docs/versioned_docs/version-6.x/realtime.md
One of the most often asked questions of RedwoodJS before and after the launch of V1 was, “When will RedwoodJS support a realtime solution?”
The answer is: now.
RedwoodJS's initial real-time solution leverages GraphQL and relies on a serverful deployment to maintain a long-running connection between the client and server.
:::note This means that your cannot currently use RedwoodJS Realtime when deployed to Netlify or Vercel.
More information about deploying a serverful RedwoodJS application is forthcoming. :::
RedwoodJS's GraphQL Server uses GraphQL over Server-Sent Events spec "distinct connections mode" for subscriptions.
Advantages of SSE over WebSockets include:
In GraphQL, there are two options for real-time updates: live queries and subscriptions.
Subscriptions are part of the GraphQL specification, whereas live queries are not.
There are times where subscriptions are well-suited for a realtime problem — and in some cases live queries may be a better fit. Later we’ll explore the pros and cons of each approach and how best to decide that to use and when.
Stream and defer are directives that allow you to improve latency for clients by sending data the most important data as soon as it's ready.
As applications grow, the GraphQL operation documents can get bigger. The server will only send the response back once all the data requested in the query is ready. However, not all requested data is of equal importance, and the client may not need all of the data at once.
The `@defer`` directive allows you to post-pone the delivery of one or more (slow) fields grouped in an inlined or spread fragment.
The '@stream' directive allows you to stream the individual items of a field of the list type as the items are available.
:::info
The @stream directive is currently not supported by Apollo GraphQL client.
:::
RedwoodJS Realtime handles the hard parts of a GraphQL Realtime implementation by automatically:
@requireAuth directives already protecting other queries and mutations (or you can implement your own validator directive)@live query directive to your GraphQL schema and setting up the useLiveQuery envelop plugin to handle requests, invalidation, and managing the storage mechanism needed@defer and @stream directivesIt provides a first-class developer experience for real-time updates with GraphQL so you can easily
and have the latest data reflected in your app.
Lastly, the Redwood CLI has commands to
Regardless of the implementation chosen, a stateful server and store are needed to track changes, invalidation, or who wants to be informed about the change.
To setup Realtime in an existing RedwoodJS project, run the following commands:
yarn rw exp setup-server-fileyarn rw exp setup-realtimeYou will get:
api/server.ts where you configure your Fastify server and GraphQLapi/lib/realtime.ts where you consume your subscriptions and configure realtime with an in-memory or Redis storeauction live query examplecountdown timer subscription examplechat subscription examplesalphabet stream exampleslow and fast field defer example:::note There is no UI setup for these examples. You can find information on how to try them out using the GraphiQL playground. :::
Now that how have a serverful project, you will configure your GraphQL server in the api/server.ts file.
:::important That means you must manually configure your GraphQL server accordingly :::
For example, you will have to setup any authentication and the realtime config:
await fastify.register(redwoodFastifyGraphQLServer, {
// If authenticating, be sure to import and add in
// authDecoder,
// getCurrentUser,
loggerConfig: {
logger: logger,
options: {
query: true,
data: true,
operationName: true,
requestId: true,
},
},
graphiQLEndpoint: enableWeb ? '/.redwood/functions/graphql' : '/graphql',
sdls,
services,
directives,
allowIntrospection: true,
allowGraphiQL: true,
// Configure if using RedwoodJS Realtime
realtime,
})
You can now remove the GraphQL handler function that resides in api/functions/graphql.ts.
By default, RedwoodJS realtime configures an in-memory store for the Pub Sub client used with subscriptions and live query invalidation.
Realtime supports in-memory and Redis stores:
To enable defer and streaming, set enableDeferStream to true.
Configure a Redis store and defer and stream in:
// api/lib/realtime.ts
import { RedwoodRealtimeOptions } from '@redwoodjs/realtime'
import subscriptions from 'src/subscriptions/**/*.{js,ts}'
// if using a Redis store
// import { Redis } from 'ioredis'
// const publishClient = new Redis()
// const subscribeClient = new Redis()
/**
* Configure RedwoodJS Realtime
*
* See https://redwoodjs.com/docs/realtime
*
* Realtime supports Live Queries and Subscriptions over GraphQL SSE.
*
* Live Queries are GraphQL queries that are automatically re-run when the data they depend on changes.
*
* Subscriptions are GraphQL queries that are run when a client subscribes to a channel.
*
* Redwood Realtime
* - uses a publish/subscribe model to broadcast data to clients.
* - uses a store to persist Live Query and Subscription data.
