Nuxt comes with two composables and a built-in library to perform data-fetching in browser or server environments: `useFetch`, [`useAsyncData`](/docs/api/composables/use-async-data) and `$fetch`.
Nuxt is a framework which can run isomorphic (or universal) code in both server and client environments. If the [`$fetch` function](/docs/api/utils/dollarfetch) is used to perform data fetching in the setup function of a Vue component, this may cause data to be fetched twice, once on the server (to render the HTML) and once again on the client (when the HTML is hydrated). This is why Nuxt offers specific data fetching composables so data is fetched only once.
The [`useFetch`](/docs/api/composables/use-fetch) and [`useAsyncData`](/docs/api/composables/use-async-data) composables ensure that once an API call is made on the server, the data is properly forwarded to the client in the payload.
The payload is a JavaScript object accessible through [`useNuxtApp().payload`](/docs/api/composables/use-nuxt-app#payload). It is used on the client to avoid refetching the same data when the code is executed in the browser [during hydration](/docs/guide/concepts/rendering#universal-rendering).
Nuxt uses Vue’s [`<Suspense>`](https://vuejs.org/guide/built-ins/suspense) component under the hood to prevent navigation before every async data is available to the view. The data fetching composables can help you leverage this feature and use what suits best on a per-calls basis.
Nuxt includes the [ofetch](https://github.com/unjs/ofetch) library, and is auto-imported as the `$fetch` alias globally across your application. It's what `useFetch` uses behind the scenes.
Beware that using only `$fetch` will not provide [network calls de-duplication and navigation prevention](#why-use-specific-composables-for-data-fetching). :br
It is recommended to use `$fetch` for client-side interactions (event based) or combined with [`useAsyncData`](#useasyncdata) when fetching the initial component data.
There are some cases when using the [`useFetch`](/docs/api/composables/use-fetch) composable is not appropriate, for example when a CMS or a third-party provide their own query layer. In this case, you can use [`useAsyncData`](/docs/api/composables/use-async-data) to wrap your calls and still keep the benefits provided by the composable.
The first argument of [`useAsyncData`](/docs/api/composables/use-async-data) is a unique key used to cache the response of the second argument, the querying function. This key can be ignored by directly passing the querying function, the key will be auto-generated.
:br :br
Since the autogenerated key only takes into account the file and line where `useAsyncData` is invoked, it is recommended to always create your own key to avoid unwanted behavior, like when you are creating your own custom composable wrapping `useAsyncData`.
:br :br
Setting a key can be useful to share the same data between components using [`useNuxtData`](/docs/api/composables/use-nuxt-data) or to [refresh specific data](/docs/api/utils/refresh-nuxt-data#refresh-specific-data).
-`clear`: a function that can be used to set `data` to `undefined`, set `error` to `null`, set `status` to `idle`, and mark any currently pending requests as cancelled.
If you have not fetched data on the server (for example, with `server: false`), then the data _will not_ be fetched until hydration completes. This means even if you await `useFetch` on client-side, `data` will remain null within `<script setup>`.
[`useAsyncData`](/docs/api/composables/use-async-data) and [`useFetch`](/docs/api/composables/use-fetch) return the same object type and accept a common set of options as their last argument. They can help you control the composables behavior, such as navigation blocking, caching or execution.
By default, data fetching composables will wait for the resolution of their asynchronous function before navigating to a new page by using Vue’s Suspense. This feature can be ignored on client-side navigation with the `lazy` option. In that case, you will have to manually handle loading state using the `status` value.
By default, data fetching composables will perform their asynchronous function on both client and server environments. Set the `server` option to `false` to only perform the call on the client-side. On initial load, the data will not be fetched before hydration is complete so you have to handle a pending state, though on subsequent client-side navigation the data will be awaited before loading the page.
Combined with the `lazy` option, this can be useful for data that is not needed on the first render (for example, non-SEO sensitive data).
The `useFetch` composable is meant to be invoked in setup method or called directly at the top level of a function in lifecycle hooks, otherwise you should use [`$fetch` method](#fetch).
The `pick` option helps you to minimize the payload size stored in your HTML document by only selecting the fields that you want returned from the composables.
Both `pick` and `transform` don't prevent the unwanted data from being fetched initially. But they will prevent unwanted data from being added to the payload transferred from server to client.
- [`useFetch`](/docs/api/composables/use-fetch) uses the provided URL as a key. Alternatively, a `key` value can be provided in the `options` object passed as a last argument.
- [`useAsyncData`](/docs/api/composables/use-async-data) uses its first argument as a key if it is a string. If the first argument is the handler function that performs the query, then a key that is unique to the file name and line number of the instance of`useAsyncData`will be generated for you.
