Excluding routes

Some pages shouldn't be swapped – an admin panel, a checkout, a logout route, anything that assumes a fresh page load. Opt them out and they fall back to ordinary full navigation.

Side-effect routes must not be GET

Sparke preloads by firing a real, credentialed GET at reachable links while the browser is idle - before any click. So a GET /logout (or any GET that deletes, unsubscribes, clears a cart, charges a card) gets hit speculatively and runs on page load, logging your users out before they touch anything. This is exactly why GET is defined as a safe method: Chrome's prefetcher, antivirus link scanners, and chat-app link unfurlers already break GET logout for the same reason.

The real fix is to make the action a POST. Sparke never preloads or intercepts non-GET forms, so it just works with no exclusion needed:

<form method="post" action="/logout">
  <button>Log out</button>
</form>

If the route must stay a GET link, exclude it with the mechanisms below (data-ignore="/logout" or rel="external"), or no-op it server-side when the request carries the X-Sparke: preload header - real clicks send X-Sparke: navigate instead, so you can tell a speculative preload apart from a genuine logout.

Centrally, on the script tag

Pass space-separated path patterns to data-ignore. A trailing * is a prefix wildcard.

<script
  src="sparke.min.js"
  defer
  data-ignore="/admin/* /checkout /logout"
></script>

Per link

For one-off exclusions, use standard HTML – no Sparke API needed:

<a href="/legacy/billing" rel="external">Billing portal</a>

A whole page, Sparke-free

To make an entire page behave like a plain MPA page, simply omit the <script> tag on it.