The overall API architecture is made up of two GraphQL servers, one exposing subscription/mutation functionality and the other exposing query functionality. Both GraphQL servers use the Spotify REST API as their datasource, but we are hosting the subscription server on dedicated infrastructure (Railway) and the other on serverless functions (Netlify).
The Apollo Router routes incoming traffic from the client application and integrates with GraphOS to receive schema updates and report usage metrics.
graph LR;
subgraph "Netlify"
web["Website\n(client app)"]
end
subgraph "Railway"
router{"Apollo Router"}
playbackSubgraph["Playback Subgraph\n(subriptions/mutations)"]
spotifySubgraph["Spotify Subgraph\n(queries)"]
end
subgraph "Apollo"
schema["Schema Pipeline"]
usage["Usage Reporting"]
end
subgraph "Spotify"
spotifyREST[Spotify REST API]
end
web <--> router
router <-->|Schema Updates\nUsage Reporting| Apollo
router <--> spotifySubgraph
router <--> playbackSubgraph
playbackSubgraph <--> spotifyREST
spotifySubgraph <--> spotifyREST
classDef spotifyBox color:#FFFFFF,fill:#1DB954,stroke:#FFFFFF,stroke-width:2px;
classDef netlifyBox color:#014847,fill:#FFFFFF,stroke:#32e6e2,stroke-width:2px;
classDef railwayBox color:#000000,fill:#FFFFFF,stroke:#000000,stroke-width:2px;
classDef apolloBox color:#3f20ba,fill:#FFFFFF,stroke:#3f20ba,stroke-width:2px;
class Spotify spotifyBox
class Netlify netlifyBox
class Netlify-Function netlifyBox
class Railway railwayBox
class Apollo apolloBox
*Note: We are using only the Spotify REST API as our datasource for demonstration purposes. The subscriptions subgraph implements a polling mechanism that we host on a dedicated infrastructure while the "query" subgraph is hosted on serverless infrastructure*
What do you want to do next with this demo app?
- Visit the public Apollo Explorer instance to interact with the graph (No GraphOS account required)
- Query the graph (Spotify account required) - OAuth workflow with be initiated from Apollo Explorer to login to our Spotify account to run any operation
- Try having your Spotify app playing on your phone or desktop and then run this mutation
- View the graph's schema
Note: To create graphs, you must have the Org Administrator or Graph Administrator role for your GraphOS organization.
-
Clone this repo.
-
Create a personal API key.
-
Run the script below using your personal API key.
AUTH={YOUR_API_KEY} npm run graphos-demo
If you have multiple organizations, the script will prompt you to select which organization the graph should be in.
-
Clone this repo.
-
From the root of the folder, install the dependencies.
npm install
-
From the root of the folder, start the client app.
npm start
-
Visit the website at http://localhost:3000.
By default, the client app is pointing to the locally-running backend URL. Follow the section below for "I want to run the backend locally" (you can choose to use Docker, or not).
Local subscriptions using rover dev
is currently a work-in-progress, so features that use subscriptions are not functional (such as listening to playback state). You can change the URL the client application is pointing at by editing the .env.development
file with VITE_SERVER_HOST
and setting it to the production URL (uncomment the second line).
You will need a GraphOS organization with an Enterprise plan or Enterprise trial plan to utilize all the GraphOS features.
-
Follow the steps in the "I want to re-create this demo in my GraphOS account" section above.
-
Find the newly-created graph in GraphOS Studio. Create a Graph API key from the Settings page, with the default "Graph Admin" permissions.
-
Create a
.env
file at the root of this repository and add in the following variables:APOLLO_GRAPH_REF={YOUR_DEMO_GRAPH_ID}@main APOLLO_KEY={YOUR_GRAPH_API_KEY} CALLBACK_URL=http://router:4000
-
Start the router and subgraphs:
npm run docker:run
- Supergraph - http://localhost:4000
- Spotify Subgraph - http://spotify:4001 (inaccessible in cluster)
- Playback Subgraph - http://playback:4002 (inaccessible in cluster)
-
If you want to run any queries through Apollo Explorer, you'll need to disable persisted queries. You can disable this in
router/router.yaml
by commenting out thepersisted_queries
configuration.# persisted_queries: # enabled: true # safelist: # enabled: true # require_id: false
Re-run the
docker:run
command.
You will need a GraphOS organization with an Enterprise plan or Enterprise trial plan to utilize all the GraphOS features.
-
Follow the steps in the "I want to re-create this demo in my GraphOS account" section above.
-
Find the newly-created graph in GraphOS Studio. Create a Graph API key from the Settings page, with the default "Graph Admin" permissions.
