``` --- title: Service workers --- Service workers act as proxy servers that handle network requests inside your app. This makes it possible to make your app work offline, but even if you don't need offline support (or can't realistically implement it because of the type of app you're building), it's often worth using service workers to speed up navigation by precaching your built JS and CSS. In SvelteKit, if you have a `src/service-worker.js` file (or `src/service-worker/index.js`) it will be bundled and automatically registered. You can change the [location of your service worker](configuration#files) if you need to. You can [disable automatic registration](configuration#serviceWorker) if you need to register the service worker with your own logic or use another solution. The default registration looks something like this: ```js if ('serviceWorker' in navigator) { addEventListener('load', function () { navigator.serviceWorker.register('./path/to/service-worker.js'); }); } ``` ## Inside the service worker Inside the service worker you have access to the [`$service-worker` module]($service-worker), which provides you with the paths to all static assets, build files and prerendered pages. You're also provided with an app version string, which you can use for creating a unique cache name, and the deployment's `base` path. If your Vite config specifies `define` (used for global variable replacements), this will be applied to service workers as well as your server/client builds. The following example caches the built app and any files in `static` eagerly, and caches all other requests as they happen. This would make each page work offline once visited. ```js // @errors: 2339 ///
import { build, files, version } from '$service-worker'; // Create a unique cache name for this deployment const CACHE = `cache-${version}`; const ASSETS = [ ...build, // the app itself ...files // everything in `static` ]; self.addEventListener('install', (event) => { // Create a new cache and add all files to it async function addFilesToCache() { const cache = await caches.open(CACHE); await cache.addAll(ASSETS); } event.waitUntil(addFilesToCache()); }); self.addEventListener('activate', (event) => { // Remove previous cached data from disk async function deleteOldCaches() { for (const key of await caches.keys()) { if (key !== CACHE) await caches.delete(key); } } event.waitUntil(deleteOldCaches()); }); self.addEventListener('fetch', (event) => { // ignore POST requests etc if (event.request.method !== 'GET') return; async function respond() { const url = new URL(event.request.url); const cache = await caches.open(CACHE); // `build`/`files` can always be served from the cache if (ASSETS.includes(url.pathname)) { const response = await cache.match(url.pathname); if (response) { return response; } } // for everything else, try the network first, but // fall back to the cache if we're offline try { const response = await fetch(event.request); // if we're offline, fetch can return a value that is not a Response // instead of throwing - and we can't pass this non-Response to respondWith if (!(response instanceof Response)) { throw new Error('invalid response from fetch'); } if (response.status === 200) { cache.put(event.request, response.clone()); } return response; } catch (err) { const response = await cache.match(event.request); if (response) { return response; } // if there's no cache, then just error out // as there is nothing we can do to respond to this request throw err; } } event.respondWith(respond()); }); ``` > [!NOTE] Be careful when caching! In some cases, stale data might be worse than data that's unavailable while offline. Since browsers will empty caches if they get too full, you should also be careful about caching large assets like video files. ## During development The service worker is bundled for production, but not during development. For that reason, only browsers that support [modules in service workers](https://web.dev/es-modules-in-sw) will be able to use them at dev time. If you are manually registering your service worker, you will need to pass the `{ type: 'module' }` option in development: ```js import { dev } from '$app/environment'; navigator.serviceWorker.register('/service-worker.js', { type: dev ? 'module' : 'classic' }); ``` > [!NOTE] `build` and `prerendered` are empty arrays during development ## Type safety Setting up proper types for service workers requires some manual setup. Inside your `service-worker.js`, add the following to the top of your file: ```original-js ///
///
///
///
const sw = /** @type {ServiceWorkerGlobalScope} */ (/** @type {unknown} */ (self)); ``` ```generated-ts ///
///
///
///
const sw = self as unknown as ServiceWorkerGlobalScope; ``` This disables access to DOM typings like `HTMLElement` which are not available inside a service worker and instantiates the correct globals. The reassignment of `self` to `sw` allows you to type cast it in the process (there are a couple of ways to do this, but this is the easiest that requires no additional files). Use `sw` instead of `self` in the rest of the file. The reference to the SvelteKit types ensures that the `$service-worker` import has proper type definitions. If you import `$env/static/public` you either have to `// @ts-ignore` the import or add `///
` to the reference types. ## Other solutions SvelteKit's service worker implementation is designed to be easy to work with and is probably a good solution for most users. However, outside of SvelteKit, many PWA applications leverage the [Workbox](https://web.dev/learn/pwa/workbox) library. If you're used to using Workbox you may prefer [Vite PWA plugin](https://vite-pwa-org.netlify.app/frameworks/sveltekit.html). ## References For more general information on service workers, we recommend [the MDN web docs](https://developer.mozilla.org/en-US/docs/Web/API/Service_Worker_API/Using_Service_Workers). --- title: Server-only modules --- Like a good friend, SvelteKit keeps your secrets. When writing your backend and frontend in the same repository, it can be easy to accidentally import sensitive data into your front-end code (environment variables containing API keys, for example). SvelteKit provides a way to prevent this entirely: server-only modules. ## Private environment variables The [`$env/static/private`]($env-static-private) and [`$env/dynamic/private`]($env-dynamic-private) modules can only be imported into modules that only run on the server, such as [`hooks.server.js`](hooks#Server-hooks) or [`+page.server.js`](routing#page-page.server.js). ## Server-only utilities The [`$app/server`]($app-server) module, which contains a [`read`]($app-server#read) function for reading assets from the filesystem, can likewise only be imported by code that runs on the server. ## Your modules You can make your own modules server-only in two ways: - adding `.server` to the filename, e.g. `secrets.server.js` - placing them in `$lib/server`, e.g. `$lib/server/secrets.js` ## How it works Any time you have public-facing code that imports server-only code (whether directly or indirectly)... ```js // @errors: 7005 /// file: $lib/server/secrets.js export const atlantisCoordinates = [/* redacted */]; ``` ```js // @errors: 2307 7006 7005 /// file: src/routes/utils.js export { atlantisCoordinates } from '$lib/server/secrets.js'; export const add = (a, b) => a + b; ``` ```html /// file: src/routes/+page.svelte ``` ...SvelteKit will error: ``` Cannot import $lib/server/secrets.js into public-facing code: - src/routes/+page.svelte - src/routes/utils.js - $lib/server/secrets.js ``` Even though the public-facing code — `src/routes/+page.svelte` — only uses the `add` export and not the secret `atlantisCoordinates` export, the secret code could end up in JavaScript that the browser downloads, and so the import chain is considered unsafe. This feature also works with dynamic imports, even interpolated ones like ``await import(`./${foo}.js`)``, with one small caveat: during development, if there are two or more dynamic imports between the public-facing code and the server-only module, the illegal import will not be detected the first time the code is loaded. > [!NOTE] Unit testing frameworks like Vitest do not distinguish between server-only and public-facing code. For this reason, illegal import detection is disabled when running tests, as determined by `process.env.TEST === 'true'`. ## Further reading - [Tutorial: Environment variables](/tutorial/kit/env-static-private) --- title: Snapshots --- Ephemeral DOM state — like scroll positions on sidebars, the content of `
` elements and so on — is discarded when you navigate from one page to another. For example, if the user fills out a form but navigates away and then back before submitting, or if the user refreshes the page, the values they filled in will be lost. In cases where it's valuable to preserve that input, you can take a _snapshot_ of DOM state, which can then be restored if the user navigates back. To do this, export a `snapshot` object with `capture` and `restore` methods from a `+page.svelte` or `+layout.svelte`: ```svelte
``` When you navigate away from this page, the `capture` function is called immediately before the page updates, and the returned value is associated with the current entry in the browser's history stack. If you navigate back, the `restore` function is called with the stored value as soon as the page is updated. The data must be serializable as JSON so that it can be persisted to `sessionStorage`. This allows the state to be restored when the page is reloaded, or when the user navigates back from a different site. > [!NOTE] Avoid returning very large objects from `capture` — once captured, objects will be retained in memory for the duration of the session, and in extreme cases may be too large to persist to `sessionStorage`. --- title: Shallow routing --- As you navigate around a SvelteKit app, you create _history entries_. Clicking the back and forward buttons traverses through this list of entries, re-running any `load` functions and replacing page components as necessary. Sometimes, it's useful to create history entries _without_ navigating. For example, you might want to show a modal dialog that the user can dismiss by navigating back. This is particularly valuable on mobile devices, where swipe gestures are often more natural than interacting directly with the UI. In these cases, a modal that is _not_ associated with a history entry can be a source of frustration, as a user may swipe backwards in an attempt to dismiss it and find themselves on the wrong page. SvelteKit makes this possible with the [`pushState`]($app-navigation#pushState) and [`replaceState`]($app-navigation#replaceState) functions, which allow you to associate state with a history entry without navigating. For example, to implement a history-driven modal: ```svelte {#if page.state.showModal}
history.back()} /> {/if} ``` The modal can be dismissed by navigating back (unsetting `page.state.showModal`) or by interacting with it in a way that causes the `close` callback to run, which will navigate back programmatically. ## API The first argument to `pushState` is the URL, relative to the current URL. To stay on the current URL, use `''`. The second argument is the new page state, which can be accessed via the [page object]($app-state#page) as `page.state`. You can make page state type-safe by declaring an [`App.PageState`](types#PageState) interface (usually in `src/app.d.ts`). To set page state without creating a new history entry, use `replaceState` instead of `pushState`. > [!LEGACY] > `page.state` from `$app/state` was added in SvelteKit 2.12. If you're using an earlier version or are using Svelte 4, use `$page.state` from `$app/stores` instead. ## Loading data for a route When shallow routing, you may want to render another `+page.svelte` inside the current page. For example, clicking on a photo thumbnail could pop up the detail view without navigating to the photo page. For this to work, you need to load the data that the `+page.svelte` expects. A convenient way to do this is to use [`preloadData`]($app-navigation#preloadData) inside the `click` handler of an `` element. If the element (or a parent) uses [`data-sveltekit-preload-data`](link-options#data-sveltekit-preload-data), the data will have already been requested, and `preloadData` will reuse that request. ```svelte {#each data.thumbnails as thumbnail} { if (innerWidth < 640 // bail if the screen is too small || e.shiftKey // or the link is opened in a new window || e.metaKey || e.ctrlKey // or a new tab (mac: metaKey, win/linux: ctrlKey) // should also consider clicking with a mouse scroll wheel ) return; // prevent navigation e.preventDefault(); const { href } = e.currentTarget; // run `load` functions (or rather, get the result of the `load` functions // that are already running because of `data-sveltekit-preload-data`) const result = await preloadData(href); if (result.type === 'loaded' && result.status === 200) { pushState(href, { selected: result.data }); } else { // something bad happened! try