The `execute` function is an alias for `refresh` that works in exactly the same way but is more semantic for cases when the fetch is [not immediate](#not-immediate).
To globally refetch or invalidate cached data, see [`clearNuxtData`](/docs/api/utils/clear-nuxt-data) and [`refreshNuxtData`](/docs/api/utils/refresh-nuxt-data).
If you want to clear the data provided, for whatever reason, without needing to know the specific key to pass to `clearNuxtData`, you can use the `clear` function provided by the composables.
To re-run your fetching function each time other reactive values in your application change, use the `watch` option. You can use it for one or multiple _watchable_ elements.
Note that **watching a reactive value won't change the URL fetched**. For example, this will keep fetching the same initial ID of the user because the URL is constructed at the moment the function is invoked.
If you need to change the URL based on a reactive value, you may want to use a [computed URL](#computed-url) instead.
#### Computed URL
Sometimes you may need to compute an URL from reactive values, and refresh the data each time these change. Instead of juggling your way around, you can attach each param as a reactive value. Nuxt will automatically use the reactive value and re-fetch each time it changes.
In the case of more complex URL construction, you may use a callback as a [computed getter](https://vuejs.org/guide/essentials/computed.html) that returns the URL string.
Every time a dependency changes, the data will be fetched using the newly constructed URL. Combine this with [not-immediate](#not-immediate), and you can wait until the reactive element changes before fetching.
The `useFetch` composable will start fetching data the moment is invoked. You may prevent this by setting `immediate: false`, for example, to wait for user interaction.
With that, you will need both the `status` to handle the fetch lifecycle, and `execute` to start the data fetch.
When we call `$fetch` in the browser, user headers like `cookie` will be directly sent to the API. But during server-side-rendering, since the `$fetch` request takes place 'internally' within the server, it doesn't include the user's browser cookies, nor does it pass on cookies from the fetch response.
The example below adds the request headers to an isomorphic `$fetch` call to ensure that the API endpoint has access to the same `cookie` header originally sent by the user.
Be very careful before proxying headers to an external API and just include headers that you need. Not all headers are safe to be bypassed and might introduce unwanted behavior. Here is a list of common headers that are NOT to be proxied:
Nuxt provides a way to perform `asyncData` fetching within the Options API. You must wrap your component definition within `defineNuxtComponent` for this to work.
When using `useAsyncData` and `useLazyAsyncData` to transfer data fetched on server to the client (as well as anything else that utilizes [the Nuxt payload](/docs/api/composables/use-nuxt-app#payload)), the payload is serialized with [`devalue`](https://github.com/Rich-Harris/devalue). This allows us to transfer not just basic JSON but also to serialize and revive/deserialize more advanced kinds of data, such as regular expressions, Dates, Map and Set, `ref`, `reactive`, `shallowRef`, `shallowReactive` and `NuxtError` - and more.
It is also possible to define your own serializer/deserializer for types that are not supported by Nuxt. You can read more in the [`useNuxtApp`](/docs/api/composables/use-nuxt-app#payload) docs.
Note that this _does not apply_ to data passed from your server routes when fetched with `$fetch` or `useFetch` - see the next section for more information.
When fetching data from the `server` directory, the response is serialized using `JSON.stringify`. However, since serialization is limited to only JavaScript primitive types, Nuxt does its best to convert the return type of `$fetch` and [`useFetch`](/docs/api/composables/use-fetch) to match the actual value.
// Type of `data` is inferred as string even though we returned a Date object
const { data } = await useFetch('/api/foo')
</script>
```
### Custom serializer function
To customize the serialization behavior, you can define a `toJSON` function on your returned object. If you define a `toJSON` method, Nuxt will respect the return type of the function and will not try to convert the types.
```ts [server/api/bar.ts]
export default defineEventHandler(() => {
const data = {
createdAt: new Date(),
toJSON() {
return {
createdAt: {
year: this.createdAt.getFullYear(),
month: this.createdAt.getMonth(),
day: this.createdAt.getDate(),
},
}
},
}
return data
})
```
```vue [app.vue]
<scriptsetuplang="ts">
// Type of `data` is inferred as
// {
// createdAt: {
// year: number
// month: number
// day: number
// }
// }
const { data } = await useFetch('/api/bar')
</script>
```
### Using an alternative serializer
Nuxt does not currently support an alternative serializer to `JSON.stringify`. However, you can return your payload as a normal string and utilize the `toJSON` method to maintain type safety.
### Consuming SSE (Server Sent Events) via POST request
::tip
If you're consuming SSE via GET request, you can use [`EventSource`](https://developer.mozilla.org/en-US/docs/Web/API/EventSource) or VueUse composable [`useEventSource`](https://vueuse.org/core/useEventSource/).
::
When consuming SSE via POST request, you need to handle the connection manually. Here's how you can do it: