Examples

A collection of example apps for reference.

Below is a collection of example SST apps. These are available in the examples/ directory of the repo.

The descriptions for these examples are generated using the comments in the sst.config.ts of the app.

Contributing

To contribute an example or to edit one, submit a PR to the repo. Make sure to document the sst.config.ts in your example.

AWS Astro container with Redis

Creates a hit counter app with Astro and Redis.

This deploys Astro as a Fargate service to ECS and it’s linked to Redis.

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

Since our Redis cluster is in a VPC, we’ll need a tunnel to connect to it from our local machine.

sudo npx sst tunnel install

This needs sudo to create a network interface on your machine. You’ll only need to do this once on your machine.

To start your app locally run.

npx sst dev

Now if you go to http://localhost:4321 you’ll see a counter update as you refresh the page.

Finally, you can deploy it by adding the Dockerfile that’s included in this example and running npx sst deploy --stage production.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true });
const redis = new sst.aws.Redis("MyRedis", { vpc });
const cluster = new sst.aws.Cluster("MyCluster", { vpc });

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "4321/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

View the full example.

IAM permissions boundaries

Use permissions boundaries to set the maximum permissions for all IAM roles that’ll be created in your app.

In this example, the Function has the s3:ListAllMyBuckets and sqs:ListQueues permissions. However, we create a permissions boundary that only allows s3:ListAllMyBuckets. And we apply it to all Roles in the app using the global $transform.

As a result, the Function is only allowed to list S3 buckets. If you open the deployed URL, you’ll see that the SQS list call fails.

Learn more about AWS IAM permissions boundaries.

// Create a permission boundary
const permissionsBoundary = new aws.iam.Policy("MyPermissionsBoundary", {
  policy: aws.iam.getPolicyDocumentOutput({
    statements: [
      {
        actions: ["s3:ListAllMyBuckets"],
        resources: ["*"],
      },
    ],
  }).json,
});

// Apply the boundary to all roles
$transform(aws.iam.Role, (args) => {
  args.permissionsBoundary = permissionsBoundary;
});

// The boundary automatically applies to this Function's role
const app = new sst.aws.Function("MyApp", {
  handler: "index.handler",
  permissions: [
    {
      actions: ["s3:ListAllMyBuckets", "sqs:ListQueues"],
      resources: ["*"],
    },
  ],
  url: true,
});

return {
  app: app.url,
};

View the full example.

Current AWS account

You can use the aws.getXXXXOutput() provider functions to get info about the current AWS account. Learn more about provider functions.

return {
  region: aws.getRegionOutput().name,
  account: aws.getCallerIdentityOutput({}).accountId,
};

View the full example.

AWS JSX Email

Uses JSX Email and the Email component to design and send emails.

To test this example, change the sst.config.ts to use your own email address.

sender: "email@example.com"

Then run.

npm install
npx sst dev

You’ll get an email from AWS asking you to confirm your email address. Click the link to verify it.

Next, go to the URL in the sst dev CLI output. You should now receive an email rendered using JSX Email.

import { Template } from "./templates/email";

await render(Template({
  email: "spongebob@example.com",
  name: "Spongebob Squarepants"
}))

Once you are ready to go to production, you can:

Request production access for SES
And use your domain to send emails

const email = new sst.aws.Email("MyEmail", {
  sender: "email@example.com",
});
const api = new sst.aws.Function("MyApi", {
  handler: "index.handler",
  link: [email],
  url: true,
});

return {
  api: api.url,
};

View the full example.

Kinesis streams

Create a Kinesis stream, and subscribe to it with a function.

const stream = new sst.aws.KinesisStream("MyStream");

// Create a function subscribing to all events
stream.subscribe("AllSub", "subscriber.all");

// Create a function subscribing to events of `bar` type
stream.subscribe("FilteredSub", "subscriber.filtered", {
  filters: [
    {
      data: {
        type: ["bar"],
      },
    },
  ],
});

const app = new sst.aws.Function("MyApp", {
  handler: "publisher.handler",
  link: [stream],
  url: true,
});

return {
  app: app.url,
  stream: stream.name,
};

View the full example.

AWS Lambda Go

This example shows how to use the go runtime in your Lambda functions.

