE. Some bundled skills have different license terms, including non-commercial terms; see NOTICE for attribution and details.