How to make fetch have special behaviour in the webview?

[myarcana]

I would like to additionally patch fetch so that fetch('http://a-server.com/this/path.js') first looks in the capacitor file system for capacitor://localhost/this/path.js and serves that if it exists, otherwise it hits the internet.

How could I set that up?

[MILLERMARRU]

Patching window.fetch directly will only catch requests made from that one JS context, and it's fragile against anything Capacitor or a library does at a lower level, so I'd go with a Service Worker instead, that's the platform-level place to intercept requests before they hit the network, and it works the same way on web/iOS/Android.

Rough shape:

// sw.js
self.addEventListener('fetch', (event) => {
  const url = new URL(event.request.url)
  if (url.origin === 'http://a-server.com') {
    event.respondWith(
      caches.match(event.request).then((cached) => cached || fetch(event.request))
    )
  }
})

and populate the cache ahead of time from files you already have via the Filesystem plugin (Filesystem.readFile + wrapping the result in a Response, then cache.put(request, response)), rather than trying to synchronously answer from the filesystem inside the fetch handler.

Two Capacitor-specific things to know before you go this route:

• On Android, since Capacitor 7 there's a resolveServiceWorkerRequests config option (server.resolveServiceWorkerRequests, default true) that routes service worker fetch interception through the Capacitor bridge. If you're on an older version or have it disabled, service worker requests may not behave the way you'd expect on Android specifically.
• Also on Android, there's a known issue where registering a service worker can prevent Capacitor's own plugin bridge from injecting into that execution context (Plugin not implemented errors), since the service worker runs in a separate context from your main page. There's a documented workaround linked from Capacitor's Android troubleshooting docs if you hit that, but it's worth testing early rather than after you've built the whole interception logic.

If your actual goal is simpler than a generic prefix rewrite, i.e. you always know upfront which specific bundled files should be served locally, precaching them in the service worker's install event (standard Cache API, no Capacitor-specific API needed) is less fragile than trying to intercept and redirect arbitrary URLs at request time.