@sanity/preview-kit
Sanity.io toolkit for building live-as-you-type content preview experiences and visual editing.
Installation
npm i @sanity/preview-kit @sanity/client
pnpm i @sanity/preview-kit @sanity/client
yarn add @sanity/preview-kit @sanity/client
@sanity/preview-kit/client
Visual Editing with Content Source Maps
Note
Content Source Maps are available as an API for select Sanity enterprise customers. Contact our sales team for more information.
You can use Visual Editing with any framework, not just React. Read our guide for how to get started.
Enhanced Sanity client with createClient
Preview Kit's enhanced Sanity client is built on top of @sanity/client
and is designed to be a drop-in replacement. It extends the client configuration with options for returning encoded metadata from Content Source Maps.
import { createClient, type ClientConfig } from '@sanity/preview-kit/client'
const config: ClientConfig = {
studioUrl: '/studio',
encodeSourceMap: 'auto',
}
const client = createClient(config)
studioUrl
Required when encodeSourceMap
is active, and can either be an absolute URL:
import { createClient } from '@sanity/preview-kit/client'
const client = createClient({
...config,
studioUrl: 'https://your-company.com/studio',
})
Or a relative path if the Studio is hosted on the same deployment, or embedded in the same app:
import { createClient } from '@sanity/preview-kit/client'
const client = createClient({
...config,
studioUrl: '/studio',
})
encodeSourceMap
Accepts "auto"
, the default, or a boolean
. Controls when to encode the content source map into strings using @vercel/stega
encoding. When "auto"
is used a best-effort environment detection is used to see if the environment is a Vercel Preview deployment. On a different hosting provider, or in local development, configure this option to make sure it is only enabled in non-production deployments.
import { createClient } from '@sanity/preview-kit/client'
const client = createClient({
...config,
encodeSourceMap: process.env.VERCEL_ENV === 'preview',
})
encodeSourceMapAtPath
By default source maps are encoded into all strings that can be traced back to a document field, except for URLs and ISO dates. We also make some exceptions for fields like, document._type
, document._id
and document.slug.current
, that we've seen leading to breakage if the string is altered as well as for Portable Text.
You can customize this behavior using encodeSourceMapAtPath
:
import { createClient } from '@sanity/preview-kit/client'
const client = createClient({
...config,
encodeSourceMapAtPath: (props) => {
if (props.path[0] === 'externalUrl') {
return false
}
return props.filterDefault(props)
},
})
logger
Pass a console
into logger
to get detailed debug info and reports on which fields are encoded and which are skipped:
import { createClient } from '@sanity/preview-kit/client'
const client = createClient({
...config,
logger: console,
})
An example report:
[@sanity/preview-kit]: Creating source map enabled client
[@sanity/preview-kit]: Stega encoding source map into result
[@sanity/preview-kit]: Paths encoded: 3, skipped: 17
[@sanity/preview-kit]: Table of encoded paths
┌─────────┬──────────────────────────────┬───────────────────────────┬────────┐
│ (index) │ path │ value │ length │
├─────────┼──────────────────────────────┼───────────────────────────┼────────┤
│ 0 │ "footer[0].children[0].text" │ '"The future is alrea...' │ 67 │
│ 1 │ "footer[1].children[0].text" │ 'Robin Williams' │ 14 │
│ 2 │ "title" │ 'Visual Editing' │ 14 │
└─────────┴──────────────────────────────┴───────────────────────────┴────────┘
[@sanity/preview-kit]: List of skipped paths [
'footer[]._key',
'footer[].children[]._key',
'footer[].children[]._type',
'footer[]._type',
'footer[].style',
'_type',
'slug.current',
]
resultSourceMap
This option is always enabled if encodeSourceMap
. It's exposed here to be compatible with @sanity/client
and custom use cases where you want content source maps, but not the encoding.
const client = createClient({
...config,
resultSourceMap: true,
})
const { result, resultSourceMap } = await client.fetch(query, params, {
filterResponse: false,
})
console.log(resultSourceMap)
Using the Content Source Map with custom logic
If you're building your own custom preview logic you can use mapToEditLinks
to skip encoding hidden metadata into strings, and access the edit links directly:
import { createClient, mapToEditLinks } from '@sanity/preview-kit/client'
const client = createClient({
...config,
resultSourceMap: true,
encodeSourceMap: false,
})
const { result, resultSourceMap } = await client.fetch(
query,
params,
{ filterResponse: false }
)
const studioUrl = 'https://your-company.com/studio'
const editLinks = mapToEditLinks({ result, resultSourceMap }, studioUrl)
const title = result.title
const titleEditLink = editLinks.title
console.log(title, titleEditLink)
Using Perspectives
The perspective
option can be used to specify special filtering behavior for queries. The default value is raw
, which means no special filtering is applied, while published
and previewDrafts
can be used to optimize for specific use cases. Read more about this option:
@sanity/preview-kit
Note
This is the new docs for @sanity/preview-kit
v2. If you're looking for docs for v1 APIs, like definePreview
and usePreview
, they're available on the v1 branch..