*
* Redwood Realtime supports in-memory and Redis stores:
* - In-memory stores are useful for development and testing.
* - Redis stores are useful for production.
*/
export const realtime: RedwoodRealtimeOptions = {
subscriptions: {
subscriptions,
store: 'in-memory',
// if using a Redis store
// store: { redis: { publishClient, subscribeClient } },
},
liveQueries: {
store: 'in-memory',
// if using a Redis store
// store: { redis: { publishClient, subscribeClient } },
},
// To enable defer and streaming, set to true.
// enableDeferStream: true,
}
By setting up RedwoodJS Realtime, the GraphQL server adds two helpers on the context:
With context.pubSub you can subscribe to and publish messages via context.pubSub.publish('the-topic', id, id2).
With context.liveQueryStore. you can context.liveQueryStore.invalidate(key) where your key may be a reference or schema coordinate:
Where the query is: auction(id: ID!): Auction @requireAuth:
"Auction:123"When the query is: auctions: [Auction!]! @requireAuth:
"Query.auctions"RedwoodJS has a first-class developer experience for GraphQL subscriptions.
type Subscription {
newMessage(roomId: ID!): Message! @requireAuth
}
Counts down from a starting values by an interval.
subscription CountdownFromInterval {
countdown(from: 100, interval: 10)
}
This example showcases how a subscription can yields its own response.
RedwoodJS has made it super easy to add live queries to your GraphQL server! You can push new data to your clients automatically once the data selected by a GraphQL operation becomes stale by annotating your query operation with the @live directive.
The invalidation mechanism is based on GraphQL ID fields and schema coordinates. Once a query operation has been invalidated, the query is re-executed, and the result is pushed to the client.
query GetCurrentAuctionBids @live {
auction(id: "1") {
bids {
amount
}
highestBid {
amount
}
id
title
}
}
mutation MakeBid {
bid(input: { auctionId: "1", amount: 10 }) {
amount
}
}
The @defer directive allows you to post-pone the delivery of one or more (slow) fields grouped in an inlined or spread fragment.
Here, the GraphQL schema defines two queries for a "fast" and a "slow" (ie, delayed) information.
export const schema = gql`
type Query {
"""
A field that resolves fast.
"""
fastField: String! @skipAuth
"""
A field that resolves slowly.
Maybe you want to @defer this field ;)
"""
slowField(waitFor: Int! = 5000): String @skipAuth
}
`
The Redwood services for these queries return the fastField immediately and the showField after some delay.
import { logger } from 'src/lib/logger'
const wait = (time: number) =>
new Promise((resolve) => setTimeout(resolve, time))
export const fastField = async () => {
return 'I am speedy'
}
export const slowField = async (_, { waitFor = 5000 }) => {
logger.debug('deferring slowField until ...')
await wait(waitFor)
logger.debug('now!')
return 'I am slow'
}
When making the query:
query SlowAndFastFieldWithDefer {
... on Query @defer {
slowField
}
fastField
}
The response returns:
{
"data": {
"fastField": "I am speedy"
}
}
and will await the deferred field to then present:
{
"data": {
"fastField": "I am speedy",
"slowField": "I am slow"
}
}
The @stream directive allows you to stream the individual items of a field of the list type as the items are available.
Here, the GraphQL schema defines a query to return the letters of the alphabet:
export const schema = gql`
type Query {
alphabet: [String!]! @skipAuth
`
The service uses Repeater to write a safe stream resolver.
:::info
AsyncGenerators as declared via the async * keywords are prone to memory leaks and leaking timers. For real-world usage, use Repeater.
:::
import { Repeater } from '@redwoodjs/realtime'
import { logger } from 'src/lib/logger'
export const alphabet = async () => {
return new Repeater<string>(async (push, stop) => {
const values = ['a', 'b', 'c', 'd', 'e', 'f', 'g']
const publish = () => {
const value = values.shift()
if (value) {
logger.debug({ value }, 'publishing')
push(value)
}
if (values.length === 0) {
stop()
}
}
const interval = setInterval(publish, 1000)
stop.then(() => {
logger.debug('cancel')
clearInterval(interval)
})
publish()
})
}
Since Apollo Client does not yet support the @stream directive, you can use them in the GraphiQL Playground or see them in action via CURL.
When making the request with the @stream directive:
curl -g -X POST \
-H "accept:multipart/mixed" \
-H "content-type: application/json" \
-d '{"query":"query StreamAlphabet { alphabet @stream }"}' \
http://localhost:8911/graphql
Here you see the initial response has [] for alphabet data.
Then on each push to the Repeater, an incremental update the the list of letters is sent.
