poplarity/dsh-science-workbench
30Last commit Aug 14, 2026
dsh-science-workbench DSH plugin
This plugin blends Jupyter's inline figures, an agent as execution engine, and Nextflow-style provenance into a single workbench. It provides tools to run self-contained code cells, attach structured feedback to figures, and redraw derived versions, all tracked in a plain-text manifest.json with SHA-256 hashes and automatic git versioning.
How to install the dsh-science-workbench DSH plugin
dsh plugin --profile web add dsh-science-workbenchCopying does not run this command. Review the repository and version before installing the dsh-science-workbench DSH plugin.
dsh-science-workbench DSH plugin data source
dsh-science-workbench DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-science-workbench DSH plugin can do
- Run self-contained code cells and automatically discover figures in the output
- Attach structured feedback to any artifact and trigger a redraw as a derived version
- Record full provenance: cell, SHA-256 hashes, parameters, seed, and lineage across reruns
- Auto-commit every step to a local git repository with no manual push needed
- Cross-platform host shell: uses bash on macOS/Linux and PowerShell on Windows
Where the dsh-science-workbench DSH plugin fits
- Bioinformatics analysis: e.g., plotting TSS enrichment profiles and iteratively adjusting colors or parameters
- Iterative figure refinement: get feedback from reviewers or yourself, then ask the agent to redraw with modifications
- Reproducible research: keep a complete audit trail of every figure, code, and input, so results can be replayed at any time
Who the dsh-science-workbench DSH plugin is for
- Bioinformaticians and computational biologists who need traceable, replayable analyses
- Data scientists who want a structured code-figure-feedback loop without leaving the DSH agent environment
dsh-science-workbench DSH plugin limitations
- Requires a running DSH web instance with the plugin installed and configured
- Only supports a limited set of figure formats (PNG, JPEG, SVG, PDF, TIFF, BMP) – other formats may not be previewed inline
dsh-science-workbench DSH plugin: from the repository README
Quoted from the poplarity/dsh-science-workbench README, the upstream source of the dsh-science-workbench DSH plugin. Copyright remains with the original authors.
English | [中文](README.zh.md) [](https://www.npmjs.com/package/dsh-science-workbench) [](./LICENSE) [](https://github.com/deepseek-ai/deepseek-harness) A **reproducible science workbench** plugin for the [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness). It blends the best of three worlds: - **Jupyter** — cells and inline figures you can see and re-run; - **Claude Science** — an agent as the execution engine; - **Nextflow / nf-core** — every artifact carries full provenance. > **Core promise**: every figure and artifact is traceable and replayable. You can always answer *“it = which code + which inputs + which environment + which params/seed”*, and re-run it in one click. --- ## ✨ Features - **Code → figure → feedback → redraw** — the agent runs a self-contained cell to produce figures shown inline; you attach structured feedback to a figure, and `bio_rerun_cell` regenerates a derived version (`v1 → v2 → v3`). - **One ledger per project** — a plain-tex
Read the full READMERepository license: MIT
dsh-science-workbench DSH plugin questions
How do I install dsh-science-workbench?
Run `dsh plugin --profile web add dsh-science-workbench` in your terminal. This command installs the package from npm and automatically adds it to your DSH profile bundles. After installation, restart `dsh web` to activate the plugin.
Can I use this plugin without DSH?
No, it is designed exclusively for the DeepSeek Harness environment. The plugin registers agent-facing tools and a workbench UI that rely on DSH's host and client infrastructure. You must have DSH installed and running to use it.
What happens when I rerun a cell?
Rerunning a cell with `bio_rerun_cell` creates a derived version (e.g., cell_0001_v2). The new cell is recorded in the manifest with a link to its parent cell, and the figures it produces are given new names. The original cell and its outputs remain intact, so you can always trace back.
How does the plugin ensure reproducibility?
Each cell is run as a self-contained script in a fresh subprocess with a fixed seed. The host records the script, SHA-256 hashes of all inputs and outputs, parameters, environment lock, and creates a git commit after every step. This provides a complete audit trail that can be replayed later.
Which file formats are supported for inline figure preview?
The workbench can preview PNG, JPEG, SVG, PDF, TIFF, and BMP images inline. Other formats may still be generated by your code but will not appear in the artifact preview panel. You can download them from the file system or use the 'open in Finder' action.