There's a full migration guide available here.
If you're looking for React Server Component and Next.js docs, they're in the next-sanity
readme.
Live real-time preview for React
Note
The examples in this README use Remix, you can find Next.js specific examples in the next-sanity
README. Including information on how to build live previews in React Server Components with the new app-router.
Write GROQ queries like @sanity/client and have them resolve in-memory, locally. Updates from Content Lake are streamed in real-time with sub-second latency.
Requires React 18, support for other libraries like Solid, Svelte, Vue etc are planned. For now you can use @sanity/groq-store directly.
Get started in 3 steps:
- Create a
getClient
utility that returns a @sanity/client
instance. - Define a
<LiveQueryProvider />
configuration. - Refactor the root layout of your app to conditionally wrap it in
<LiveQueryProvider />
when it's asked to preview drafts. - Use the
useLiveQuery
hook in components that you want to re-render in real-time as your documents are edited.
1. Create a getClient
utility
As <LiveQueryProvider />
is configured with a @sanity/client
instance it makes sense to create a utility for it. Doing so makes it easy to ensure the server-side and client-side client are configured the same way.
app/lib/sanity.ts
import { createClient } from '@sanity/client'
import type { SanityClient } from '@sanity/client'
export function getClient({
preview,
}: {
preview?: { token: string }
}): SanityClient {
const client = createClient({
projectId: 'your-project-id',
dataset: 'production',
apiVersion: '2023-06-20',
useCdn: true,
perspective: 'published',
})
if (preview) {
if (!preview.token) {
throw new Error('You must provide a token to preview drafts')
}
return client.withConfig({
token: preview.token,
useCdn: false,
ignoreBrowserTokenWarning: true,
perspective: 'previewDrafts',
})
}
return client
}
2. Define a <LiveQueryProvider />
component
Create a new file for the provider, so it can be loaded with React.lazy
and avoid increasing the bundle size in production. Ensuring code needed for live previewing drafts are only loaded when needed.
app/PreviewProvider.tsx
import { LiveQueryProvider } from '@sanity/preview-kit'
import { useMemo } from 'react'
import { getClient } from '~/lib/sanity'
export default function PreviewProvider({
children,
token,
}: {
children: React.ReactNode
token: string
}) {
const client = useMemo(() => getClient({ preview: { token } }), [token])
return <LiveQueryProvider client={client}>{children}</LiveQueryProvider>
}
Only the client
prop is required. For debugging you can pass a logger={console}
prop.
3. Making a Remix route conditionally preview draft
Here's the Remix route we'll be adding live preview of drafts, it's pretty basic:
import type { LoaderArgs } from '@vercel/remix'
import { useLoaderData } from '@remix-run/react'
import { getClient } from '~/lib/sanity'
import type { UsersResponse } from '~/UsersList'
import { UsersList, usersQuery } from '~/UsersList'
import { Layout } from '~/ui'
export async function loader({ request }: LoaderArgs) {
const client = getClient({})
const url = new URL(request.url)
const lastId = url.searchParams.get('lastId') || ''
const users = await client.fetch<UsersResponse>(usersQuery, { lastId })
return { users, lastId }
}
export default function Index() {
const { users, lastId } = useLoaderData<typeof loader>()
return (
<Layout>
<UsersList data={users} lastId={lastId} />
</Layout>
)
}
Now let's import the PreviewProvider
component we created in the previous step. To ensure we don't increase the production bundle size, we'll use React.lazy
to code-split the component. The React.lazy
API requires a React.Suspense
boundary, so we'll add that too.
import { lazy, Suspense } from 'react'
const PreviewProvider = lazy(() => import('~/PreviewProvider'))
Before we can add <PreviewProvider />
to the layout we need to update the loader
to include the props it needs. We'll use an environment variable called SANITY_API_PREVIEW_DRAFTS
to control when to live preview drafts, and store a viewer
API token in SANITY_API_READ_TOKEN
.
