Skip to content
dsh.fish
Bundle

@onpeople/dsh-academic-research

Evidence-grounded bilingual research tools and workflows for DeepSeek Harness.

Source
userInner
stars
2 stars
License
MIT
Updated
Updated 15 hours ago

Readme

# `@onpeople/dsh-academic-research`

An independent `dsh-plugin` for DeepSeek Harness and OnPeople. It adds a bilingual, evidence-grounded academic research collaborator without requiring an account or a licensed database connection.

The plugin provides:

- literature discovery through public Crossref and OpenAlex APIs;
- retrieval of a specific public HTTPS research record or paper page;
- source status and coverage information;
- evidence-aware bilingual literature review, academic writing, citation checking, and reviewer-response guidance;
- research-integrity rules that prohibit fabricated sources, data, quotations, experiments, and peer-review claims.

The package contains both the standard `.codex-plugin/plugin.json` metadata and a native DeepSeek Harness Cordis plugin. The `.onpeople/industry.json` companion lets OnPeople expose the same package as a selectable industry plugin; the DSH runtime uses `src/index.ts` and `cordis.patch.yml`.

## Install in DeepSeek Harness

The package is designed for profile-scoped installation. From a DSH checkout or an installed `dsh` command:

```sh
dsh plugin --profile web add @onpeople/dsh-academic-research
```

Until the package is published, install a local checkout instead:

```sh
dsh plugin --profile web add /absolute/path/to/dsh-academic-research
```

For a GitHub release, publish this directory as its own repository and add the
`dsh-plugin` topic so it can be discovered with the rest of the ecosystem.

After installation, restart the profile. The package declares a `dsh.bundle` patch, so the profile automatically mounts the plugin row. It is not enabled in unrelated profiles.

## Use it

Ask the agent directly, for example:

> Search 2020–2024 English papers on graph neural networks for recommender systems. Build an evidence matrix with title, authors, year, DOI, source URL, and evidence limits.

The plugin's tools are:

| Tool | Purpose |
| --- | --- |
| `research_search` | Search Crossref and OpenAlex public metadata. |
| `research_fetch` | Read one allowlisted public HTTPS record or page. |
| `research_source_status` | Describe the public source coverage and limitations. |

The search tools are read-only and do not accept credentials. Public metadata, abstracts, and full-text links are reported as different evidence levels; a link is never treated as proof that the full paper was read.

## Development

```sh
npm install
npm run build
npm pack
```

The package intentionally keeps provider calls in the plugin and uses only standard `fetch`; no custom UI is required. A host can compose it with its existing tool presentation, Skills, permissions, approvals, and model settings.

## Compatibility

DeepSeek Harness is in developer preview and its plugin APIs may change. This package targets the profile-bundle contract available in the current `master` line and keeps the host-specific metadata declarative so a future manifest change can be handled without changing the research workflow content.

## License

MIT. See [LICENSE](LICENSE).

Install

dsh plugin --profile web add github:userInner/dsh-academic-research

Profile: web

  • This package builds from source on install. pnpm will ask you to allow its build script — that is permission to run the package’s code on your machine, outside the agent sandbox. Only allow sources you trust.
  • This source has no pinned commit, so a later push upstream changes what installs. Prefer pinning a commit.
Source