navigating goto(href); } }} > {/each} {#if page.state.selected} history.back()}> {/if} ``` ## Caveats During server-side rendering, `page.state` is always an empty object. The same is true for the first page the user lands on — if the user reloads the page (or returns from another document), state will _not_ be applied until they navigate. Shallow routing is a feature that requires JavaScript to work. Be mindful when using it and try to think of sensible fallback behavior in case JavaScript isn't available. --- title: Packaging --- You can use SvelteKit to build apps as well as component libraries, using the `@sveltejs/package` package (`npx sv create` has an option to set this up for you). When you're creating an app, the contents of `src/routes` is the public-facing stuff; [`src/lib`]($lib) contains your app's internal library. A component library has the exact same structure as a SvelteKit app, except that `src/lib` is the public-facing bit, and your root `package.json` is used to publish the package. `src/routes` might be a documentation or demo site that accompanies the library, or it might just be a sandbox you use during development. Running the `svelte-package` command from `@sveltejs/package` will take the contents of `src/lib` and generate a `dist` directory (which can be [configured](#Options)) containing the following: - All the files in `src/lib`. Svelte components will be preprocessed, TypeScript files will be transpiled to JavaScript. - Type definitions (`d.ts` files) which are generated for Svelte, JavaScript and TypeScript files. You need to install `typescript >= 4.0.0` for this. Type definitions are placed next to their implementation, hand-written `d.ts` files are copied over as is. You can [disable generation](#Options), but we strongly recommend against it — people using your library might use TypeScript, for which they require these type definition files. > [!NOTE] `@sveltejs/package` version 1 generated a `package.json`. This is no longer the case and it will now use the `package.json` from your project and validate that it is correct instead. If you're still on version 1, see [this PR](https://github.com/sveltejs/kit/pull/8922) for migration instructions. ## Anatomy of a package.json Since you're now building a library for public use, the contents of your `package.json` will become more important. Through it, you configure the entry points of your package, which files are published to npm, and which dependencies your library has. Let's go through the most important fields one by one. ### name This is the name of your package. It will be available for others to install using that name, and visible on `https://npmjs.com/package/`. ```json { "name": "your-library" } ``` Read more about it [here](https://docs.npmjs.com/cli/v9/configuring-npm/package-json#name). ### license Every package should have a license field so people know how they are allowed to use it. A very popular license which is also very permissive in terms of distribution and reuse without warranty is `MIT`. ```json { "license": "MIT" } ``` Read more about it [here](https://docs.npmjs.com/cli/v9/configuring-npm/package-json#license). Note that you should also include a `LICENSE` file in your package. ### files This tells npm which files it will pack up and upload to npm. It should contain your output folder (`dist` by default). Your `package.json` and `README` and `LICENSE` will always be included, so you don't need to specify them. ```json { "files": ["dist"] } ``` To exclude unnecessary files (such as unit tests, or modules that are only imported from `src/routes` etc) you can add them to an `.npmignore` file. This will result in smaller packages that are faster to install. Read more about it [here](https://docs.npmjs.com/cli/v9/configuring-npm/package-json#files). ### exports The `"exports"` field contains the package's entry points. If you set up a new library project through `npx sv create`, it's set to a single export, the package root: ```json { "exports": { ".": { "types": "./dist/index.d.ts", "svelte": "./dist/index.js" } } } ``` This tells bundlers and tooling that your package only has one entry point, the root, and everything should be imported through that, like this: ```js // @errors: 2307 import { Something } from 'your-library'; ``` The `types` and `svelte` keys are [export conditions](https://nodejs.org/api/packages.html#conditional-exports). They tell tooling what file to import when they look up the `your-library` import: - TypeScript sees the `types` condition and looks up the type definition file. If you don't publish type definitions, omit this condition. - Svelte-aware tooling sees the `svelte` condition and knows this is a Svelte component library. If you publish a library that does not export any Svelte components and that could also work in non-Svelte projects (for example a Svelte store library), you can replace this condition with `default`. > [!NOTE] Previous versions of `@sveltejs/package` also added a `package.json` export. This is no longer part of the template because all tooling can now deal with a `package.json` not being explicitly exported. You can adjust `exports` to your liking and provide more entry points. For example, if instead of a `src/lib/index.js` file that re-exported components you wanted to expose a `src/lib/Foo.svelte` component directly, you could create the following export map... ```json { "exports": { "./Foo.svelte": { "types": "./dist/Foo.svelte.d.ts", "svelte": "./dist/Foo.svelte" } } } ``` ...and a consumer of your library could import the component like so: ```js // @filename: ambient.d.ts declare module 'your-library/Foo.svelte'; // @filename: index.js // ---cut--- import Foo from 'your-library/Foo.svelte'; ``` > [!NOTE] Beware that doing this will need additional care if you provide type definitions. Read more about the caveat [here](#TypeScript) In general, each key of the exports map is the path the user will have to use to import something from your package, and the value is the path to the file that will be imported or a map of export conditions which in turn contains these file paths. Read more about `exports` [here](https://nodejs.org/docs/latest-v18.x/api/packages.html#package-entry-points). ### svelte This is a legacy field that enabled tooling to recognise Svelte component libraries. It's no longer necessary when using the `svelte` [export condition](#Anatomy-of-a-package.json-exports), but for backwards compatibility with outdated tooling that doesn't yet know about export conditions it's good to keep it around. It should point towards your root entry point. ```json { "svelte": "./dist/index.js" } ``` ### sideEffects The `sideEffects` field in `package.json` is used by bundlers to determine if a module may contain code that has side effects. A module is considered to have side effects if it makes changes that are observable from other scripts outside the module when it's imported. For example, side effects include modifying global variables or the prototype of built-in JavaScript objects. Because a side effect could potentially affect the behavior of other parts of the application, these files/modules will be included in the final bundle regardless of whether their exports are used in the application. It is a best practice to avoid side effects in your code. Setting the `sideEffects` field in `package.json` can help the bundler to be more aggressive in eliminating unused exports from the final bundle, a process known as tree-shaking. This results in smaller and more efficient bundles. Different bundlers handle `sideEffects` in various manners. While not necessary for Vite, we recommend that libraries state that all CSS files have side effects so that your library will be [compatible with webpack](https://webpack.js.org/guides/tree-shaking/#mark-the-file-as-side-effect-free). This is the configuration that comes with newly created projects: ```json /// file: package.json { "sideEffects": ["**/*.css"] } ``` > If the scripts in your library have side effects, ensure that you update the `sideEffects` field. All scripts are marked as side effect free by default in newly created projects. If a file with side effects is incorrectly marked as having no side effects, it can result in broken functionality. If your package has files with side effects, you can specify them in an array: ```json /// file: package.json { "sideEffects": [ "**/*.css", "./dist/sideEffectfulFile.js" ] } ``` This will treat only the specified files as having side effects. ## TypeScript You should ship type definitions for your library even if you don't use TypeScript yourself so that people who do get proper intellisense when using your library. `@sveltejs/package` makes the process of generating types mostly opaque to you. By default, when packaging your library, type definitions are auto-generated for JavaScript, TypeScript and Svelte files. All you need to ensure is that the `types` condition in the [exports](#Anatomy-of-a-package.json-exports) map points to the correct files. When initialising a library project through `npx sv create`, this is automatically setup for the root export. If you have something else than a root export however — for example providing a `your-library/foo` import — you need to take additional care for providing type definitions. Unfortunately, TypeScript by default will _not_ resolve the `types` condition for an export like `{ "./foo": { "types": "./dist/foo.d.ts", ... }}`. Instead, it will search for a `foo.d.ts` relative to the root of your library (i.e. `your-library/foo.d.ts` instead of `your-library/dist/foo.d.ts`). To fix this, you have two options: The first option is to require people using your library to set the `moduleResolution` option in their `tsconfig.json` (or `jsconfig.json`) to `bundler` (available since TypeScript 5, the best and recommended option in the future), `node16` or `nodenext`. This opts TypeScript into actually looking at the exports map and resolving the types correctly. The second option is to (ab)use the `typesVersions` feature from TypeScript to wire up the types. This is a field inside `package.json` TypeScript uses to check for different type definitions depending on the TypeScript version, and also contains a path mapping feature for that. We leverage that path mapping feature to get what we want. For the mentioned `foo` export above, the corresponding `typesVersions` looks like this: ```json { "exports": { "./foo": { "types": "./dist/foo.d.ts", "svelte": "./dist/foo.js" } }, "typesVersions": { ">4.0": { "foo": ["./dist/foo.d.ts"] } } } ``` `>4.0` tells TypeScript to check the inner map if the used TypeScript version is greater than 4 (which should in practice always be true). The inner map tells TypeScript that the typings for `your-library/foo` are found within `./dist/foo.d.ts`, which essentially replicates the `exports` condition. You also have `*` as a wildcard at your disposal to make many type definitions at once available without repeating yourself. Note that if you opt into `typesVersions` you have to declare all type imports through it, including the root import (which is defined as `"index.d.ts": [..]`). You can read more about that feature [here](https://www.typescriptlang.org/docs/handbook/declaration-files/publishing.html#version-selection-with-typesversions). ## Best practices You should avoid using SvelteKit-specific modules like `$app/environment` in your packages unless you intend for them to only be consumable by other SvelteKit projects. E.g. rather than using `import { browser } from '$app/environment'` you could use `import { BROWSER } from 'esm-env'` ([see esm-env docs](https://github.com/benmccann/esm-env)). You may also wish to pass in things like the current URL or a navigation action as a prop rather than relying directly on `$app/state`, `$app/navigation`, etc. Writing your app in this more generic fashion will also make it easier to setup tools for testing, UI demos and so on. Ensure that you add [aliases](configuration#alias) via `svelte.config.js` (not `vite.config.js` or `tsconfig.json`), so that they are processed by `svelte-package`. You should think carefully about whether or not the changes you make to your package are a bug fix, a new feature, or a breaking change, and update the package version accordingly. Note that if you remove any paths from `exports` or any `export` conditions inside them from your existing library, that should be regarded as a breaking change. ```json { "exports": { ".": { "types": "./dist/index.d.ts", // changing `svelte` to `default` is a breaking change: --- "svelte": "./dist/index.js"--- +++ "default": "./dist/index.js"+++ }, // removing this is a breaking change: --- "./foo": { "types": "./dist/foo.d.ts", "svelte": "./dist/foo.js", "default": "./dist/foo.js" },--- // adding this is ok: +++ "./bar": { "types": "./dist/bar.d.ts", "svelte": "./dist/bar.js", "default": "./dist/bar.js" }+++ } } ``` ## Source maps You can create so-called declaration maps (`d.ts.map` files) by setting `"declarationMap": true` in your `tsconfig.json`. This will allow editors such as VS Code to go to the original `.ts` or `.svelte` file when using features like _Go to Definition_. This means you also need to publish your source files alongside your dist folder in a way that the relative path inside the declaration files leads to a file on disk. Assuming that you have all your library code inside `src/lib` as suggested by Svelte's CLI, this is as simple as adding `src/lib` to `files` in your `package.json`: ```json { "files": [ "dist", "!dist/**/*.test.