This section explains how the Atlas graph is maintained, reviewed, corrected, and carried through downstream docs and tooling. Read it when you want the operating model behind the catalog rather than a single domain fact.

Tests, evidence-boundary checks, versioning, CI/CD, and downstream artifact distribution. The phase that enforces every guarantee the earlier phases promise.

Generators turn graph queries into derived artifacts. Every page of prose, every TypeScript type, every OpenAPI document, every scaffolded code skeleton in the v6 ecosystem should be produced from graph data rather than hand-maintained copies.

This section collects pages that are produced directly from graph data to explain a concrete domain, map, or reference set. These pages are where Atlas feels most like a handbook instead of a schema browser.

The site is a static, file-based build. Output lands under wiki/generated/. Top-level navigation:

MATCH X reads as "all nodes of NodeKind X". → edgename → Y reads "follow edgename from each match to nodes of kind Y". Multi-hop chains are read left-to-right.

Atlas wiki mirror of the repository docs site. The existing docs site remains unchanged; these pages are generated from docs markdown during the Atlas build.

Reusable views are wiki-backed projections that render graph records with a named UI template instead of plain markdown alone.