Update the const client = getClient({})
call to:
const token = process.env.SANITY_API_READ_TOKEN
const preview =
process.env.SANITY_API_PREVIEW_DRAFTS === 'true' ? { token } : undefined
const client = getClient({ preview })
Update the loader
return statement from return {users, lastId}
to:
return { preview, users, lastId }
And add preview
to useLoaderData
:
const { preview, users, lastId } = useLoaderData<typeof loader>()
Then make the render conditional based on wether preview
is set:
const children = <UsersList data={users} lastId={lastId} />
return (
<Layout>
{preview ? (
<Suspense fallback={children}>
<PreviewProvider token={preview.token}>{children}</PreviewProvider>
</Suspense>
) : (
children
)}
</Layout>
)
After putting everything together the route should now look like this:
import type { LoaderArgs } from '@vercel/remix'
import { useLoaderData } from '@remix-run/react'
import { lazy, Suspense } from 'react'
import { getClient } from '~/lib/sanity'
import type { UsersResponse } from '~/UsersList'
import { UsersList, usersQuery } from '~/UsersList'
import { Layout } from '~/ui'
const PreviewProvider = lazy(() => import('~/PreviewProvider'))
export async function loader({ request }: LoaderArgs) {
const token = process.env.SANITY_API_READ_TOKEN
const preview =
process.env.SANITY_API_PREVIEW_DRAFTS === 'true' ? { token } : undefined
const client = getClient({ preview })
const url = new URL(request.url)
const lastId = url.searchParams.get('lastId') || ''
const users = await client.fetch<UsersResponse>(usersQuery, { lastId })
return { preview, users, lastId }
}
export default function Index() {
const { preview, users, lastId } = useLoaderData<typeof loader>()
const children = <UsersList data={users} lastId={lastId} />
return (
<Layout>
{preview ? (
<Suspense fallback={children}>
<PreviewProvider token={preview.token!}>{children}</PreviewProvider>
</Suspense>
) : (
children
)}
</Layout>
)
}
4. Adding the useLiveQuery
hook to components that need to re-render in real-time
Let's look at what the <UsersList>
component looks like, before we add the hook:
import groq from 'groq'
import { ListView, ListPagination } from '~/ui'
export const usersQuery = groq`{
"list": *[_type == "user" && _id > $lastId] | order(_id) [0...20],
"total": count(*[_type == "user"]),
}`
export interface UsersResponse {
list: User[]
total: number
}
export interface UsersListProps {
data: UsersResponse
lastId: string
}
export function UsersList(props: UsersListProps) {
const { data, lastId } = props
return (
<>
<ListView list={data.list} />
<ListPagination total={data.total} lastId={lastId} />
</>
)
}
To make this component connect to your preview provider you need to add the useLiveQuery
. You don't have to refactor your components so that the hook is only called when there's a parent <LiveQueryProvider />
, it's safe to call it unconditionally.
If there's no <LiveQueryProvider />
it behaves as if the hook had this implementation:
export function useLiveQuery(initialData) {
return [initialData, false]
}
Thus it's fairly easy to add conditional live preview capabilities to UsersList
, simply add hook to your imports:
import { useLiveQuery } from '@sanity/preview-kit'
And replace this:
const { data, lastId } = props
With this:
const { data: initialData, lastId } = props
const [data] = useLiveQuery(initialData, usersQuery, { lastId })
All together now:
import { useLiveQuery } from '@sanity/preview-kit'
import groq from 'groq'
import { ListView, ListPagination } from '~/ui'
export const usersQuery = groq`{
"list": *[_type == "user" && _id > $lastId] | order(_id) [0...20],
"total": count(*[_type == "user"]),
}`
export interface UsersResponse {
list: User[]
total: number
}
export interface UsersListProps {
data: UsersResponse
lastId: string
}
export function UsersList(props: UsersListProps) {
const { data: initialData, lastId } = props
const [data] = useLiveQuery(initialData, usersQuery, { lastId })
return (
<>
<ListView list={data.list} />
<ListPagination total={data.total} lastId={lastId} />
</>
)
}
And done! You can optionally optimize it further by adding a loading UI while it loads, or improve performance by adding a custom isEqual
function to reduce React re-renders if there's a lot of data that changes but isn't user visible (SEO metadata and such).