Our Go function is in the src directory and we point to it in our function.

new sst.aws.Function("MyFunction", {
  url: true,
  runtime: "go",
  link: [bucket],
  handler: "./src",
});

We are also linking it to an S3 bucket. We can reference the bucket in our function.

func handler() (string, error) {
  bucket, err := resource.Get("MyBucket", "name")
  if err != nil {
    return "", err
  }
  return bucket.(string), nil
}

The resource.Get function is from the SST Go SDK.

import (
  "github.com/sst/sst/v3/sdk/golang/resource"
)

The sst dev CLI also supports running your Go function Live.

const bucket = new sst.aws.Bucket("MyBucket");

new sst.aws.Function("MyFunction", {
  url: true,
  runtime: "go",
  link: [bucket],
  handler: "./src",
});

View the full example.

AWS Lambda retry with queues

An example on how to retry Lambda invocations using SQS queues.

Create a SQS retry queue which will be set as the destination for the Lambda function.

const retryQueue = new sst.aws.Queue("retryQueue");

const bus = new sst.aws.Bus("bus");

const busSubscriber = bus.subscribe("busSubscriber", {
  handler: "src/bus-subscriber.handler",
  environment: {
    RETRIES: "2", // set the number of retries
  },
  link: [retryQueue], // so the function can send messages to the retry queue
});

new aws.lambda.FunctionEventInvokeConfig("eventConfig", {
  functionName: $resolve([busSubscriber.nodes.function.name]).apply(
    ([name]) => name,
  ),
  maximumRetryAttempts: 2, // default is 2, must be between 0 and 2
  destinationConfig: {
    onFailure: {
      destination: retryQueue.arn,
    },
  },
});

Create a bus subscriber which will publish messages to the bus. Include a DLQ for messages that continue to fail.

const dlq = new sst.aws.Queue("dlq");

retryQueue.subscribe({
  handler: "src/retry.handler",
  link: [busSubscriber.nodes.function, retryQueue, dlq],
  timeout: "30 seconds",
  environment: {
    RETRIER_QUEUE_URL: retryQueue.url,
  },
  permissions: [
    {
      actions: ["lambda:GetFunction", "lambda:InvokeFunction"],
      resources: [
        $interpolate`arn:aws:lambda:${aws.getRegionOutput().name}:${
          aws.getCallerIdentityOutput().accountId
        }:function:*`,
      ],
    },
  ],
  transform: {
    function: {
      deadLetterConfig: {
        targetArn: dlq.arn,
      },
    },
  },
});

The Retry function will read mesaages and send back to the queue to be retried with a backoff.

export const handler: SQSHandler = async (evt) => {
  for (const record of evt.Records) {
    const parsed = JSON.parse(record.body);
    console.log("body", parsed);
    const functionName = parsed.requestContext.functionArn
      .replace(":$LATEST", "")
      .split(":")
      .pop();
    if (parsed.responsePayload) {
      const attempt = (parsed.requestPayload.attempts || 0) + 1;

      const info = await lambda.send(
        new GetFunctionCommand({
          FunctionName: functionName,
        }),
      );
      const max =
        Number.parseInt(
          info.Configuration?.Environment?.Variables?.RETRIES || "",
        ) || 0;
      console.log("max retries", max);
      if (attempt > max) {
        console.log(`giving up after ${attempt} retries`);
        // send to dlq
        await sqs.send(
          new SendMessageCommand({
            QueueUrl: Resource.dlq.url,
            MessageBody: JSON.stringify({
              requestPayload: parsed.requestPayload,
              requestContext: parsed.requestContext,
              responsePayload: parsed.responsePayload,
            }),
          }),
        );
        return;
      }
      const seconds = Math.min(Math.pow(2, attempt), 900);
      console.log(
        "delaying retry by ",
        seconds,
        "seconds for attempt",
        attempt,
      );
      parsed.requestPayload.attempts = attempt;
      await sqs.send(
        new SendMessageCommand({
          QueueUrl: Resource.retryQueue.url,
          DelaySeconds: seconds,
          MessageBody: JSON.stringify({
            requestPayload: parsed.requestPayload,
            requestContext: parsed.requestContext,
          }),
        }),
      );
    }

    if (!parsed.responsePayload) {
      console.log("triggering function");
      try {
        await lambda.send(
          new InvokeCommand({
            InvocationType: "Event",
            Payload: Buffer.from(JSON.stringify(parsed.requestPayload)),
            FunctionName: functionName,
          }),
        );
      } catch (e) {
        if (e instanceof ResourceNotFoundException) {
          return;
        }
        throw e;
      }
    }
  }
};

const dlq = new sst.aws.Queue("dlq");