*", "!dist/**/*.spec.*", +++"src/lib", "!src/lib/**/*.test.*", "!src/lib/**/*.spec.*"+++ ] } ``` ## Options `svelte-package` accepts the following options: - `-w`/`--watch` — watch files in `src/lib` for changes and rebuild the package - `-i`/`--input` — the input directory which contains all the files of the package. Defaults to `src/lib` - `-o`/`--output` — the output directory where the processed files are written to. Your `package.json`'s `exports` should point to files inside there, and the `files` array should include that folder. Defaults to `dist` - `-t`/`--types` — whether or not to create type definitions (`d.ts` files). We strongly recommend doing this as it fosters ecosystem library quality. Defaults to `true` - `--tsconfig` - the path to a tsconfig or jsconfig. When not provided, searches for the next upper tsconfig/jsconfig in the workspace path. ## Publishing To publish the generated package: ```sh npm publish ``` ## Caveats All relative file imports need to be fully specified, adhering to Node's ESM algorithm. This means that for a file like `src/lib/something/index.js`, you must include the filename with the extension: ```js // @errors: 2307 import { something } from './something+++/index.js+++'; ``` If you are using TypeScript, you need to import `.ts` files the same way, but using a `.js` file ending, _not_ a `.ts` file ending. (This is a TypeScript design decision outside our control.) Setting `"moduleResolution": "NodeNext"` in your `tsconfig.json` or `jsconfig.json` will help you with this. All files except Svelte files (preprocessed) and TypeScript files (transpiled to JavaScript) are copied across as-is. --- title: Advanced --- --- title: Auth --- Auth refers to authentication and authorization, which are common needs when building a web application. Authentication means verifying that the user is who they say they are based on their provided credentials. Authorization means determining which actions they are allowed to take. ## Sessions vs tokens After the user has provided their credentials such as a username and password, we want to allow them to use the application without needing to provide their credentials again for future requests. Users are commonly authenticated on subsequent requests with either a session identifier or signed token such as a JSON Web Token (JWT). Session IDs are most commonly stored in a database. They can be immediately revoked, but require a database query to be made on each request. In contrast, JWT generally are not checked against a datastore, which means they cannot be immediately revoked. The advantage of this method is improved latency and reduced load on your datastore. ## Integration points Auth [cookies](@sveltejs-kit#Cookies) can be checked inside [server hooks](hooks#Server-hooks). If a user is found matching the provided credentials, the user information can be stored in [`locals`](hooks#Server-hooks-handle). ## Guides [Lucia](https://lucia-auth.com/) is a good reference for session-based web app auth. It contains example code snippets and projects for implementing session-based auth within SvelteKit and other JS projects. You can add code which follows the Lucia guide to your project with `npx sv create` when creating a new project or `npx sv add lucia` for an existing project. An auth system is tightly coupled to a web framework because most of the code lies in validating user input, handling errors, and directing users to the appropriate next page. As a result, many of the generic JS auth libraries include one or more web frameworks within them. For this reason, many users will find it preferrable to follow a SvelteKit-specific guide such as the examples found in [Lucia](https://lucia-auth.com/) rather than having multiple web frameworks inside their project. --- title: Performance --- Out of the box, SvelteKit does a lot of work to make your applications as performant as possible: - Code-splitting, so that only the code you need for the current page is loaded - Asset preloading, so that 'waterfalls' (of files requesting other files) are prevented - File hashing, so that your assets can be cached forever - Request coalescing, so that data fetched from separate server `load` functions is grouped into a single HTTP request - Parallel loading, so that separate universal `load` functions fetch data simultaneously - Data inlining, so that requests made with `fetch` during server rendering can be replayed in the browser without issuing a new request - Conservative invalidation, so that `load` functions are only re-run when necessary - Prerendering (configurable on a per-route basis, if necessary) so that pages without dynamic data can be served instantaneously - Link preloading, so that data and code requirements for a client-side navigation are eagerly anticipated Nevertheless, we can't (yet) eliminate all sources of slowness. To eke out maximum performance, you should be mindful of the following tips. ## Diagnosing issues Google's [PageSpeed Insights](https://pagespeed.web.dev/) and (for more advanced analysis) [WebPageTest](https://www.webpagetest.org/) are excellent ways to understand the performance characteristics of a site that is already deployed to the internet. Your browser also includes useful developer tools for analysing your site, whether deployed or running locally: * Chrome - [Lighthouse](https://developer.chrome.com/docs/lighthouse/overview#devtools), [Network](https://developer.chrome.com/docs/devtools/network), and [Performance](https://developer.chrome.com/docs/devtools/performance) devtools * Edge - [Lighthouse](https://learn.microsoft.com/en-us/microsoft-edge/devtools-guide-chromium/lighthouse/lighthouse-tool), [Network](https://learn.microsoft.com/en-us/microsoft-edge/devtools-guide-chromium/network/), and [Performance](https://learn.microsoft.com/en-us/microsoft-edge/devtools-guide-chromium/evaluate-performance/) devtools * Firefox - [Network](https://firefox-source-docs.mozilla.org/devtools-user/network_monitor/) and [Performance](https://hacks.mozilla.org/2022/03/performance-tool-in-firefox-devtools-reloaded/) devtools * Safari - [enhancing the performance of your webpage](https://developer.apple.com/library/archive/documentation/NetworkingInternetWeb/Conceptual/Web_Inspector_Tutorial/EnhancingyourWebpagesPerformance/EnhancingyourWebpagesPerformance.html) Note that your site running locally in `dev` mode will exhibit different behaviour than your production app, so you should do performance testing in [preview](building-your-app#Preview-your-app) mode after building. ### Instrumenting If you see in the network tab of your browser that an API call is taking a long time and you'd like to understand why, you may consider instrumenting your backend with a tool like [OpenTelemetry](https://opentelemetry.io/) or [Server-Timing headers](https://developer.mozilla.org/en-US/docs/Web/HTTP/Headers/Server-Timing). ## Optimizing assets ### Images Reducing the size of image files is often one of the most impactful changes you can make to a site's performance. Svelte provides the `@sveltejs/enhanced-img` package, detailed on the [images](images) page, for making this easier. Additionally, Lighthouse is useful for identifying the worst offenders. ### Videos Video files can be very large, so extra care should be taken to ensure that they're optimized: - Compress videos with tools such as [Handbrake](https://handbrake.fr/). Consider converting the videos to web-friendly formats such as `.webm` or `.mp4`. - You can [lazy-load videos](https://web.dev/articles/lazy-loading-video) located below the fold with `preload="none"` (though note that this will slow down playback when the user _does_ initiate it). - Strip the audio track out of muted videos using a tool like [FFmpeg](https://ffmpeg.org/). ### Fonts SvelteKit automatically preloads critical `.js` and `.css` files when the user visits a page, but it does _not_ preload fonts by default, since this may cause unnecessary files (such as font weights that are referenced by your CSS but not actually used on the current page) to be downloaded. Having said that, preloading fonts correctly can make a big difference to how fast your site feels. In your [`handle`](hooks#Server-hooks-handle) hook, you can call `resolve` with a `preload` filter that includes your fonts. You can reduce the size of font files by [subsetting](https://web.dev/learn/performance/optimize-web-fonts#subset_your_web_fonts) your fonts. ## Reducing code size ### Svelte version We recommend running the latest version of Svelte. Svelte 5 is smaller and faster than Svelte 4, which is smaller and faster than Svelte 3. ### Packages [`rollup-plugin-visualizer`](https://www.npmjs.com/package/rollup-plugin-visualizer) can be helpful for identifying which packages are contributing the most to the size of your site. You may also find opportunities to remove code by manually inspecting the build output (use `build: { minify: false }` in your [Vite config](https://vitejs.dev/config/build-options.html#build-minify) to make the output readable, but remember to undo that before deploying your app), or via the network tab of your browser's devtools. ### External scripts Try to minimize the number of third-party scripts running in the browser. For example, instead of using JavaScript-based analytics consider using server-side implementations, such as those offered by many platforms with SvelteKit adapters including [Cloudflare](https://www.cloudflare.com/web-analytics/), [Netlify](https://docs.netlify.com/monitor-sites/site-analytics/), and [Vercel](https://vercel.com/docs/analytics). To run third party scripts in a web worker (which avoids blocking the main thread), use [Partytown's SvelteKit integration](https://partytown.builder.io/sveltekit). ### Selective loading Code imported with static `import` declarations will be automatically bundled with the rest of your page. If there is a piece of code you need only when some condition is met, use the dynamic `import(...)` form to selectively lazy-load the component. ## Navigation ### Preloading You can speed up client-side navigations by eagerly preloading the necessary code and data, using [link options](link-options). This is configured by default on the `` element when you create a new SvelteKit app. ### Non-essential data For slow-loading data that isn't needed immediately, the object returned from your `load` function can contain promises rather than the data itself. For server `load` functions, this will cause the data to [stream](load#Streaming-with-promises) in after the navigation (or initial page load). ### Preventing waterfalls One of the biggest performance killers is what is referred to as a _waterfall_, which is a series of requests that is made sequentially. This can happen on the server or in the browser. - Asset waterfalls can occur in the browser when your HTML requests JS which requests CSS which requests a background image and web font. SvelteKit will largely solve this class of problems for you by adding [`modulepreload`](https://developer.mozilla.org/en-US/docs/Web/HTML/Attributes/rel/modulepreload) tags or headers, but you should view [the network tab in your devtools](#Diagnosing-issues) to check whether additional resources need to be preloaded. Pay special attention to this if you use web [fonts](#Optimizing-assets-Fonts) since they need to be handled manually. - If a universal `load` function makes an API call to fetch the current user, then uses the details from that response to fetch a list of saved items, and then uses _that_ response to fetch the details for each item, the browser will end up making multiple sequential requests. This is deadly for performance, especially for users that are physically located far from your backend. Avoid this issue by using [server `load` functions](load#Universal-vs-server) where possible. - Server `load` functions are also not immune to waterfalls (though they are much less costly since they rarely involve roundtrips with high latency). For example if you query a database to get the current user and then use that data to make a second query for a list of saved items, it will typically be more performant to issue a single query with a database join. ## Hosting Your frontend should be located in the same data center as your backend to minimize latency. For sites with no central backend, many SvelteKit adapters support deploying to the _edge_, which means handling each user's requests from a nearby server. This can reduce load times significantly. Some adapters even support [configuring deployment on a per-route basis](page-options#config). You