Implementing a Loading UI with useLiveQuery
The best way to do this is to add a wrapper component that is only used in preview mode that calls the useLiveQuery
hook.
export function UsersList(props: UsersListProps) {
const { data, lastId } = props
return (
<>
<ListView list={data.list} />
<ListPagination total={data.total} lastId={lastId} />
</>
)
}
export function PreviewUsersList(props: UsersListProps) {
const { data: initialData, lastId } = props
const [data, loading] = useLiveQuery(initialData, usersQuery, { lastId })
return (
<>
<PreviewStatus loading={loading} />
<UsersList data={users} lastId={lastId} />
</>
)
}
Change the layout from:
const children = <UsersList data={users} lastId={lastId} />
return (
<Layout>
{preview ? (
<Suspense fallback={children}>
<PreviewProvider token={preview.token!}>{children}</PreviewProvider>
</Suspense>
) : (
children
)}
</Layout>
)
To this:
return (
<Layout>
{preview ? (
<Suspense fallback={children}>
<PreviewProvider token={preview.token!}>
<PreviewUsersList data={users} lastId={lastId} />
</PreviewProvider>
</Suspense>
) : (
<UsersList data={users} lastId={lastId} />
)}
</Layout>
)
Optimizing performance
Out of the box it'll only trigger a re-render of UsersList
if the query response changed, using react-fast-compare
under the hood. You can tweak this behavior by passing a custom isEqual
function as the third argument to useLiveQuery
if there's only some changes you want to trigger a re-render.
const [data] = useLiveQuery(
initialData,
usersQuery,
{ lastId },
{
isEqual: (a, b) =>
a.list.every((aItem, index) => {
const bItem = b.list[index]
return aItem._id === bItem._id && aItem.name === bItem.name
}),
}
)
You can also use the React.useDeferredValue
hook and a React.memo
wrapper to further optimize performance by letting React give other state updates higher priority than the preview updates. It prevents the rest of your app from slowing down should there be too much Studio activity for the previews to keep up with:
import { memo, useDeferredValue } from 'react'
export function PreviewUsersList(props: UsersListProps) {
const { data: initialData, lastId } = props
const [snapshot] = useLiveQuery(initialData, usersQuery, { lastId })
const data = useDeferredValue(snapshot)
return <UsersList data={data} lastId={lastId} />
}
export const UsersList = memo(function UsersList(props: UsersListProps) {
const { data, lastId } = props
return (
<>
<ListView list={data.list} />
<ListPagination total={data.total} lastId={lastId} />
</>
)
})
Advanced usage
Fine-tuning cache
The defaults set for the cache
prop are optimized for most use cases, but are conservative since the size of your documents can vary a lot. And your queries might only use some document types and it's not necessary to cache every type.
Thus you can fine-tune the cache by passing a custom cache
prop to LiveQueryProvider
:
import { LiveQueryProvider } from '@sanity/preview-kit'
return (
<LiveQueryProvider
client={client}
cache={{
// default: 3000, increased to 10000 for this example app as each document is small
maxDocuments: 10000,
// The default cache includes all document types, you can reduce the amount of documents
// by only including the ones you need.
// You can run the `array::unique(*._type)` GROQ query in `Vision` in your Studio to see how many types are in your dataset.
// Just be careful that you don't forget the document types you might be using in strong references, such as `project` or `sanity.imageAsset`
includeTypes: ['page', 'product', 'sanity.imageAsset'],
// Turn off using a mutation EventSource listener, this means content updates will require a manual refresh
listen: false,
}}
// If the cache is full it'll fallback to a polling interval mode, that refreshes every 10 seconds by default.
// You can opt-in to having an error thrown instead by setting this to `0`.
refreshInterval={10000}
// Passing a logger gives you more information on what to expect based on your configuration
logger={console}
>
{children}
</LiveQueryProvider>
)
Content Source Map features
When the client
instance is configured to client.config().resultSourceMap == true
the LiveQueryProvider
will opt-in to a faster and smarter cache than the default mode.
It'll only listen for mutations on the documents that you are using in your queries, and apply the mutations to the cache in real-time.
This mode is best-effort, and if you're relying on features such as upper
you may want to disable this mode.
import { LiveQueryProvider } from '@sanity/preview-kit'
return (
<LiveQueryProvider
client={client}
turboSourceMap={false}
// Passing a logger gives you more information on what to expect based on your configuration
logger={console}
>
{children}
</LiveQueryProvider>
)
For data that can't be traced with Content Source Maps there's a background refresh interval. Depending on your queries you might want to tweak this interval to get the best performance.
import { LiveQueryProvider } from '@sanity/preview-kit'
return (
<LiveQueryProvider
client={client}
// Refetch all queries every minute instead of the default 10 seconds
refreshInterval={1000 * 60}
// Passing a logger gives you more information on what to expect based on your configuration
logger={console}
>
{children}
</LiveQueryProvider>
)
Release new version
Run "CI & Release" workflow.
Make sure to select the main branch and check "Release new version".
Semantic release will only release on configured branches, so it is safe to run release on any branch.
License
MIT-licensed. See LICENSE.