Inspect the knowledge behind the publication
GraphArch collects source statements, authored fragments, generated explanations, relationships, publication hints, and provenance into an RDF knowledge graph. A generated page or PDF is a view of that graph, rather than a separate copy of the documentation's meaning.
Query content and provenance
Use the publication CLI's SPARQL query operation to inspect the KG. You can ask which fragments explain a resource, which source supplied an annotation, which components a target selected, or which quality findings remain in a revision. The publication pipeline guide shows the query, planning, archive, and rendering commands.
Source contributions remain separate. Git inputs retain the selected commit and file location; local working-tree inputs identify the content actually read. RDF identities, authored text, and generated explanations stay connected instead of being flattened into an untraceable HTML export.
Publish an exact revision
Assembly accepts a complete source revision. Planning selects a publication's content and order against that revision. Rendering then reads the pinned plan and KG, so a source edit during rendering cannot silently change part of the result.
A later build can accept new sources while earlier plans remain readable. Specifications, companion sites, and PDFs can share an input revision while retaining their own content selection and arrangement.
Retain and restore the graph
Export the KG and restore it into an empty store to render without reopening the original files or contacting the original image hosts. The archive includes ownership and revision metadata as well as publication content; it is an administrative archive, not automatically a filtered public data download.
A local persistent store supports repeated builds. A compatible transactional RDF4J service can also hold the documentation KG. This requires stronger isolation guarantees than an ordinary SPARQL query/update service; see SPARQL endpoints before choosing that backend.