should also consider serving images from a CDN (which are typically edge networks) — the hosts for many SvelteKit adapters will do this automatically. Ensure your host uses HTTP/2 or newer. Vite's code splitting creates numerous small files for improved cacheability, which results in excellent performance, but this does assume that your files can be loaded in parallel with HTTP/2. ## Further reading For the most part, building a performant SvelteKit app is the same as building any performant web app. You should be able to apply information from general performance resources such as [Core Web Vitals](https://web.dev/explore/learn-core-web-vitals) to any web experience you build. --- title: Icons --- ## CSS A great way to use icons is to define them purely via CSS. Iconify offers support for [many popular icon sets](https://icon-sets.iconify.design/) that [can be included via CSS](https://iconify.design/docs/usage/css/). This method can also be used with popular CSS frameworks by leveraging the Iconify [Tailwind CSS plugin](https://iconify.design/docs/usage/css/tailwind/) or [UnoCSS plugin](https://iconify.design/docs/usage/css/unocss/). As opposed to libraries based on Svelte components, it doesn't require each icon to be imported into your `.svelte` file. ## Svelte There are many [icon libraries for Svelte](https://www.sveltesociety.dev/packages?category=icons). When choosing an icon library, it is recommended to avoid those that provide a `.svelte` file per icon as these libraries can have thousands of `.svelte` files which really slow down [Vite's dependency optimization](https://vite.dev/guide/dep-pre-bundling.html). This can become especially pathological if the icons are imported both via an umbrella import and subpath import [as described in the `vite-plugin-svelte` FAQ](https://github.com/sveltejs/vite-plugin-svelte/blob/main/docs/faq.md#what-is-going-on-with-vite-and-pre-bundling-dependencies). --- title: Images --- Images can have a big impact on your app's performance. For best results, you should optimize them by doing the following: - generate optimal formats like `.avif` and `.webp` - create different sizes for different screens - ensure that assets can be cached effectively Doing this manually is tedious. There are a variety of techniques you can use, depending on your needs and preferences. ## Vite's built-in handling [Vite will automatically process imported assets](https://vitejs.dev/guide/assets.html) for improved performance. This includes assets referenced via the CSS `url()` function. Hashes will be added to the filenames so that they can be cached, and assets smaller than `assetsInlineLimit` will be inlined. Vite's asset handling is most often used for images, but is also useful for video, audio, etc. ```svelte ` rather than `` tag will be replaced with an `` providing multiple image types and sizes. It's only possible to downscale images without losing quality, which means that you should provide the highest resolution image that you need — smaller versions will be generated for the various device types that may request an image. You should provide your image at 2x resolution for HiDPI displays (a.k.a. retina displays). `` will automatically take care of serving smaller versions to smaller devices. If you wish to add styles to your ``, you should add a `class` and target that. ### Dynamically choosing an image You can also manually import an image asset and pass it to an ``. This is useful when you have a collection of static images and would like to dynamically choose one or [iterate over them](https://github.com/sveltejs/kit/blob/0ab1733e394b6310895a1d3bf0f126ce34531170/sites/kit.svelte.dev/src/routes/home/Showcase.svelte). In this case you will need to update both the `import` statement and `` tag is preprocessed. With these attributes, the browser can reserve the correct amount of space, preventing [layout shift](https://web.dev/articles/cls). If you'd like to use a different `width` and `height` you can style the image with CSS. Because the preprocessor adds a `width` and `height` for you, if you'd like one of the dimensions to be automatically calculated then you will need to specify that: ```svelte ``` ### `srcset` and `sizes` If you have a large image, such as a hero image taking the width of the design, you should specify `sizes` so that smaller versions are requested on smaller devices. E.g. if you have a 1280px image you may want to specify something like: ```svelte ` will generate small images for smaller devices and populate the `srcset` attribute. The smallest picture generated automatically will have a width of 540px. If you'd like smaller images or would otherwise like to specify custom widths, you can do that with the `w` query parameter: ```svelte ` tags with multiple sources. Finally, some CDNs may generate images lazily, which could have a negative performance impact for sites with low traffic and frequently changing images. CDNs can generally be used without any need for a library. However, there are a number of libraries with Svelte support that make it easier. [`@unpic/svelte`](https://unpic.pics/img/svelte/) is a CDN-agnostic library with support for a large number of providers. You may also find that specific CDNs like [Cloudinary](https://svelte.cloudinary.dev/) have Svelte support. Finally, some content management systems (CMS) which support Svelte (such as [Contentful](https://www.contentful.com/sveltekit-starter-guide/), [Storyblok](https://github.com/storyblok/storyblok-svelte), and [Contentstack](https://www.contentstack.com/docs/developers/sample-apps/build-a-starter-website-with-sveltekit-and-contentstack)) have built-in support for image handling. ## Best practices - For each image type, use the appropriate solution from those discussed above. You can mix and match all three solutions in one project. For example, you may use Vite's built-in handling to provide images for `` tag) triggers a full page reload. When this happens, screen readers and other assistive technology will read out the new page's title so that users understand that the page has changed. Since navigation between pages in SvelteKit happens without reloading the page (known as [client-side routing](glossary#Routing)), SvelteKit injects a [live region](https://developer.mozilla.org/en-US/docs/Web/Accessibility/ARIA/ARIA_Live_Regions) onto the page that will read out the new page name after each navigation. This determines the page name to announce by inspecting the `` element. Because of this behavior, every page in your app should have a unique, descriptive title. In SvelteKit, you can do this by placing a `<svelte:head>` element on each page: ```svelte <!--- file: src/routes/+page.svelte ---> <svelte:head> <title>Todo List ``` This will allow screen readers and other assistive technology to identify the new page after a navigation occurs. Providing a descriptive title is also important for [SEO](seo#Manual-setup-title-and-meta). ## Focus management In traditional server-rendered applications, every navigation will reset focus to the top of the page. This ensures that people browsing the web with a keyboard or screen reader will start interacting with the page from the beginning. To simulate this behavior during client-side routing, SvelteKit focuses the `` element after each navigation and [enhanced form submission](form-actions#Progressive-enhancement). There is one exception - if an element with the [`autofocus`](https://developer.mozilla.org/en-US/docs/Web/HTML/Global_attributes/autofocus) attribute is present, SvelteKit will focus that element instead. Make sure to [consider the implications for assistive technology](https://developer.mozilla.org/en-US/docs/Web/HTML/Global_attributes/autofocus#accessibility_considerations) when using that attribute. If you want to customize SvelteKit's focus management, you can use the `afterNavigate` hook: ```js /// ` and `<meta name="description">` elements inside a [`<svelte:head>`](../svelte/svelte-head). Guidance on how to write descriptive titles and descriptions, along with other suggestions on making content understandable by search engines, can be found on Google's [Lighthouse SEO audits](https://web.dev/lighthouse-seo/) documentation. > [!NOTE] A common pattern is to return SEO-related `data` from page [`load`](load) functions, then use it (as [`page.data`]($app-state)) in a `<svelte:head>` in your root [layout](routing#layout). ### Sitemaps [Sitemaps](https://developers.google.com/search/docs/advanced/sitemaps/build-sitemap) help search engines prioritize pages within your site, particularly when you have a large amount of content. You can create a sitemap dynamically using an endpoint: ```js /// file: src/routes/sitemap.xml/+server.js export async function GET() { return new Response( ` <?xml version="1.0" encoding="UTF-8" ?> <urlset xmlns="https://www.sitemaps.org/schemas/sitemap/0.9" xmlns:xhtml="https://www.w3.org/1999/xhtml" xmlns:mobile="https://www.google.com/schemas/sitemap-mobile/1.0" xmlns:news="https://www.google.com/schemas/sitemap-news/0.9" xmlns:image="https://www.google.com/schemas/sitemap-image/1.1" xmlns:video="https://www.google.com/schemas/sitemap-video/1.1" > <!-- <url> elements go here --> </urlset>`.trim(), { headers: { 'Content-Type': 'application/xml' } } ); } ``` ### AMP An unfortunate reality of modern web development is that it is sometimes necessary to create an [Accelerated Mobile Pages (AMP)](https://amp.dev/) version of your site. In SvelteKit this can be done by setting the [`inlineStyleThreshold`](configuration#inlineStyleThreshold) option... ```js /// file: svelte.config.js /** @type {import('@sveltejs/kit').Config} */ const config = { kit: { // since <link rel="stylesheet"> isn't // allowed, inline all styles inlineStyleThreshold: Infinity } }; export default config; ``` ...disabling `csr` in your root `+layout.js`/`+layout.server.js`... ```js /// file: src/routes/+layout.server.js export const csr = false; ``` ...adding `amp` to your `app.html` ```html <html amp> ... ``` ...and transforming the HTML using `transformPageChunk` along with `transform` imported from `@sveltejs/amp`: ```js /// file: src/hooks.server.js import * as amp from '@sveltejs/amp'; /** @type {import('@sveltejs/kit').Handle} */ export async function handle({ event, resolve }) { let buffer = ''; return await resolve(event, { transformPageChunk: ({ html, done }) => { buffer += html; if (done) return amp.transform(buffer); } }); } ``` To prevent shipping any unused CSS as a result of transforming the page to amp, we can use [`dropcss`](https://www.npmjs.com/package/dropcss): ```js // @filename: ambient.d.ts declare module 'dropcss'; // @filename: index.js // ---cut--- /// file: src/hooks.server.js // @errors: 2307 import * as amp from '@sveltejs/amp'; import dropcss from 'dropcss'; /** @type {import('@sveltejs/kit').Handle} */ export async function handle({ event, resolve }) { let buffer = ''; return await resolve(event, { transformPageChunk: ({ html, done }) => { buffer += html; if (done) { let css = ''; const markup = amp .transform(buffer) .replace('⚡', 'amp') // dropcss can't handle this character .replace(/<style amp-custom([^>]*?)>([^]+?)