const retryQueue = new sst.aws.Queue("retryQueue");

const bus = new sst.aws.Bus("bus");

const busSubscriber = bus.subscribe("busSubscriber", {
  handler: "src/bus-subscriber.handler",
  environment: {
    RETRIES: "2",
  },
  link: [retryQueue], // so the function can send messages to the queue
});

const publisher = new sst.aws.Function("publisher", {
  handler: "src/publisher.handler",
  link: [bus],
  url: true,
});

new aws.lambda.FunctionEventInvokeConfig("eventConfig", {
  functionName: $resolve([busSubscriber.nodes.function.name]).apply(
    ([name]) => name,
  ),
  maximumRetryAttempts: 1,
  destinationConfig: {
    onFailure: {
      destination: retryQueue.arn,
    },
  },
});

retryQueue.subscribe({
  handler: "src/retry.handler",
  link: [busSubscriber.nodes.function, retryQueue, dlq],
  timeout: "30 seconds",
  environment: {
    RETRIER_QUEUE_URL: retryQueue.url,
  },
  permissions: [
    {
      actions: ["lambda:GetFunction", "lambda:InvokeFunction"],
      resources: [
        $interpolate`arn:aws:lambda:${aws.getRegionOutput().name}:${
          aws.getCallerIdentityOutput().accountId
        }:function:*`,
      ],
    },
  ],
  transform: {
    function: {
      deadLetterConfig: {
        targetArn: dlq.arn,
      },
    },
  },
});

return {
  publisher: publisher.url,
  dlq: dlq.url,
  retryQueue: retryQueue.url,
};

View the full example.

AWS Lambda streaming

An example on how to enable streaming for Lambda functions.

{
  streaming: true
}

While sst dev doesn’t support streaming, you can use the lambda-stream package to test locally.

npm install lambda-stream

Then, you can use the streamifyResponse function to wrap your handler:

import { APIGatewayProxyEventV2 } from "aws-lambda";
import { streamifyResponse, ResponseStream } from "lambda-stream";

export const handler = streamifyResponse(myHandler);

async function myHandler(
  _event: APIGatewayProxyEventV2,
  responseStream: ResponseStream
): Promise<void> {
  return new Promise((resolve, _reject) => {
    responseStream.setContentType('text/plain')
    responseStream.write('Hello')
    setTimeout(() => {
      responseStream.write(' World')
      responseStream.end()
      resolve()
    }, 3000)
  })
}

When deployed, this will use the awslambda.streamifyResponse.

To test this in your terminal, use the curl command with the --no-buffer option.

curl --no-buffer https://u3dyblk457ghskwbmzrbylpxoi0ayrbb.lambda-url.us-east-1.on.aws

Here we are using a Function URL directly because API Gateway doesn’t support streaming.

const fn = new sst.aws.Function("MyFunction", {
  url: true,
  streaming: true,
  timeout: "15 minutes",
  handler: "index.handler",
});

return {
  url: fn.url,
};

View the full example.

AWS Lambda in a VPC

You can use SST to locally work on Lambda functions that are in a VPC. To do so, you’ll need to enable bastion and nat on the Vpc component.

new sst.aws.Vpc("MyVpc", { bastion: true, nat: "managed" });

The NAT gateway is necessary to allow your Lambda function to connect to the internet. While, the bastion host is necessary for your local machine to be able to tunnel to the VPC.

You’ll need to install the tunnel, if you haven’t done this before.

sudo sst tunnel install

This needs sudo to create the network interface on your machine. You’ll only need to do this once.

Now you can run sst dev, your function can access resources in the VPC. For example, here we are connecting to a Redis cluster.

const redis = new Cluster(
  [{ host: Resource.MyRedis.host, port: Resource.MyRedis.port }],
  {
    dnsLookup: (address, callback) => callback(null, address),
    redisOptions: {
      tls: {},
      username: Resource.MyRedis.username,
      password: Resource.MyRedis.password,
    },
  }
);

The Redis cluster is in the same VPC as the function.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true, nat: "managed" });
const redis = new sst.aws.Redis("MyRedis", { vpc });
const api = new sst.aws.Function("MyFunction", {
  vpc,
  url: true,
  link: [redis],
  handler: "index.handler"
});

return {
  url: api.url,
};

View the full example.

