SvelteKitBuild and deploy

Vercel

To deploy to Vercel, use adapter-vercel.

This adapter will be installed by default when you use adapter-auto, but adding it to your project allows you to specify Vercel-specific options.

Usage

Install with npm i -D @sveltejs/adapter-vercel, then add the adapter to your svelte.config.js:

svelte.config

import function adapter(config?: Config): Adapteradapter from '@sveltejs/adapter-vercel';

export default {
	kit: {
    adapter: Adapter;
}kit: {
		adapter: Adapteradapter: function adapter(config?: Config): Adapteradapter({
			// see below for options that can be set here
		})
	}
};

Deployment configuration

To control how your routes are deployed to Vercel as functions, you can specify deployment configuration, either through the option shown above or with export const config inside +server.js, +page(.server).js and +layout(.server).js files.

For example you could deploy some parts of your app as Edge Functions...

about/+page

/** @type {import('@sveltejs/adapter-vercel').Config} */
export const const config: {
    runtime: string;
}@type{import('@sveltejs/adapter-vercel').Config}config = {
	runtime: stringruntime: 'edge'
};

import type { type Config = (EdgeConfig | ServerlessConfig) & {
    images?: ImagesConfig;
}Config } from '@sveltejs/adapter-vercel';

export const const config: Configconfig: type Config = (EdgeConfig | ServerlessConfig) & {
    images?: ImagesConfig;
}Config = {
	runtime: "edge"runtime: 'edge'
};

...and others as Serverless Functions (note that by specifying config inside a layout, it applies to all child pages):

admin/+layout

/** @type {import('@sveltejs/adapter-vercel').Config} */
export const const config: {
    runtime: string;
}@type{import('@sveltejs/adapter-vercel').Config}config = {
	runtime: stringruntime: 'nodejs22.x'
};

import type { type Config = (EdgeConfig | ServerlessConfig) & {
    images?: ImagesConfig;
}Config } from '@sveltejs/adapter-vercel';

export const const config: Configconfig: type Config = (EdgeConfig | ServerlessConfig) & {
    images?: ImagesConfig;
}Config = {
	ServerlessConfig.runtime?: `nodejs${number}.x` | undefinedWhether to use Edge Functions ('edge') or Serverless Functions ('nodejs18.x', 'nodejs20.x' etc).
@defaultSame as the build environmentruntime: 'nodejs22.x'
};

The following options apply to all functions:

runtime: 'edge', 'nodejs18.x', 'nodejs20.x' or 'nodejs22.x'. By default, the adapter will select the 'nodejs<version>.x' corresponding to the Node version your project is configured to use on the Vercel dashboard
regions: an array of edge network regions (defaulting to ["iad1"] for serverless functions) or 'all' if runtime is edge (its default). Note that multiple regions for serverless functions are only supported on Enterprise plans
split: if true, causes a route to be deployed as an individual function. If split is set to true at the adapter level, all routes will be deployed as individual functions

Additionally, the following option applies to edge functions:

external: an array of dependencies that esbuild should treat as external when bundling functions. This should only be used to exclude optional dependencies that will not run outside Node

And the following option apply to serverless functions:

memory: the amount of memory available to the function. Defaults to 1024 Mb, and can be decreased to 128 Mb or increased in 64Mb increments up to 3008 Mb on Pro or Enterprise accounts
maxDuration: maximum execution duration of the function. Defaults to 10 seconds for Hobby accounts, 15 for Pro and 900 for Enterprise
isr: configuration Incremental Static Regeneration, described below

If your functions need to access data in a specific region, it’s recommended that they be deployed in the same region (or close to it) for optimal performance.

Image Optimization

You may set the images config to control how Vercel builds your images. See the image configuration reference for full details. As an example, you may set:

svelte.config

import function adapter(config?: Config): Adapteradapter from '@sveltejs/adapter-vercel';

export default {
	kit: {
    adapter({ images: { sizes: [], 640: , 828: , 1200: , 1920: , 3840:  } }: {
        images: {
 sizes: Iterable<any>;
 640: any;
 828: any;
 1200: any;
 1920: any;
 3840: any;
        };
    }): any;
    formats: string[];
    minimumCacheTTL: number;
    domains: string[];
}kit: {
		function adapter({ images: { sizes: [], 640: , 828: , 1200: , 1920: , 3840:  } }: {
    images: {
        sizes: Iterable<any>;
        640: any;
        828: any;
        1200: any;
        1920: any;
        3840: any;
    };
}): anyadapter({
			images: {
    sizes: Iterable<any>;
    640: any;
    828: any;
    1200: any;
    1920: any;
    3840: any;
}images: {
				sizes: Iterable<any>sizes: [640, 828, 1200, 1920, 3840],






				formats: string[]formats: ['image/avif', 'image/webp'],
				minimumCacheTTL: numberminimumCacheTTL: 300,
				domains: string[]domains: ['example-app.vercel.app'],
			}
		})
	}
};

Incremental Static Regeneration

Vercel supports Incremental Static Regeneration (ISR), which provides the performance and cost advantages of prerendered content with the flexibility of dynamically rendered content.