<\/style>/, (match, attributes, contents) => { css = contents; return `<style amp-custom${attributes}></style>`; }); css = dropcss({ css, html: markup }).css; return markup.replace('</style>', `${css}</style>`); } } }); } ``` > [!NOTE] It's a good idea to use the `handle` hook to validate the transformed HTML using `amphtml-validator`, but only if you're prerendering pages since it's very slow. --- title: Best practices --- --- title: Frequently asked questions --- ## Other resources Please see [the Svelte FAQ](../svelte/faq) and [`vite-plugin-svelte` FAQ](https://github.com/sveltejs/vite-plugin-svelte/blob/main/docs/faq.md) as well for the answers to questions deriving from those libraries. ## What can I make with SvelteKit? See [the documentation regarding project types](project-types) for more details. ## How do I include details from package.json in my application? If you'd like to include your application's version number or other information from `package.json` in your application, you can load JSON like so: ```ts // @errors: 2732 /// file: svelte.config.js import pkg from './package.json' with { type: 'json' }; ``` ## How do I fix the error I'm getting trying to include a package? Most issues related to including a library are due to incorrect packaging. You can check if a library's packaging is compatible with Node.js by entering it into [the publint website](https://publint.dev/). Here are a few things to keep in mind when checking if a library is packaged correctly: - `exports` takes precedence over the other entry point fields such as `main` and `module`. Adding an `exports` field may not be backwards-compatible as it prevents deep imports. - ESM files should end with `.mjs` unless `"type": "module"` is set in which any case CommonJS files should end with `.cjs`. - `main` should be defined if `exports` is not. It should be either a CommonJS or ESM file and adhere to the previous bullet. If a `module` field is defined, it should refer to an ESM file. - Svelte components should be distributed as uncompiled `.svelte` files with any JS in the package written as ESM only. Custom script and style languages, like TypeScript and SCSS, should be preprocessed as vanilla JS and CSS respectively. We recommend using [`svelte-package`](./packaging) for packaging Svelte libraries, which will do this for you. Libraries work best in the browser with Vite when they distribute an ESM version, especially if they are dependencies of a Svelte component library. You may wish to suggest to library authors that they provide an ESM version. However, CommonJS (CJS) dependencies should work as well since, by default, [`vite-plugin-svelte` will ask Vite to pre-bundle them](https://github.com/sveltejs/vite-plugin-svelte/blob/main/docs/faq.md#what-is-going-on-with-vite-and-pre-bundling-dependencies) using `esbuild` to convert them to ESM. If you are still encountering issues we recommend searching both [the Vite issue tracker](https://github.com/vitejs/vite/issues) and the issue tracker of the library in question. Sometimes issues can be worked around by fiddling with the [`optimizeDeps`](https://vitejs.dev/config/#dep-optimization-options) or [`ssr`](https://vitejs.dev/config/#ssr-options) config values though we recommend this as only a short-term workaround in favor of fixing the library in question. ## How do I use the view transitions API? While SvelteKit does not have any specific integration with [view transitions](https://developer.chrome.com/docs/web-platform/view-transitions/), you can call `document.startViewTransition` in [`onNavigate`]($app-navigation#onNavigate) to trigger a view transition on every client-side navigation. ```js // @errors: 2339 2810 import { onNavigate } from '$app/navigation'; onNavigate((navigation) => { if (!document.startViewTransition) return; return new Promise((resolve) => { document.startViewTransition(async () => { resolve(); await navigation.complete; }); }); }); ``` For more, see ["Unlocking view transitions"](/blog/view-transitions) on the Svelte blog. ## How do I set up a database? Put the code to query your database in a [server route](./routing#server) - don't query the database in .svelte files. You can create a `db.js` or similar that sets up a connection immediately and makes the client accessible throughout the app as a singleton. You can execute any one-time setup code in `hooks.server.js` and import your database helpers into any endpoint that needs them. You can use [the Svelte CLI](/docs/cli/overview) to automatically set up database integrations. ## How do I use a client-side library accessing `document` or `window`? If you need access to the `document` or `window` variables or otherwise need code to run only on the client-side you can wrap it in a `browser` check: ```js /// <reference types="@sveltejs/kit" /> // ---cut--- import { browser } from '$app/environment'; if (browser) { // client-only code here } ``` You can also run code in `onMount` if you'd like to run it after the component has been first rendered to the DOM: ```js // @filename: ambient.d.ts // @lib: ES2015 declare module 'some-browser-only-library'; // @filename: index.js // ---cut--- import { onMount } from 'svelte'; onMount(async () => { const { method } = await import('some-browser-only-library'); method('hello world'); }); ``` If the library you'd like to use is side-effect free you can also statically import it and it will be tree-shaken out in the server-side build where `onMount` will be automatically replaced with a no-op: ```js // @filename: ambient.d.ts // @lib: ES2015 declare module 'some-browser-only-library'; // @filename: index.js // ---cut--- import { onMount } from 'svelte'; import { method } from 'some-browser-only-library'; onMount(() => { method('hello world'); }); ``` Finally, you may also consider using an `{#await}` block: ```svelte <!--- file: index.svelte ---> <script> import { browser } from '$app/environment'; const ComponentConstructor = browser ? import('some-browser-only-library').then((module) => module.Component) : new Promise(() => {}); </script> {#await ComponentConstructor} <p>Loading...</p> {:then component} <svelte:component this={component} /> {:catch error} <p>Something went wrong: {error.message}</p> {/await} ``` ## How do I use a different backend API server? You can use [`event.fetch`](./load#Making-fetch-requests) to request data from an external API server, but be aware that you would need to deal with [CORS](https://developer.mozilla.org/en-US/docs/Web/HTTP/CORS), which will result in complications such as generally requiring requests to be preflighted resulting in higher latency. Requests to a separate subdomain may also increase latency due to an additional DNS lookup, TLS setup, etc. If you wish to use this method, you may find [`handleFetch`](./hooks#Server-hooks-handleFetch) helpful. Another approach is to set up a proxy to bypass CORS headaches. In production, you would rewrite a path like `/api` to the API server; for local development, use Vite's [`server.proxy`](https://vitejs.dev/config/server-options.html#server-proxy) option. How to setup rewrites in production will depend on your deployment platform. If rewrites aren't an option, you could alternatively add an [API route](./routing#server): ```js /// file: src/routes/api/[...path]/+server.js /** @type {import('./$types').RequestHandler} */ export function GET({ params, url }) { return fetch(`https://my-api-server.com/${params.path + url.search}`); } ``` (Note that you may also need to proxy `POST`/`PATCH` etc requests, and forward `request.headers`, depending on your needs.) ## How do I use middleware? `adapter-node` builds a middleware that you can use with your own server for production mode. In dev, you can add middleware to Vite by using a Vite plugin. For example: ```js // @errors: 2322 // @filename: ambient.d.ts declare module '@sveltejs/kit/vite'; // TODO this feels unnecessary, why can't it 'see' the declarations? // @filename: index.js // ---cut--- import { sveltekit } from '@sveltejs/kit/vite'; /** @type {import('vite').Plugin} */ const myPlugin = { name: 'log-request-middleware', configureServer(server) { server.middlewares.use((req, res, next) => { console.log(`Got request ${req.url}`); next(); }); } }; /** @type {import('vite').UserConfig} */ const config = { plugins: [myPlugin, sveltekit()] }; export default config; ``` See [Vite's `configureServer` docs](https://vitejs.dev/guide/api-plugin.html#configureserver) for more details including how to control ordering. ## How do I use Yarn? ### Does it work with Yarn 2? Sort of. The Plug'n'Play feature, aka 'pnp', is broken (it deviates from the Node module resolution algorithm, and [doesn't yet work with native JavaScript modules](https://github.com/yarnpkg/berry/issues/638) which SvelteKit — along with an [increasing number of packages](https://blog.sindresorhus.com/get-ready-for-esm-aa53530b3f77) — uses). You can use `nodeLinker: 'node-modules'` in your [`.yarnrc.yml`](https://yarnpkg.com/configuration/yarnrc#nodeLinker) file to disable pnp, but it's probably easier to just use npm or [pnpm](https://pnpm.io/), which is similarly fast and efficient but without the compatibility headaches. ### How do I use with Yarn 3? Currently ESM Support within the latest Yarn (version 3) is considered [experimental](https://github.com/yarnpkg/berry/pull/2161). The below seems to work although your results may vary. First create a new application: ```sh yarn create svelte myapp cd myapp ``` And enable Yarn Berry: ```sh yarn set version berry yarn install ``` #### Yarn 3 global cache One of the more interesting features of Yarn Berry is the ability to have a single global cache for packages, instead of having multiple copies for each project on the disk. However, setting `enableGlobalCache` to true causes building to fail, so it is recommended to add the following to the `.yarnrc.yml` file: ```yaml nodeLinker: node-modules ``` This will cause packages to be downloaded into a local node_modules directory but avoids the above problem and is your best bet for using version 3 of Yarn at this point in time. --- title: Integrations --- ## `vitePreprocess` Including [`vitePreprocess`](https://github.com/sveltejs/vite-plugin-svelte/blob/main/docs/preprocess.md) in your project will allow you to use the various flavors of CSS that Vite supports: PostCSS, SCSS, Less, Stylus, and SugarSS. If you set your project up with TypeScript it will be included by default: ```js // svelte.config.js import { vitePreprocess } from '@sveltejs/vite-plugin-svelte'; export default { preprocess: [vitePreprocess()] }; ``` You will also need to use a preprocessor if you're using TypeScript with Svelte 4. TypeScript is supported natively in Svelte 5 if you're using only the type syntax. To use more complex TypeScript syntax in Svelte 5, you will need still need a preprocessor and can use `vitePreprocess({ script: true })`. ## Adders Run `npx sv add` to setup many different complex integrations with a single command including: - prettier (formatting) - eslint (linting) - vitest (unit testing) - playwright (e2e testing) - lucia (auth) - tailwind (CSS) - drizzle (DB) - paraglide (i18n) - mdsvex (markdown) - storybook (frontend workshop) ## Directory See [sveltesociety.dev](https://sveltesociety.dev/) for a full listing of [packages](https://sveltesociety.dev/packages) and [templates](https://sveltesociety.dev/templates) available for use with Svelte and SvelteKit. ## Additional integrations ### `svelte-preprocess` `svelte-preprocess` has some additional functionality not found in `vitePreprocess` such as support for Pug, Babel, and global styles. However, `vitePreprocess` may be faster and require less configuration, so it is used by default. Note that CoffeeScript is [not supported](https://github.com/sveltejs/kit/issues/2920#issuecomment-996469815) by SvelteKit. You will need to install `svelte-preprocess` with `npm install --save-dev svelte-preprocess` and [add it to your `svelte.config.js`](https://github.com/sveltejs/svelte-preprocess/blob/main/docs/usage.md#with-svelte-config). After that, you will often need to [install the corresponding library](https://github.com/sveltejs/svelte-preprocess/blob/main/docs/getting-started.md) such as `npm install -D sass` or `npm install -D less`. ## Vite plugins Since SvelteKit projects are built with Vite, you can use Vite plugins to enhance your project. See a list of available plugins at [`vitejs/awesome-vite`](https://github.com/vitejs/awesome-vite?tab=readme-ov-file#plugins). ## Integration FAQs [The SvelteKit FAQ](./faq) answers many questions about how to do X with SvelteKit, which may be helpful if you still have questions. --- title: Breakpoint Debugging --- In addition to the [`@debug`](../svelte/@debug) tag, you can also debug Svelte and SvelteKit projects using breakpoints within various tools and development environments. This includes both frontend and backend code. The following guides assume your JavaScript runtime environment is Node.js. ## Visual Studio Code With the built-in debug terminal, you can set up breakpoints in source files within VSCode. 1. Open the command palette: `CMD/Ctrl` + `Shift` + `P`. 2. Find and launch "Debug: JavaScript Debug Terminal". 3. Start your project using the debug terminal. For example: `npm run dev`. 4. Set some breakpoints in your client or server-side source code. 5. Trigger the breakpoint. ### Launch via debug pane You may alternatively set up a `.vscode/launch.json` in your project. To set one up automatically: 1. Go to the "Run and Debug" pane. 2. In the "Run" select menu, choose "Node.js...". 3. Select the "run script" that corresponds to your project, such as "Run script: dev". 4. Press the "Start debugging" play button, or hit `F5` to begin breakpoint debugging. Here's an example `launch.json`: ```json { "version": "0.2.0", "configurations": [ { "command": "npm run dev", "name": "Run development server", "request": "launch", "type": "node-terminal" } ] } ``` Further reading: <https://code.visualstudio.com/docs/editor/debugging>. ## Other Editors If you use a different editor, these community guides might be useful for you: - [WebStorm Svelte: Debug Your Application](https://www.jetbrains.com/help/webstorm/svelte.html#ws_svelte_debug) - [Debugging JavaScript Frameworks in Neovim](https://theosteiner.de/debugging-javascript-frameworks-in-neovim) ## Google Chrome and Microsoft Edge Developer Tools It's possible to debug Node.js applications using a browser-based debugger. > [!NOTE] Note this only works with debugging client-side SvelteKit source maps. 1. Run the `--inspect` flag when starting the Vite server with Node.js. For instance: `NODE_OPTIONS="--inspect" npm run dev` 2. Open your site in a new tab. Typically at `localhost:5173`. 