The graph page

Every pagers build writes graph.html at the site root — an interactive, dependency-free force-directed map of how pages link to each other. It's linked from the site header on every page.

What it shows

• One node per content page, sized/labelled by title.
• Directed edges for internal Markdown links ([x](other.md) and resolved clean-URL links).
• Drag nodes to rearrange; click a node for its title, URL and outgoing links; "Open page" jumps to the page.
• Dark/light theming follows the site toggle.

The renderer (crates/pagers-graph) extracts edges during the build from the same link-resolution pass that rewrites URLs, so the graph always matches what readers can actually click.

Standalone graph

pagers graph scans a directory and writes the same data without building the site — useful for auditing link structure or embedding elsewhere:

pagers graph ./docs
# → graph.html, graph.json, graph.dot

• graph.html — self-contained viewer (open directly in a browser).
• graph.json — { nodes: [...], edges: [...], stats: {...} } for tooling.
• graph.dot — Graphviz DOT if you prefer dot -Tsvg.

Reading the JSON

{
  "nodes": [{ "id": "/guide/install/", "title": "Installation", "source": "guide/install.md" }],
  "edges": [{ "from": "/", "to": "/guide/install/" }],
  "stats": { "pages": 12, "edges": 30, "orphaned": 1 }
}

Orphaned pages (no inbound links) and dangling references are the usual actionable findings.