fairscape-artifacts

Turns a finished RO-Crate into pages people can read: an HTML datasheet, an interactive evidence graph, a preview of every entity, and an AI-Ready review. It runs offline on a crate folder. You don’t need a server.

It is the view step of FAIRSCAPE. Run it once your crate is built with fairscape_models or fairscape_conversion.

Install

pip install git+https://github.com/fairscape/fairscape_artifacts

Example

fairscape-artifacts all path/to/crate

• Directorypath/to/crate/

  • ro-crate-metadata.json the crate, unchanged
  • ro-crate-datasheet.html the datasheet
  • ro-crate-preview.html every entity in the crate, grouped by kind
  • ro-crate-evidence-graph.html interactive provenance graph
  • ro-crate-evidence-graph.json the same graph as data
  • ai-ready-review.html the AI-Ready review, if the grader is installed

Each HTML file is self-contained, so you can email it, put it in a Zenodo deposit or serve it from GitHub Pages.

What they look like

The three pages, built from the Snakemake variant-calling tutorial crate.

Datasheet

Datasheet Open full page

Title, authors, license and summary at the top; then every dataset, software and computation, with an AI-Ready section when the grader is installed.

Evidence graph

Evidence graph Open full page

Every entity is a node; edges are usedDataset, usedSoftware and generated. Drag to pan, scroll to zoom, click a node for its metadata.

Preview

Entity preview Open full page

Every entity in the crate, grouped by type, with all its properties. The page an RO-Crate viewer would show.

Commands

Command	What it does
all	Builds the evidence graph, the review and the datasheet in one pass.
datasheet	Builds the datasheet and preview pages.
evidence-graph	Builds the graph page and its JSON.
review	Runs the AI-Ready grader and writes the review page.
add-io writes to the crate	Writes EVI:inputs / EVI:outputs onto the crate root. This is the only command that changes the crate.
interpret	Has an LLM read the crate’s provenance and code, then writes an annotated summary. all never runs it.

Outputs are written next to the crate unless you pass -d DIR.

Details

• AI-Ready review. Scores come from the AIreadiness-grader. Install it with pip install "fairscape-artifacts[review]". Without it, every other command still works and the datasheet leaves the review section out.
• LLM interpretation. Install pip install "fairscape-artifacts[interpret]" and run fairscape-artifacts interpret CRATE --model anthropic:claude-haiku-4-5-20251001 --api-key $ANTHROPIC_API_KEY.
• Releases and linked crates. A release crate gets one card per sub-crate. A crate linked to an upstream crate gets one evidence graph that spans both. Use --reference OTHER_CRATE to look up ids in a related crate that this one doesn’t point to.

Evidence graph across two linked crates Open full page

A one-rule Snakemake run that counted the variants the pipeline above called. Its crate was linked to the upstream crate, so the graph follows the input table back through the whole variant-calling pipeline.

How the graph is built, what the datasheet contains and how PROV-only crates are handled: docs/DETAILS.md.

Source: github.com/fairscape/fairscape_artifacts