3. Open your browser's dev tools, and click on the "Open dedicated DevTools for Node.js" icon near the top-left. It should display the Node.js logo. 4. Set up breakpoints and debug your application. You may alternatively open the debugger devtools by navigating to `chrome://inspect` in Google Chrome, or `edge://inspect` in Microsoft Edge. ## References - [Debugging Node.js](https://nodejs.org/en/learn/getting-started/debugging) --- title: Migrating to SvelteKit v2 --- Upgrading from SvelteKit version 1 to version 2 should be mostly seamless. There are a few breaking changes to note, which are listed here. You can use `npx sv migrate sveltekit-2` to migrate some of these changes automatically. We highly recommend upgrading to the most recent 1.x version before upgrading to 2.0, so that you can take advantage of targeted deprecation warnings. We also recommend [updating to Svelte 4](../svelte/v4-migration-guide) first: Later versions of SvelteKit 1.x support it, and SvelteKit 2.0 requires it. ## `redirect` and `error` are no longer thrown by you Previously, you had to `throw` the values returned from `error(...)` and `redirect(...)` yourself. In SvelteKit 2 this is no longer the case — calling the functions is sufficient. ```js import { error } from '@sveltejs/kit' // ... ---throw error(500, 'something went wrong');--- +++error(500, 'something went wrong');+++ ``` `svelte-migrate` will do these changes automatically for you. If the error or redirect is thrown inside a `try {...}` block (hint: don't do this!), you can distinguish them from unexpected errors using [`isHttpError`](@sveltejs-kit#isHttpError) and [`isRedirect`](@sveltejs-kit#isRedirect) imported from `@sveltejs/kit`. ## path is required when setting cookies When receiving a `Set-Cookie` header that doesn't specify a `path`, browsers will [set the cookie path](https://www.rfc-editor.org/rfc/rfc6265#section-5.1.4) to the parent of the resource in question. This behaviour isn't particularly helpful or intuitive, and frequently results in bugs because the developer expected the cookie to apply to the domain as a whole. As of SvelteKit 2.0, you need to set a `path` when calling `cookies.set(...)`, `cookies.delete(...)` or `cookies.serialize(...)` so that there's no ambiguity. Most of the time, you probably want to use `path: '/'`, but you can set it to whatever you like, including relative paths — `''` means 'the current path', `'.'` means 'the current directory'. ```js /** @type {import('./$types').PageServerLoad} */ export function load({ cookies }) { cookies.set(name, value, +++{ path: '/' }+++); return { response } } ``` `svelte-migrate` will add comments highlighting the locations that need to be adjusted. ## Top-level promises are no longer awaited In SvelteKit version 1, if the top-level properties of the object returned from a `load` function were promises, they were automatically awaited. With the introduction of [streaming](/blog/streaming-snapshots-sveltekit) this behavior became a bit awkward as it forces you to nest your streamed data one level deep. As of version 2, SvelteKit no longer differentiates between top-level and non-top-level promises. To get back the blocking behavior, use `await` (with `Promise.all` to prevent waterfalls, where appropriate): ```js // @filename: ambient.d.ts declare const url: string; // @filename: index.js // ---cut--- // If you have a single promise /** @type {import('./$types').PageServerLoad} */ export +++async+++ function load({ fetch }) { const response = +++await+++ fetch(url).then(r => r.json()); return { response } } ``` ```js // @filename: ambient.d.ts declare const url1: string; declare const url2: string; // @filename: index.js // ---cut--- // If you have multiple promises /** @type {import('./$types').PageServerLoad} */ export +++async+++ function load({ fetch }) { --- const a = fetch(url1).then(r => r.json());--- --- const b = fetch(url2).then(r => r.json());--- +++ const [a, b] = await Promise.all([ fetch(url1).then(r => r.json()), fetch(url2).then(r => r.json()), ]);+++ return { a, b }; } ``` ## goto(...) changes `goto(...)` no longer accepts external URLs. To navigate to an external URL, use `window.location.href = url`. The `state` object now determines `$page.state` and must adhere to the `App.PageState` interface, if declared. See [shallow routing](shallow-routing) for more details. ## paths are now relative by default In SvelteKit 1, `%sveltekit.assets%` in your `app.html` was replaced with a relative path by default (i.e. `.` or `..` or `../..` etc, depending on the path being rendered) during server-side rendering unless the [`paths.relative`](configuration#paths) config option was explicitly set to `false`. The same was true for `base` and `assets` imported from `$app/paths`, but only if the `paths.relative` option was explicitly set to `true`. This inconsistency is fixed in version 2. Paths are either always relative or always absolute, depending on the value of [`paths.relative`](configuration#paths). It defaults to `true` as this results in more portable apps: if the `base` is something other than the app expected (as is the case when viewed on the [Internet Archive](https://archive.org/), for example) or unknown at build time (as is the case when deploying to [IPFS](https://ipfs.tech/) and so on), fewer things are likely to break. ## Server fetches are not trackable anymore Previously it was possible to track URLs from `fetch`es on the server in order to rerun load functions. This poses a possible security risk (private URLs leaking), and as such it was behind the `dangerZone.trackServerFetches` setting, which is now removed. ## `preloadCode` arguments must be prefixed with `base` SvelteKit exposes two functions, [`preloadCode`]($app-navigation#preloadCode) and [`preloadData`]($app-navigation#preloadData), for programmatically loading the code and data associated with a particular path. In version 1, there was a subtle inconsistency — the path passed to `preloadCode` did not need to be prefixed with the `base` path (if set), while the path passed to `preloadData` did. This is fixed in SvelteKit 2 — in both cases, the path should be prefixed with `base` if it is set. Additionally, `preloadCode` now takes a single argument rather than _n_ arguments. ## `resolvePath` has been removed SvelteKit 1 included a function called `resolvePath` which allows you to resolve a route ID (like `/blog/[slug]`) and a set of parameters (like `{ slug: 'hello' }`) to a pathname. Unfortunately the return value didn't include the `base` path, limiting its usefulness in cases where `base` was set. As such, SvelteKit 2 replaces `resolvePath` with a (slightly better named) function called `resolveRoute`, which is imported from `$app/paths` and which takes `base` into account. ```js ---import { resolvePath } from '@sveltejs/kit'; import { base } from '$app/paths';--- +++import { resolveRoute } from '$app/paths';+++ ---const path = base + resolvePath('/blog/[slug]', { slug });--- +++const path = resolveRoute('/blog/[slug]', { slug });+++ ``` `svelte-migrate` will do the method replacement for you, though if you later prepend the result with `base`, you need to remove that yourself. ## Improved error handling Errors are handled inconsistently in SvelteKit 1. Some errors trigger the `handleError` hook but there is no good way to discern their status (for example, the only way to tell a 404 from a 500 is by seeing if `event.route.id` is `null`), while others (such as 405 errors for `POST` requests to pages without actions) don't trigger `handleError` at all, but should. In the latter case, the resulting `$page.error` will deviate from the [`App.Error`](types#Error) type, if it is specified. SvelteKit 2 cleans this up by calling `handleError` hooks with two new properties: `status` and `message`. For errors thrown from your code (or library code called by your code) the status will be `500` and the message will be `Internal Error`. While `error.message` may contain sensitive information that should not be exposed to users, `message` is safe. ## Dynamic environment variables cannot be used during prerendering The `$env/dynamic/public` and `$env/dynamic/private` modules provide access to _run time_ environment variables, as opposed to the _build time_ environment variables exposed by `$env/static/public` and `$env/static/private`. During prerendering in SvelteKit 1, they are one and the same. As such, prerendered pages that make use of 'dynamic' environment variables are really 'baking in' build time values, which is incorrect. Worse, `$env/dynamic/public` is populated in the browser with these stale values if the user happens to land on a prerendered page before navigating to dynamically-rendered pages. Because of this, dynamic environment variables can no longer be read during prerendering in SvelteKit 2 — you should use the `static` modules instead. If the user lands on a prerendered page, SvelteKit will request up-to-date values for `$env/dynamic/public` from the server (by default from a module called `/_app/env.js`) instead of reading them from the server-rendered HTML. ## `form` and `data` have been removed from `use:enhance` callbacks If you provide a callback to [`use:enhance`](form-actions#Progressive-enhancement-use:enhance), it will be called with an object containing various useful properties. In SvelteKit 1, those properties included `form` and `data`. These were deprecated some time ago in favour of `formElement` and `formData`, and have been removed altogether in SvelteKit 2. ## Forms containing file inputs must use `multipart/form-data` If a form contains an `<input type="file">` but does not have an `enctype="multipart/form-data"` attribute, non-JS submissions will omit the file. SvelteKit 2 will throw an error if it encounters a form like this during a `use:enhance` submission to ensure that your forms work correctly when JavaScript is not present. ## Generated `tsconfig.json` is more strict Previously, the generated `tsconfig.json` was trying its best to still produce a somewhat valid config when your `tsconfig.json` included `paths` or `baseUrl`. In SvelteKit 2, the validation is more strict and will warn when you use either `paths` or `baseUrl` in your `tsconfig.json`. These settings are used to generate path aliases and you should use [the `alias` config](configuration#alias) option in your `svelte.config.js` instead, to also create a corresponding alias for the bundler. ## `getRequest` no longer throws errors The `@sveltejs/kit/node` module exports helper functions for use in Node environments, including `getRequest` which turns a Node [`ClientRequest`](https://nodejs.org/api/http.html#class-httpclientrequest) into a standard [`Request`](https://developer.mozilla.org/en-US/docs/Web/API/Request) object. In SvelteKit 1, `getRequest` could throw if the `Content-Length` header exceeded the specified size limit. In SvelteKit 2, the error will not be thrown until later, when the request body (if any) is being read. This enables better diagnostics and simpler code. ## `vitePreprocess` is no longer exported from `@sveltejs/kit/vite` Since `@sveltejs/vite-plugin-svelte` is now a peer dependency, SvelteKit 2 no longer re-exports `vitePreprocess`. You should import it directly from `@sveltejs/vite-plugin-svelte`. ## Updated dependency requirements SvelteKit 2 requires Node `18.13` or higher, and the following minimum dependency versions: - `svelte@4` - `vite@5` - `typescript@5` - `@sveltejs/vite-plugin-svelte@3` (this is now required as a `peerDependency` of SvelteKit — previously it was directly depended upon) - `@sveltejs/adapter-cloudflare@3` (if you're using these adapters) - `@sveltejs/adapter-cloudflare-workers@2` - `@sveltejs/adapter-netlify@3` - `@sveltejs/adapter-node@2` - `@sveltejs/adapter-static@3` - `@sveltejs/adapter-vercel@4` `svelte-migrate` will update your `package.json` for you. As part of the TypeScript upgrade, the generated `tsconfig.json` (the one your `tsconfig.json` extends from) now uses `"moduleResolution": "bundler"` (which is recommended by the TypeScript team, as it properly resolves types from packages with an `exports` map in package.json) and `verbatimModuleSyntax` (which replaces the existing `importsNotUsedAsValues ` and `preserveValueImports` flags — if you have those in your `tsconfig.json`, remove them. `svelte-migrate` will do this for you). ## SvelteKit 2.12: $app/stores deprecated SvelteKit 2.12 introduced `$app/state` based on the [Svelte 5 runes API](/docs/svelte/what-are-runes). `$app/state` provides everything that `$app/stores` provides but with more flexibility as to where and how you use it. Most importantly, the `page` object is now fine-grained, e.g. updates to `page.state` will not invalidate `page.data` and vice-versa. As a consequence, `$app/stores` is deprecated and subject to be removed in SvelteKit 3. We recommend [upgrading to Svelte 5](/docs/svelte/v5-migration-guide), if you haven't already, and then migrate away from `$app/stores`. Most of the replacements should be pretty simple: Replace the `$app/stores` import with `$app/state` and remove the `$` prefixes from the usage sites. ```svelte <script> ---import { page } from '$app/stores';--- +++import { page } from '$app/state';+++ </script> ---{$page.data}--- +++{page.data}+++ ``` Use `npx sv migrate