AWS multi-region

To deploy resources to multiple AWS regions, you can create a new provider for the region you want to deploy to.

const provider = new aws.Provider("MyProvider", { region: "us-west-2" });

And then pass that in to the resource.

new sst.aws.Function("MyFunction", { handler: "index.handler" }, { provider });

If no provider is passed in, the default provider will be used. And if no region is specified, the default region from your credentials will be used.

const east = new sst.aws.Function("MyEastFunction", {
  url: true,
  handler: "index.handler",
});

const provider = new aws.Provider("MyWestProvider", { region: "us-west-2" });
const west = new sst.aws.Function(
  "MyWestFunction",
  {
    url: true,
    handler: "index.handler",
  },
  { provider }
);

return {
  east: east.url,
  west: west.url,
};

View the full example.

AWS Next.js add behavior

Here’s how to add additional routes or cache behaviors to the CDN of a Next.js app deployed with OpenNext to AWS.

Specify the path pattern that you want to forward to your new origin. For example, to forward all requests to the /blog path to a different origin.

pathPattern: "/blog/*"

And then specify the domain of the new origin.

domainName: "blog.example.com"

We use this to transform our site’s CDN and add the additional behaviors.

const blogOrigin = {
  // The domain of the new origin
  domainName: "blog.example.com",
  originId: "blogCustomOrigin",
  customOriginConfig: {
    httpPort: 80,
    httpsPort: 443,
    originSslProtocols: ["TLSv1.2"],
    // If HTTPS is supported
    originProtocolPolicy: "https-only",
  },
};

const cacheBehavior = {
  // The path to forward to the new origin
  pathPattern: "/blog/*",
  targetOriginId: blogOrigin.originId,
  viewerProtocolPolicy: "redirect-to-https",
  allowedMethods: ["GET", "HEAD", "OPTIONS"],
  cachedMethods: ["GET", "HEAD"],
  forwardedValues: {
    queryString: true,
    cookies: {
      forward: "all",
    },
  },
};

new sst.aws.Nextjs("MyWeb", {
  transform: {
    cdn: (options: sst.aws.CdnArgs) => {
      options.origins = $resolve(options.origins).apply(val => [...val, blogOrigin]);

      options.orderedCacheBehaviors = $resolve(
        options.orderedCacheBehaviors || []
      ).apply(val => [...val, cacheBehavior]);
    },
  },
});

View the full example.

AWS Next.js basic auth

Deploys a simple Next.js app and adds basic auth to it.

This is useful for dev environments where you want to share your app your team but ensure that it’s not publicly accessible.

This works by injecting some code into a CloudFront function that checks the basic auth header and matches it against the USERNAME and PASSWORD secrets.

{
  injection: $interpolate`
    if (
        !event.request.headers.authorization
          || event.request.headers.authorization.value !== "Basic ${basicAuth}"
       ) {
      return {
        statusCode: 401,
        headers: {
          "www-authenticate": { value: "Basic" }
        }
      };
    }`,
}

To deploy this, you need to first set the USERNAME and PASSWORD secrets.

sst secret set USERNAME my-username
sst secret set PASSWORD my-password

If you are deploying this to preview environments, you might want to set the secrets using the --fallback flag.

const username = new sst.Secret("USERNAME");
const password = new sst.Secret("PASSWORD");
const basicAuth = $resolve([username.value, password.value]).apply(
  ([username, password]) =>
    Buffer.from(`${username}:${password}`).toString("base64")
);

new sst.aws.Nextjs("MyWeb", {
  server: {
    // Don't password protect prod
    edge: $app.stage !== "production"
      ? {
        viewerRequest: {
          injection: $interpolate`
            if (
                !event.request.headers.authorization
                  || event.request.headers.authorization.value !== "Basic ${basicAuth}"
               ) {
              return {
                statusCode: 401,
                headers: {
                  "www-authenticate": { value: "Basic" }
                }
              };
            }`,
        },
      }
      : undefined,
  },
});

View the full example.

AWS Next.js container with Redis

Creates a hit counter app with Next.js and Redis.

This deploys Next.js as a Fargate service to ECS and it’s linked to Redis.

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

Since our Redis cluster is in a VPC, we’ll need a tunnel to connect to it from our local machine.

sudo npx sst tunnel install

This needs sudo to create a network interface on your machine. You’ll only need to do this once on your machine.

To start your app locally run.

npx sst dev

Now if you go to http://localhost:3000 you’ll see a counter update as you refresh the page.