-
Create a
.env
file at the root of this repository and add in the following variables:APOLLO_GRAPH_REF={YOUR_DEMO_GRAPH_ID}@main APOLLO_KEY={YOUR_GRAPH_API_KEY} CALLBACK_URL=http://router:4000
-
Download the latest router binary to the
router
folder. -
Build and start the subgraphs. In the root of the project, run:
npm run build
npm run start:spotify
and in another terminal:
npm run start:playback
- Spotify Subgraph - http://localhost:4001
- Playback Subgraph - http://localhost:4002
Note: If you make any edits to the subgraph files, you'll need to re-build and restart the subgraphs manually. If you're looking to do local development, you may want to navigate to each subgraph folder and run
npm run dev
instead, which will watch for changes and restart the subgraph automatically. -
In a new terminal, start the router:
APOLLO_KEY={YOUR_DEMO_GRAPH_ID}@main APOLLO_GRAPH_REF={YOUR_GRAPH_API_KEY} CALLBACK_URL=http://127.0.0.1:4000 npm run start:router
- Supergraph - http://localhost:4000
-
If you want to run any queries through Apollo Explorer, you'll need to disable persisted queries. You can disable this in
router/router.yaml
by commenting out thepersisted_queries
configuration.# persisted_queries: # enabled: true # safelist: # enabled: true # require_id: false
Restart the router to run it with the latest config changes.
There are launch configurations for the client project and subgraph projects. You can navigate to the debug tab of VS Code and launch any of the projects. They will default to the following urls:
- Client App - http://localhost:3000
- Spotify Subgraph - http://localhost:4001
- Playback Subgraph - http://localhost:4002
playback - This subgraph has been designed to handle the Subscription/Mutation operations for our graph. The subgraph is hosted on a dedicated piece of infrastructure (Railway) because it needs to be long lived with subscriptions support.
spotify - This subgraph handles all of the Query operations for our graph and is hosted on serverless infrastructure (Netlify/AWS Lambda). Hosting in serverless is more cost effective for this single service in our overall architecture.
If you used the React + Apollo Spotify Showcase and have two minutes then we'd really appreciate it if you filled out this survey - it really helps us improve!
If you're exploring the codebase and not sure where to get started, try the following:
Client
client/src/router.tsx
- This defines all routes used in the app. To view the source code for a given route, follow the import for the route component.client/src/index.tsx
- This is the entry point to the client app. This defines the providers used in the app.
Server
server.ts
- This defines the Apollo GraphQL server used to serve the Spotify GraphQL schema.server/src/resolvers/
- This defines the resolvers used to resolve types and fields in the schema. The file names correspond to their respective types in the schema.server/src/dataSources/spotify.ts
- Defines the Spotify client used to make REST calls to Spotify's REST API.
Apollo Client's newest features unlock powerful UI development capabilities when
used with React 18. Using Suspense via useSuspenseQuery
is one such
capability, as is useBackgroundQuery
. Both of these will be shipped in 3.8.0.
These hooks, along with the already-available useFragment
hook and the GraphQL
@defer
directive, comprise a toolkit for the UI developer to implement the
render-as-you-fetch pattern throughout the component tree.
@jerelmiller started building this application
while building useSuspenseQuery
in an effort to dogfood the changes with a
non-trivial app. Apollo Client and GraphQL are built to be both approachable and
scalable; to-do apps are the former but not the latter. A Spotify clone - and it
really is a clone (๐ that CSS) - offered a much more robust proving ground for
the functionality we were building. As the team used it more and more, we
decided that if we open-sourced it then the community could use it to try things
out for themselves.
So, here you go! It's our hope that you are able to use this app to do any or all of these things:
- Listen to music ๐ง
- Learn how to use React Suspense
- See how the features in Apollo Client 3.7 and 3.8 work
- Try the GraphQL
@defer
directive - Experiment with GraphOS by turning a monograph into a supergraph
- Get some concrete code samples to inspire your own applications
- Use as a template for reporting bugs
Name | Username |
---|---|
Ben Newman | @benjamn |
Alessia Bellisario | @alessbell |
Jeff Auriemma | @bignimbus |
Hugh Willson | @hwillson |
Jerel Miller | @jerelmiller |
Lenz Weber-Tronic | @phryneas |
This app implements a GraphQL API on top of Spotify's REST API. The GraphQL server aims to mirror the REST API as much as possible, including the field names and returned values. While it's tempting to patch the REST API in areas that make it difficult to consume (such as a separate endpoint to check if a track is in the user's library), this presented a good opportunity to showcase how a developer can use Apollo Client's capabilities to effectively build an app with these kinds of shortcomings.
There are, however, a few cases where the GraphQL API differs from the REST API.
-
This Spotify GraphQL API returns full object types in some areas where Spotify returns "simplified" object types. For example, fetching a track via the
/tracks/:trackId
endpoint gives you the full track data, but fetching tracks through the/albums/:albumId
endpoint gives you a simplified track type. In these cases, the GraphQL API consolidates these distinct types into a the full object type (i.e.Track
.) -
Paginated fields use a Relay-style connection type. This allows the GraphQL API to express edge-specific data and pagination information in a natural way.
-
Endpoints that accept a
market
parameter are omitted in the equivalent GraphQL field. This is because, according to the documentation:If a valid user access token is specified in the request header, the country associated with the user account will take priority over this parameter.
The GraphQL server only works with authenticated users via an access token, so maintaining this parameter was unnecessary overhead.
-
The GraphQL serves fields using camelCase. The Spotify REST API returns fields using snake_case. While not strictly enforced in the spec, GraphQL fields are commonly written in camelCase form.