app-state` to auto-migrate most of your `$app/stores` usages inside `.svelte` components. --- title: Migrating from Sapper rank: 1 --- SvelteKit is the successor to Sapper and shares many elements of its design. If you have an existing Sapper app that you plan to migrate to SvelteKit, there are a number of changes you will need to make. You may find it helpful to view [some examples](additional-resources#Examples) while migrating. ## package.json ### type: "module" Add `"type": "module"` to your `package.json`. You can do this step separately from the rest as part of an incremental migration if you are using Sapper 0.29.3 or newer. ### dependencies Remove `polka` or `express`, if you're using one of those, and any middleware such as `sirv` or `compression`. ### devDependencies Remove `sapper` from your `devDependencies` and replace it with `@sveltejs/kit` and whichever [adapter](adapters) you plan to use (see [next section](migrating#Project-files-Configuration)). ### scripts Any scripts that reference `sapper` should be updated: - `sapper build` should become `vite build` using the Node [adapter](adapters) - `sapper export` should become `vite build` using the static [adapter](adapters) - `sapper dev` should become `vite dev` - `node __sapper__/build` should become `node build` ## Project files The bulk of your app, in `src/routes`, can be left where it is, but several project files will need to be moved or updated. ### Configuration Your `webpack.config.js` or `rollup.config.js` should be replaced with a `svelte.config.js`, as documented [here](configuration). Svelte preprocessor options should be moved to `config.preprocess`. You will need to add an [adapter](adapters). `sapper build` is roughly equivalent to [adapter-node](adapter-node) while `sapper export` is roughly equivalent to [adapter-static](adapter-static), though you might prefer to use an adapter designed for the platform you're deploying to. If you were using plugins for filetypes that are not automatically handled by [Vite](https://vitejs.dev), you will need to find Vite equivalents and add them to the [Vite config](project-structure#Project-files-vite.config.js). ### src/client.js This file has no equivalent in SvelteKit. Any custom logic (beyond `sapper.start(...)`) should be expressed in your `+layout.svelte` file, inside an `onMount` callback. ### src/server.js When using `adapter-node` the equivalent is a [custom server](adapter-node#Custom-server). Otherwise, this file has no direct equivalent, since SvelteKit apps can run in serverless environments. ### src/service-worker.js Most imports from `@sapper/service-worker` have equivalents in [`$service-worker`]($service-worker): - `files` is unchanged - `routes` has been removed - `shell` is now `build` - `timestamp` is now `version` ### src/template.html The `src/template.html` file should be renamed `src/app.html`. Remove `%sapper.base%`, `%sapper.scripts%` and `%sapper.styles%`. Replace `%sapper.head%` with `%sveltekit.head%` and `%sapper.html%` with `%sveltekit.body%`. The `<div id="sapper">` is no longer necessary. ### src/node_modules A common pattern in Sapper apps is to put your internal library in a directory inside `src/node_modules`. This doesn't work with Vite, so we use [`src/lib`]($lib) instead. ## Pages and layouts ### Renamed files Routes now are made up of the folder name exclusively to remove ambiguity, the folder names leading up to a `+page.svelte` correspond to the route. See [the routing docs](routing) for an overview. The following shows a old/new comparison: | Old | New | | ------------------------- | ------------------------- | | routes/about/index.svelte | routes/about/+page.svelte | | routes/about.svelte | routes/about/+page.svelte | Your custom error page component should be renamed from `_error.svelte` to `+error.svelte`. Any `_layout.svelte` files should likewise be renamed `+layout.svelte`. [Any other files are ignored](routing#Other-files). ### Imports The `goto`, `prefetch` and `prefetchRoutes` imports from `@sapper/app` should be replaced with `goto`, `preloadData` and `preloadCode` imports respectively from [`$app/navigation`]($app-navigation). The `stores` import from `@sapper/app` should be replaced — see the [Stores](migrating#Pages-and-layouts-Stores) section below. Any files you previously imported from directories in `src/node_modules` will need to be replaced with [`$lib`]($lib) imports. ### Preload As before, pages and layouts can export a function that allows data to be loaded before rendering takes place. This function has been renamed from `preload` to [`load`](load), it now lives in a `+page.js` (or `+layout.js`) next to its `+page.svelte` (or `+layout.svelte`), and its API has changed. Instead of two arguments — `page` and `session` — there is a single `event` argument. There is no more `this` object, and consequently no `this.fetch`, `this.error` or `this.redirect`. Instead, you can get [`fetch`](load#Making-fetch-requests) from the input methods, and both [`error`](load#Errors) and [`redirect`](load#Redirects) are now thrown. ### Stores In Sapper, you would get references to provided stores like so: ```js // @filename: ambient.d.ts declare module '@sapper/app'; // @filename: index.js // ---cut--- import { stores } from '@sapper/app'; const { preloading, page, session } = stores(); ``` The `page` store still exists; `preloading` has been replaced with a `navigating` store that contains `from` and `to` properties. `page` now has `url` and `params` properties, but no `path` or `query`. You access them differently in SvelteKit. `stores` is now `getStores`, but in most cases it is unnecessary since you can import `navigating`, and `page` directly from [`$app/stores`]($app-stores). If you're on Svelte 5 and SvelteKit 2.12 or higher, consider using [`$app/state`]($app-state) instead. ### Routing Regex routes are no longer supported. Instead, use [advanced route matching](advanced-routing#Matching). ### Segments Previously, layout components received a `segment` prop indicating the child segment. This has been removed; you should use the more flexible `$page.url.pathname` (or `page.url.pathname`) value to derive the segment you're interested in. ### URLs In Sapper, all relative URLs were resolved against the base URL — usually `/`, unless the `basepath` option was used — rather than against the current page. This caused problems and is no longer the case in SvelteKit. Instead, relative URLs are resolved against the current page (or the destination page, for `fetch` URLs in `load` functions) instead. In most cases, it's easier to use root-relative (i.e. starts with `/`) URLs, since their meaning is not context-dependent. ### <a> attributes - `sapper:prefetch` is now `data-sveltekit-preload-data` - `sapper:noscroll` is now `data-sveltekit-noscroll` ## Endpoints In Sapper, [server routes](routing#server) received the `req` and `res` objects exposed by Node's `http` module (or the augmented versions provided by frameworks like Polka and Express). SvelteKit is designed to be agnostic as to where the app is running — it could be running on a Node server, but could equally be running on a serverless platform or in a Cloudflare Worker. For that reason, you no longer interact directly with `req` and `res`. Your endpoints will need to be updated to match the new signature. To support this environment-agnostic behavior, `fetch` is now available in the global context, so you don't need to import `node-fetch`, `cross-fetch`, or similar server-side fetch implementations in order to use it. ## Integrations See [integrations](./integrations) for detailed information about integrations. ### HTML minifier Sapper includes `html-minifier` by default. SvelteKit does not include this, but you can add it as a prod dependency and then use it through a [hook](hooks#Server-hooks-handle): ```js // @filename: ambient.d.ts /// <reference types="@sveltejs/kit" /> declare module 'html-minifier'; // @filename: index.js // ---cut--- import { minify } from 'html-minifier'; import { building } from '$app/environment'; const minification_options = { collapseBooleanAttributes: true, collapseWhitespace: true, conservativeCollapse: true, decodeEntities: true, html5: true, ignoreCustomComments: [/^#/], minifyCSS: true, minifyJS: false, removeAttributeQuotes: true, removeComments: false, // some hydration code needs comments, so leave them in removeOptionalTags: true, removeRedundantAttributes: true, removeScriptTypeAttributes: true, removeStyleLinkTypeAttributes: true, sortAttributes: true, sortClassName: true }; /** @type {import('@sveltejs/kit').Handle} */ export async function handle({ event, resolve }) { let page = ''; return resolve(event, { transformPageChunk: ({ html, done }) => { page += html; if (done) { return building ? minify(page, minification_options) : page; } } }); } ``` Note that `prerendering` is `false` when using `vite preview` to test the production build of the site, so to verify the results of minifying, you'll need to inspect the built HTML files directly. --- title: Additional resources --- ## FAQs Please see the [SvelteKit FAQ](faq) for solutions to common issues and helpful tips and tricks. The [Svelte FAQ](../svelte/faq) and [`vite-plugin-svelte` FAQ](https://github.com/sveltejs/vite-plugin-svelte/blob/main/docs/faq.md) may also be helpful for questions deriving from those libraries. ## Examples We've written and published a few different SvelteKit sites as examples: - [`sveltejs/realworld`](https://github.com/sveltejs/realworld) contains an example blog site - [A HackerNews clone](https://github.com/sveltejs/sites/tree/master/sites/hn.svelte.dev) - [`svelte.dev`](https://github.com/sveltejs/svelte.dev) SvelteKit users have also published plenty of examples on GitHub, under the [#sveltekit](https://github.com/topics/sveltekit) and [#sveltekit-template](https://github.com/topics/sveltekit-template) topics, as well as on [the Svelte Society site](https://sveltesociety.dev/templates?category=sveltekit). Note that these have not been vetted by the maintainers and may not be up to date. ## Support You can ask for help on [Discord](/chat) and [StackOverflow](https://stackoverflow.com/questions/tagged/sveltekit). Please first search for information related to your issue in the FAQ, Google or another search engine, issue tracker, and Discord chat history in order to be respectful of others' time. There are many more people asking questions than answering them, so this will help in allowing the community to grow in a scalable fashion. --- title: Glossary --- The core of SvelteKit provides a highly configurable rendering engine. This section describes some of the terms used when discussing rendering. A reference for setting these options is provided in the documentation above. ## CSR Client-side rendering (CSR) is the generation of the page contents in the web browser using JavaScript. In SvelteKit, client-side rendering will be used by default, but you can turn off JavaScript with [the `csr = false` page option](page-options#csr). ## Edge Rendering on the edge refers to rendering an application in a content delivery network (CDN) near the user. Edge rendering allows the request and response for a page to travel a shorter distance thus improving latency. ## Hydration Svelte components store some state and update the DOM when the state is updated. When fetching data during SSR, by default SvelteKit will store this data and transmit it to the client along with the server-rendered HTML. The components can then be initialized on the client with that data without having to call the same API endpoints again. Svelte will then check that the DOM is in the expected state and attach event listeners in a process called hydration. Once the components are fully hydrated, they can react to changes to their properties just like any newly created Svelte component. In SvelteKit, pages will be hydrated by default, but you can turn off JavaScript with [the `csr = false` page option](page-options#csr). ## ISR Incremental static regeneration (ISR) allows you to generate static pages on your site as visitors request those pages without redeploying. This may reduces build times compared to [SSG](#SSG) sites with a large number of pages. You can do [ISR with `adapter-vercel`](adapter-vercel#Incremental-Static-Regeneration). ## MPA Traditional applications that render each page view on the server — such as those written in languages other than JavaScript — are often referred to as multi-page apps (MPA). ## Prerendering Prerendering means computing the contents of a page at build time and saving the HTML for display. This approach has the same benefits as traditional server-rendered pages, but avoids recomputing the page for each visitor and so scales nearly for free as the number of visitors increases. The tradeoff is that the build process is more expensive and prerendered content can only be updated by building and deploying a new version of the application. Not all pages can be prerendered. The basic rule is this: for content to be prerenderable, any two users hitting it directly must