Backlinks

Taragen records which pages link to which, so a page can show what links to it:

// in articles/_layout.jsx
{page.backlinks.length > 0 && <aside>
  <h2>Linked from</h2>
  <ul>{page.backlinks.map(p => <li><a href={p.path}>{p.title}</a></li>)}</ul>
</aside>}

Filter by path for a section-specific count:

<p>{page.backlinks.filter(p => p.path.startsWith("/logs/")).length} logs link here</p>

links and backlinks

Every page gets two lists, empty when there's nothing:

• page.links: the sorted URLs of site pages this page's body links to, such as ["/docs/outputs", "/logs/2026/ngrok"].
• page.backlinks: the pages that link to this page, in the usual listing order . Each entry is a copy of the linking page's data, like a pages() result, but without links, backlinks, or subpages (so there are no cycles). To get more about a referrer, call pages() on its path.

pages() results carry both lists too.

What counts as a link

A link is the href of an <a> in a page's body: its own content, before layouts. Nav bars, footers, and other links added by layouts don't count, or every page would link to every nav target. If you want a link counted, put it in the page body.

Only HTML pages are link sources; feeds and sitemaps link to everything and would drown out real references.

A link counts toward a page when it resolves to that page's own URL. It doesn't count when it:

• points at an output, such as /articles/x.txt, or a static file, or nothing;
• points at an alias or redirect URL (link to the page's current path instead);
• points at the linking page itself, including #section links;
• involves a draft, in build. In serve, drafts count.

Several links from one page to the same target count once.

How hrefs are resolved

• mailto:, tel:, javascript:, and other non-HTTP schemes are skipped.

• Absolute links (https://..., //...) count only if their origin matches siteURL. Set siteURL in your root _defaults.jsx so full links to your own pages count:

  siteURL = "https://example.com";
  

  Without it, every absolute link is treated as external.

• Relative links are resolved the way a browser resolves them on the built site, where a clean-URL page lives at <path>/. So in /docs/intro, href="setup" means /docs/intro/setup, and you'd write href="../setup" or href="/docs/setup" to reach a sibling.

• Query strings and #fragments are dropped, . and .. are resolved, a trailing slash is dropped, and the path is percent-decoded.

When backlinks are available

The link graph is computed once per full parse of the site: in build, and in serve after each change. Index pages that call pages() are parsed a second time, once all pages exist, so their links to everything they list count.

Layouts always see the full graph, as does any page body in the final output. Links rendered from backlinks themselves (a "linked from" list) are not collected, so a backlinks section never feeds back into the graph.

One side effect: a page body that calls pages() runs twice per parse, so something like console.log in it prints twice.