Collect existing sources
Keep documentation close to the project it describes. GraphArch uses focused collectors for particular inputs, so a publication can combine authored prose with RDF resources and other supported constructs. Collecting content does not choose its chapter, page, or publication status; those are publication decisions.
The Collectors catalogue maps each current collector to its inputs, supported documentable things, practical examples, and entry points.
Available source paths
| Source | What is collected |
|---|---|
| Markdown | Authored explanations, headings, code examples, frontmatter, and supported images or diagrams. |
| RDF files | Native statements and annotations that describe resources such as ontologies, constraints, and taxonomies. |
| Git repositories | Existing local checkouts or a selected remote revision, with file and commit provenance. |
| SPARQL endpoints | RDF returned by explicitly configured CONSTRUCT or DESCRIBE queries. |
There is also a specialised collector for supported structures in an already-published W3C/ReSpec HTML specification. It reads the published document without running its original JavaScript generator. See ReSpec compatibility for its scope and the specification guide for an example.
Combine inputs deliberately
Use an ontology-family build for a directory, repository, or set of endpoint queries. Use a website build for a Markdown documentation tree. The staged CLI can also collect individual inputs into a shared documentation graph before planning a publication. The workflow guide links these paths.
Independent collectors can run concurrently. Each input keeps its own identity and observed revision. An updated source replaces its own successful contribution; a missing file, invalid RDF, or failed request is an explicit build failure rather than permission to publish an incomplete replacement.
Content support is specific
A file extension alone is not a promise that GraphArch understands the file's meaning. Generic CSV, JSON, XML, YAML, TOML, SQL, Cypher, and GraphQL conversion adapters are not currently supplied. Configuration files and retained code examples in those formats are different from collectors that document arbitrary records or database schemas. See extending GraphArch.