get the same content from the server, and the page must not contain [actions](form-actions). Note that you can still prerender content that is loaded based on the page's parameters as long as all users will be seeing the same prerendered content. Pre-rendered pages are not limited to static content. You can build personalized pages if user-specific data is fetched and rendered client-side. This is subject to the caveat that you will experience the downsides of not doing SSR for that content as discussed above. In SvelteKit, you can control prerendering with [the `prerender` page option](page-options#prerender) and [`prerender` config](configuration#prerender) in `svelte.config.js`. ## PWA A progressive web app (PWA) is an app that's built using web APIs and technologies, but functions like a mobile or desktop app. Sites served as [PWAs can be installed](https://web.dev/learn/pwa/installation), allowing you to add a shortcut to the application on your launcher, home screen, or start menu. Many PWAs will utilize [service workers](service-workers) to build offline capabilities. ## Routing By default, when you navigate to a new page (by clicking on a link or using the browser's forward or back buttons), SvelteKit will intercept the attempted navigation and handle it instead of allowing the browser to send a request to the server for the destination page. SvelteKit will then update the displayed contents on the client by rendering the component for the new page, which in turn can make calls to the necessary API endpoints. This process of updating the page on the client in response to attempted navigation is called client-side routing. In SvelteKit, client-side routing will be used by default, but you can skip it with [`data-sveltekit-reload`](link-options#data-sveltekit-reload). ## SPA A single-page app (SPA) is an application in which all requests to the server load a single HTML file which then does client-side rendering of the requested contents based on the requested URL. All navigation is handled on the client-side in a process called client-side routing with per-page contents being updated and common layout elements remaining largely unchanged. SPAs do not provide SSR and thus have worse performance and SEO characteristics. However, some applications are not greatly impacted by these shortcomings such as a complex business application behind a login where SEO would not be important and it is known that users will be accessing the application from a consistent computing environment. In SvelteKit, you can [build an SPA with `adapter-static`](single-page-apps). ## SSG Static Site Generation (SSG) is a term that refers to a site where every page is prerendered. One benefit of fully prerendering a site is that you do not need to maintain or pay for servers to perform SSR. Once generated, the site can be served from CDNs, leading to great “time to first byte” performance. This delivery model is often referred to as JAMstack. In SvelteKit, you can do static site generation by using [`adapter-static`](adapter-static) or by configuring every page to be prerendered using [the `prerender` page option](page-options#prerender) or [`prerender` config](configuration#prerender) in `svelte.config.js`. ## SSR Server-side rendering (SSR) is the generation of the page contents on the server. SSR is generally preferred for SEO. While some search engines can index content that is dynamically generated on the client-side it may take longer even in these cases. It also tends to improve perceived performance and makes your app accessible to users if JavaScript fails or is disabled (which happens [more often than you probably think](https://kryogenix.org/code/browser/everyonehasjs.html)). In SvelteKit, pages are server-side rendered by default. You can disable SSR with [the `ssr` page option](page-options#ssr). --- title: Appendix --- --- title: @sveltejs/kit --- > MODULE: @sveltejs/kit ## Private types The following are referenced by the public types documented above, but cannot be imported directly: > TYPES: Private types --- title: @sveltejs/kit/hooks --- > MODULE: @sveltejs/kit/hooks --- title: @sveltejs/kit/node/polyfills --- > MODULE: @sveltejs/kit/node/polyfills --- title: @sveltejs/kit/node --- > MODULE: @sveltejs/kit/node --- title: @sveltejs/kit/vite --- > MODULE: @sveltejs/kit/vite --- title: $app/environment --- > MODULE: $app/environment --- title: $app/forms --- > MODULE: $app/forms --- title: $app/navigation --- > MODULE: $app/navigation --- title: $app/paths --- > MODULE: $app/paths --- title: $app/server --- > MODULE: $app/server --- title: $app/state --- SvelteKit makes three read-only state objects available via the `$app/state` module — `page`, `navigating` and `updated`. > [!NOTE] > This module was added in 2.12. If you're using an earlier version of SvelteKit, use [`$app/stores`]($app-stores) instead. > MODULE: $app/state --- title: $app/stores --- This module contains store-based equivalents of the exports from [`$app/state`]($app-state). If you're using SvelteKit 2.12 or later, use that module instead. > MODULE: $app/stores --- title: $env/dynamic/private --- > MODULE: $env/dynamic/private --- title: $env/dynamic/public --- > MODULE: $env/dynamic/public --- title: $env/static/private --- > MODULE: $env/static/private --- title: $env/static/public --- > MODULE: $env/static/public --- title: $lib --- SvelteKit automatically makes files under `src/lib` available using the `$lib` import alias. You can change which directory this alias points to in your [config file](configuration#files). ```svelte <!--- file: src/lib/Component.svelte ---> A reusable component ``` ```svelte <!--- file: src/routes/+page.svelte ---> <script> import Component from '$lib/Component.svelte'; </script> <Component /> ``` --- title: $service-worker --- > MODULE: $service-worker --- title: Configuration --- Your project's configuration lives in a `svelte.config.js` file at the root of your project. As well as SvelteKit, this config object is used by other tooling that integrates with Svelte such as editor extensions. ```js /// file: svelte.config.js // @filename: ambient.d.ts declare module '@sveltejs/adapter-auto' { const plugin: () => import('@sveltejs/kit').Adapter; export default plugin; } // @filename: index.js // ---cut--- import adapter from '@sveltejs/adapter-auto'; /** @type {import('@sveltejs/kit').Config} */ const config = { kit: { adapter: adapter() } }; export default config; ``` ## Config > TYPES: Configuration#Config ## KitConfig The `kit` property configures SvelteKit, and can have the following properties: > EXPANDED_TYPES: Configuration#KitConfig --- title: Command Line Interface --- SvelteKit projects use [Vite](https://vitejs.dev), meaning you'll mostly use its CLI (albeit via `npm run dev/build/preview` scripts): - `vite dev` — start a development server - `vite build` — build a production version of your app - `vite preview` — run the production version locally However SvelteKit includes its own CLI for initialising your project: ## svelte-kit sync `svelte-kit sync` creates the `tsconfig.json` and all generated types (which you can import as `./$types` inside routing files) for your project. When you create a new project, it is listed as the `prepare` script and will be run automatically as part of the npm lifecycle, so you should not ordinarily have to run this command. --- title: Types --- ## Generated types The `RequestHandler` and `Load` types both accept a `Params` argument allowing you to type the `params` object. For example this endpoint expects `foo`, `bar` and `baz` params: ```js /// file: src/routes/[foo]/[bar]/[baz]/+server.js // @errors: 2355 2322 1360 /** @type {import('@sveltejs/kit').RequestHandler<{ foo: string; bar: string; baz: string }>} */ export async function GET({ params }) { // ... } ``` Needless to say, this is cumbersome to write out, and less portable (if you were to rename the `[foo]` directory to `[qux]`, the type would no longer reflect reality). To solve this problem, SvelteKit generates `.d.ts` files for each of your endpoints and pages: ```ts /// file: .svelte-kit/types/src/routes/[foo]/[bar]/[baz]/$types.d.ts /// link: true import type * as Kit from '@sveltejs/kit'; type RouteParams = { foo: string; bar: string; baz: string; }; export type RequestHandler = Kit.RequestHandler<RouteParams>; export type PageLoad = Kit.Load<RouteParams>; ``` These files can be imported into your endpoints and pages as siblings, thanks to the [`rootDirs`](https://www.typescriptlang.org/tsconfig#rootDirs) option in your TypeScript configuration: ```js /// file: src/routes/[foo]/[bar]/[baz]/+server.js // @filename: $types.d.ts import type * as Kit from '@sveltejs/kit'; type RouteParams = { foo: string; bar: string; baz: string; } export type RequestHandler = Kit.RequestHandler<RouteParams>; // @filename: index.js // @errors: 2355 2322 // ---cut--- /** @type {import('./$types').RequestHandler} */ export async function GET({ params }) { // ... } ``` ```js /// file: src/routes/[foo]/[bar]/[baz]/+page.js // @filename: $types.d.ts import type * as Kit from '@sveltejs/kit'; type RouteParams = { foo: string; bar: string; baz: string; } export type PageLoad = Kit.Load<RouteParams>; // @filename: index.js // @errors: 2355 // ---cut--- /** @type {import('./$types').PageLoad} */ export async function load({ params, fetch }) { // ... } ``` The return types of the load functions are then available through the `$types` module as `PageData` and `LayoutData` respectively, while the union of the return values of all `Actions` is available as `ActionData`. Starting with version 2.16.0, two additional helper types are provided: `PageProps` defines `data: PageData`, as well as `form: ActionData`, when there are actions defined, while `LayoutProps` defines `data: LayoutData`, as well as `children: Snippet`. ```svelte <!--- file: src/routes/+page.svelte ---> <script> /** @type {import('./$types').PageProps} */ let { data, form } = $props(); </script> ``` > [!LEGACY] > Before 2.16.0: > ```svelte > <!--- file: src/routes/+page.svelte ---> > <script> > /** @type {{ data: import('./$types').PageData, form: import('./$types').ActionData }} */ > let { data, form } = $props(); > </script> > ``` > > Using Svelte 4: > ```svelte > <!--- file: src/routes/+page.svelte ---> > <script> > /** @type {import('./$types').PageData} */ > export let data; > /** @type {import('./$types').ActionData} */ > export let form; > </script> > ``` > [!NOTE] For this to work, your own `tsconfig.json` or `jsconfig.json` should extend from the generated `.svelte-kit/tsconfig.json` (where `.svelte-kit` is your [`outDir`](configuration#outDir)): > > `{ "extends": "./.svelte-kit/tsconfig.json" }` ### Default tsconfig.json The generated `.svelte-kit/tsconfig.json` file contains a mixture of options. Some are generated programmatically based on your project configuration, and should generally not be overridden without good reason: ```json /// file: .svelte-kit/tsconfig.json { "compilerOptions": { "paths": { "$lib": ["../src/lib"], "$lib/*": ["../src/lib/*"] }, "rootDirs": ["..", "./types"] }, "include": [ "ambient.d.ts", "non-ambient.d.ts", "./types/**/$types.d.ts", "../vite.config.js", "../vite.config.ts", "../src/**/*.js", "../src/**/*.ts", "../src/**/*.svelte", "../tests/**/*.js", "../tests/**/*.ts", "../tests/**/*.svelte" ], "exclude": [ "../node_modules/**", "../src/service-worker.js", "../src/service-worker/**/*.js", "../src/service-worker.ts", "../src/service-worker/**/*.ts", "../src/service-worker.d.ts", "../src/service-worker/**/*.d.ts" ] } ``` Others are required for SvelteKit to work properly, and should also be left untouched unless you know what you're doing: ```json /// file: .svelte-kit/tsconfig.json { "compilerOptions": { // this ensures that types are explicitly // imported with `import type`, which is // necessary as Svelte/Vite cannot // otherwise compile components correctly "verbatimModuleSyntax": true, // Vite compiles one TypeScript module // at a time, rather than compiling // the entire module graph "isolatedModules": true, // Tell TS it's used only for type-checking "noEmit": true, // This ensures both `vite build` // and `svelte-package` work correctly "lib": ["esnext", "DOM", "DOM.Iterable"], "moduleResolution": "bundler", "module": "esnext", "target": "esnext" } } ``` ## $lib This is a simple alias to `src/lib`, or whatever directory is specified as [`config.kit.files.lib`](configuration#files). It allows you to access common components and utility modules without `../../../../` nonsense. ### $lib/server A subdirectory of `$lib`. SvelteKit will prevent you from importing any modules in `$lib/server` into client-side code. See [server-only modules](server-only-modules). ## app.d.ts The `app.d.ts` file is home to the ambient types of your apps, i.e. types that are available without explicitly importing them. Always part of this file is the `App` namespace. This namespace contains several types that influence the shape of certain SvelteKit features you interact with. > TYPES: App --- title: Reference --- --- title: SvelteKit ---