Use ISR only on routes where every visitor should see the same content (much like when you prerender). If there’s anything user-specific happening (like session cookies), they should happen on the client via JavaScript only to not leak sensitive information across visits

To add ISR to a route, include the isr property in your config object:

import { import BYPASS_TOKENBYPASS_TOKEN } from '$env/static/private';

export const const config: {
    isr: {
        expiration: number;
        bypassToken: any;
        allowQuery: string[];
    };
}config = {
	isr: {
    expiration: number;
    bypassToken: any;
    allowQuery: string[];
}isr: {
		expiration: numberexpiration: 60,
		bypassToken: anybypassToken: import BYPASS_TOKENBYPASS_TOKEN,
		allowQuery: string[]allowQuery: ['search']
	}
};

Using ISR on a route with export const prerender = true will have no effect, since the route is prerendered at build time

The expiration property is required; all others are optional. The properties are discussed in more detail below.

expiration

The expiration time (in seconds) before the cached asset will be re-generated by invoking the Serverless Function. Setting the value to false means it will never expire. In that case, you likely want to define a bypass token to re-generate on demand.

bypassToken

A random token that can be provided in the URL to bypass the cached version of the asset, by requesting the asset with a __prerender_bypass=<token> cookie.

Making a GET or HEAD request with x-prerender-revalidate: <token> will force the asset to be re-validated.

Note that the BYPASS_TOKEN string must be at least 32 characters long. You could generate one using the JavaScript console like so:

btoa(Math.random().toString()).substring(0,32);

Set this string as an environment variable on Vercel by logging in and going to your project then Settings > Environment Variables. For “Key” put BYPASS_TOKEN and for “value” use the string generated above, then hit “Save”.

To get this key known about for local development, you can use the Vercel CLI by running the vercel env pull command locally like so:

$ vercel env pull .env.development.local

allowQuery

A list of valid query parameters that contribute to the cache key. Other parameters (such as utm tracking codes) will be ignored, ensuring that they do not result in content being re-generated unnecessarily. By default, query parameters are ignored.

Pages that are prerendered will ignore ISR configuration.

Environment variables

Vercel makes a set of deployment-specific environment variables available. Like other environment variables, these are accessible from $env/static/private and $env/dynamic/private (sometimes — more on that later), and inaccessible from their public counterparts. To access one of these variables from the client:

+layout.server

import { import VERCEL_COMMIT_REFVERCEL_COMMIT_REF } from '$env/static/private';

/** @type {import('./$types').LayoutServerLoad} */
export function function load(): {
    deploymentGitBranch: any;
}@type{import('./$types').LayoutServerLoad}load() {
	return {
		deploymentGitBranch: anydeploymentGitBranch: import VERCEL_COMMIT_REFVERCEL_COMMIT_REF
	};
}

import { import VERCEL_COMMIT_REFVERCEL_COMMIT_REF } from '$env/static/private';
import type { type LayoutServerLoad = (event: Kit.ServerLoadEvent<Record<string, any>, Record<string, any>, string | null>) => MaybePromise<void | Record<string, any>>
type LayoutServerLoad = (event: Kit.ServerLoadEvent<Record<string, any>, Record<string, any>, string | null>) => MaybePromise<void | Record<string, any>>LayoutServerLoad } from './$types';

export const const load: LayoutServerLoadload: type LayoutServerLoad = (event: Kit.ServerLoadEvent<Record<string, any>, Record<string, any>, string | null>) => MaybePromise<void | Record<string, any>>
type LayoutServerLoad = (event: Kit.ServerLoadEvent<Record<string, any>, Record<string, any>, string | null>) => MaybePromise<void | Record<string, any>>LayoutServerLoad = () => {
	return {
		deploymentGitBranch: anydeploymentGitBranch: import VERCEL_COMMIT_REFVERCEL_COMMIT_REF
	};
};

+layout

<script>
	/** @type {import('./$types').LayoutProps} */
	let { data } = $props();
</script>

<p>This staging environment was deployed from {data.deploymentGitBranch}.</p>

<script lang="ts">
	import type { LayoutProps } from './$types';

	let { data }: LayoutProps = $props();
</script>

<p>This staging environment was deployed from {data.deploymentGitBranch}.</p>

Since all of these variables are unchanged between build time and run time when building on Vercel, we recommend using $env/static/private — which will statically replace the variables, enabling optimisations like dead code elimination — rather than $env/dynamic/private.

Skew protection

When a new version of your app is deployed, assets belonging to the previous version may no longer be accessible. If a user is actively using your app when this happens, it can cause errors when they navigate — this is known as version skew. SvelteKit mitigates this by detecting errors resulting from version skew and causing a hard reload to get the latest version of the app, but this will cause any client-side state to be lost. (You can also proactively mitigate it by observing the updated store value, which tells clients when a new version has been deployed.)

Skew protection is a Vercel feature that routes client requests to their original deployment. When a user visits your app, a cookie is set with the deployment ID, and any subsequent requests will be routed to that deployment for as long as skew protection is active. When they reload the page, they will get the newest deployment. (The updated store is exempted from this behaviour, and so will continue to report new deployments.) To enable it, visit the Advanced section of your project settings on Vercel.

Cookie-based skew protection comes with one caveat: if a user has multiple versions of your app open in multiple tabs, requests from older versions will be routed to the newer one, meaning they will fall back to SvelteKit’s built-in skew protection.

Notes

Vercel functions

If you have Vercel functions contained in the api directory at the project’s root, any requests for /api/* will not be handled by SvelteKit. You should implement these as API routes in your SvelteKit app instead, unless you need to use a non-JavaScript language in which case you will need to ensure that you don’t have any /api/* routes in your SvelteKit app.

Node version

Projects created before a certain date may default to using an older Node version than what SvelteKit currently requires. You can change the Node version in your project settings.

Troubleshooting

Accessing the file system

You can’t use fs in edge functions.

You can use it in serverless functions, but it won’t work as expected, since files are not copied from your project into your deployment. Instead, use the read function from $app/server to access your files. read does not work inside routes deployed as edge functions (this may change in future).

Alternatively, you can prerender the routes in question.

Edit this page on GitHub

previous next

Netlify Writing adapters