Smarter Payload Handling for Cached Routes

What's new#

Nuxt 4.4 fixes an expensive quirk of payload extraction on cached routes and adds a new payloadExtraction: 'client' mode plus a runtime in-memory LRU cache for _payload.json.

Why you'd care#

With the ISR/SWR payload extraction from the 4.3 series, the browser would immediately fetch _payload.json as a second request — which triggered a full SSR re-render of the same page. In serverless, that could spin up a second lambda before the first response finished streaming. Wasteful. 4.4 makes it stop.

Before#

Payload extraction on a cached route meant: render HTML → browser fetches _payload.json → server re-renders the whole page just to produce that JSON.

export default defineNuxtConfig({
  routeRules: {
    '/products/**': { isr: 3600 }
  }
})

Two renders for one page view.

After#

Inline the payload in the initial HTML, and let a runtime cache serve later _payload.json requests without re-rendering:

export default defineNuxtConfig({
  experimental: {
    payloadExtraction: 'client',
  },
})

• 'client' mode inlines the full payload in the first HTML response, while still generating _payload.json for client-side navigation.
• The in-memory LRU cache serves _payload.json without a full re-render — and it's active for all users regardless of this setting.

Do it yourself#

1. Keep your isr/swr route rules from the 4.3 series.
2. Add experimental: { payloadExtraction: 'client' } to nuxt.config.ts.
3. Build and preview (nuxt build → nuxt preview).
4. Load a cached route and watch the network tab — the initial navigation should not fire a separate _payload.json re-render.
5. Client-navigate to another cached route and confirm _payload.json is still used (served from cache).

Gotchas#

• payloadExtraction: 'client' becomes the default under compatibilityVersion: 5 — adopting it now is future-proofing.
• The LRU cache is in-memory per server instance. Across many serverless instances each keeps its own cache; that's fine, it still removes the re-render cost.

Recap#

payloadExtraction: 'client' inlines the payload and, with the runtime LRU cache, kills the redundant second render on cached routes. Turn it on if you use isr/swr.