The stream ends when hasNext is false:
* Connected to localhost (127.0.0.1) port 8911 (#0)
> POST /graphql HTTP/1.1
> Host: localhost:8911
> User-Agent: curl/8.1.2
> accept:multipart/mixed
> content-type: application/json
> Content-Length: 53
>
< HTTP/1.1 200 OK
< connection: keep-alive
< content-type: multipart/mixed; boundary="-"
< transfer-encoding: chunked
<
---
Content-Type: application/json; charset=utf-8
Content-Length: 39
{"data":{"alphabet":[]},"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["a"],"path":["alphabet",0]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["b"],"path":["alphabet",1]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["c"],"path":["alphabet",2]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["d"],"path":["alphabet",3]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["e"],"path":["alphabet",4]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["f"],"path":["alphabet",5]}],"hasNext":true}
---
Content-Type: application/json; charset=utf-8
Content-Length: 70
{"incremental":[{"items":["g"],"path":["alphabet",6]}],"hasNext":true}
---
...
---
Content-Type: application/json; charset=utf-8
Content-Length: 17
{"hasNext":false}
-----
When deciding on how to offer realtime data updates in your RedwoodJS app, you’ll want to consider:
Please see our showcase RedwoodJS Realtime app for exampes of subscriptions and live queries. It also demonstrates how you can handle streaming responses, like those used by OpenAI chat completions.
Sends a message to one of four Chat Rooms.
Each room subscribes to its new messages via the NewMessage channel aka topic.
context.pubSub.publish('newMessage', roomId, { from, body })
./scripts/simulate_chat.sh -h
Usage: ./scripts/simulate_chat.sh -r [roomId] -n [num_messages]
./scripts/simulate_chat.sh -h
Options:
-r roomId Specify the room ID (1-4) for sending chat messages.
-n num_messages Specify the number of chat messages to send. If not provided, the script will run with a random number of messages.
/**
* To test this NewMessage subscription, run the following in one GraphQL Playground to subscribe:
*
* subscription ListenForNewMessagesInRoom {
* newMessage(roomId: "1") {
* body
* from
* }
* }
*
*
* And run the following in another GraphQL Playground to publish and send a message to the room:
*
* mutation SendMessageToRoom {
* sendMessage(input: {roomId: "1", from: "hello", body: "bob"}) {
* body
* from
* }
* }
*/
Bid on a fancy pair of new sneaks!
When a bid is made, the auction updates via a Live Query due to the invalidation of the auction key.
const key = `Auction:${auctionId}`
context.liveQueryStore.invalidate(key)
./scripts/simulate_bids.sh -h
Usage: ./scripts/simulate_bids.sh [options]
Options:
-a <auctionId> Specify the auction ID (1-5) for which to send bids (optional).
-n <num_bids> Specify the number of bids to send (optional).
-h, --help Display this help message.
/**
* To test this live query, run the following in the GraphQL Playground:
*
* query GetCurrentAuctionBids @live {
* auction(id: "1") {
* bids {
* amount
* }
* highestBid {
* amount
* }
* id
* title
* }
* }
*
* And then make a bid with the following mutation:
*
* mutation MakeBid {
* bid(input: {auctionId: "1", amount: 10}) {
* amount
* }
* }
*/
It started slowly and I thought it was my heart But then I realised that this time it was for real
Counts down from a starting values by an interval.
This example showcases how a subscription can yields its own response.
/**
* To test this Countdown subscription, run the following in the GraphQL Playground:
*
* subscription CountdownFromInterval {
* countdown(from: 100, interval: 10)
* }
*/
Tell me a story about a happy, purple penguin that goes to a concert.
Showcases how to use OpenAI to stream a chat completion via a prompt that writes a bedtime story:
const PROMPT = `Write a short children's bedtime story about an Animal that is a given Color and that does a given Activity.
Give the animal a cute descriptive and memorable name.
The story should teach a lesson.
The story should be told in a quality, style and feeling of the given Adjective.
The story should be no longer than 3 paragraphs.
Format the story using Markdown.`
The story updates on each stream content delta via a newStory subscription topic event.
context.pubSub.publish('newStory', id, story)
It's Out of Africa meets Pretty Woman.
So it's a psychic, political, thriller comedy with a heart With a heart, not unlike Ghost meets Manchurian Candidate.
-- The Player, 1992
Mashup some of your favorite movies to create something new and Netflix-worthy to watch.
Powered by OpenAI, this movie tagline and treatment updates on each stream content delta via a Live Query bui invalidating the `MovieMashup key.
context.liveQueryStore.invalidate(`MovieMashup:${id}`)