Case Studies
Case studies use a common declarative-package layout so dataset
selection, operator overlay, optional CIP-57 blueprints, SHACL
invariants, presentation prose, and SPARQL evidence stay separate. The
reproduce step is a documented Unix pipe in each case study's
README.md — no per-case shell scripts, no orchestrator binaries.
A case-study directory is an application's RDF asset
bundle: the overlay.yaml, blueprints/, and
shapes/ are exactly what an app developer ships to give transaction
authors a pre-sign gate and auditors a lattice classifier.
docs/case-studies/<slug>/
├── README.md # documents the reproduce pipe
├── selection.md # txid-selection method
├── selections.txt # one selected txid per line
├── overlay.yaml # operator overlay (entities, vendors,
│ # attestations, imports:)
├── blueprints/ # optional CIP-57 files for cq-rdf blueprint
├── shapes/ # optional SHACL invariants for cq-rdf shacl
├── case.md # operator-facing presentation
└── queries/
└── q-example.md # one page per query: answer, SPARQL, result
Use the template files as the starting point for new case studies:
README,
presentation,
selection,
overlay,
selections, and
query page. Keep README.md,
selections.txt, overlay.yaml, and any referenced blueprint or shape
files in the case-study directory so they ship with the built site, but
leave them out of MkDocs navigation when they aren't reader-facing.