Finally, you can deploy it by:

Setting output: "standalone" in your next.config.mjs file.
Adding a Dockerfile that’s included in this example.
Running npx sst deploy --stage production.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true });
const redis = new sst.aws.Redis("MyRedis", { vpc });
const cluster = new sst.aws.Cluster("MyCluster", { vpc });

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

View the full example.

AWS Next.js streaming

An example of how to use streaming Next.js RSC. Uses Suspense to stream an async component.

<Suspense fallback={<div>Loading...</div>}>
  <Friends />
</Suspense>

For this demo we also need to make sure the route is not statically built.

export const dynamic = "force-dynamic";

This is deployed with OpenNext, which needs a config to enable streaming.

export default {
  default: {
    override: {
      wrapper: "aws-lambda-streaming"
    }
  }
};

You should see the friends section load after a 3 second delay.

Safari uses a different heuristic to determine when to stream data. You need to render enough initial HTML to trigger streaming. This is typically only a problem for demo apps.

new sst.aws.Nextjs("MyWeb");

View the full example.

AWS Postgres local

In this example, we connect to a locally running Postgres instance for dev. While on deploy, we use RDS.

We use the docker run CLI to start a local container with Postgres. You don’t have to use Docker, you can use Postgres.app or any other way to run Postgres locally.

docker run \
  --rm \
  -p 5432:5432 \
  -v $(pwd)/.sst/storage/postgres:/var/lib/postgresql/data \
  -e POSTGRES_USER=postgres \
  -e POSTGRES_PASSWORD=password \
  -e POSTGRES_DB=local \
  postgres:16.4

The data is saved to the .sst/storage directory. So if you restart the dev server, the data will still be there.

We then configure the dev property of the Postgres component with the settings for the local Postgres instance.

dev: {
  username: "postgres",
  password: "password",
  database: "local",
  port: 5432,
}

By providing the dev prop for Postgres, SST will use the local Postgres instance and not deploy a new RDS database when running sst dev.

It also allows us to access the database through a Reosurce link without having to conditionally check if we are running locally.

const pool = new Pool({
  host: Resource.MyPostgres.host,
  port: Resource.MyPostgres.port,
  user: Resource.MyPostgres.username,
  password: Resource.MyPostgres.password,
  database: Resource.MyPostgres.database,
});

The above will work in both sst dev and sst deploy.

const vpc = new sst.aws.Vpc("MyVpc", { nat: "ec2" });

const rds = new sst.aws.Postgres("MyPostgres", {
  dev: {
    username: "postgres",
    password: "password",
    database: "local",
    host: "localhost",
    port: 5432,
  },
  vpc,
});

new sst.aws.Function("MyFunction", {
  vpc,
  url: true,
  link: [rds],
  handler: "index.handler",
});

View the full example.

Prisma in Lambda

To use Prisma in a Lambda function you need to

Generate the Prisma Client with the right architecture
Copy the generated client to the function
Run the function inside a VPC

You can set the architecture using the binaryTargets option in prisma/schema.prisma.

// For x86
binaryTargets = ["native", "rhel-openssl-3.0.x"]
// For ARM
// binaryTargets = ["native", "linux-arm64-openssl-3.0.x"]

You can also switch to ARM, just make sure to also change the function architecture in your sst.config.ts.

{
  // For ARM
  architecture: "arm64"
}

To generate the client, you need to run prisma generate when you make changes to the schema.

Since this needs to be done on every deploy, we add a postinstall script to the package.json.

"scripts": {
  "postinstall": "prisma generate"
}

This runs the command on npm install.

We then need to copy the generated client to the function when we deploy.

{
  copyFiles: [{ from: "node_modules/.prisma/client/" }]
}

Our function also needs to run inside a VPC, since Prisma doesn’t support the Data API.

{
  vpc
}

Prisma in serverless environments

Prisma is not great in serverless environments. For a couple of reasons:

It doesn’t support Data API, so you need to manage the connection pool on your own.
Without the Data API, your functions need to run inside a VPC.
- You cannot use sst dev without connecting to the VPC.
Due to the internal architecture of their client, it’s also has slower cold starts.

Instead we recommend using Drizzle. This example is here for reference for people that are already using Prisma.

const vpc = new sst.aws.Vpc("MyVpc", { nat: "managed" });
const rds = new sst.aws.Postgres("MyPostgres", { vpc });

const api = new sst.aws.Function("MyApi", {
  vpc,
  url: true,
  link: [rds],
  // For ARM
  // architecture: "arm64",
  handler: "index.handler",
  copyFiles: [{ from: "node_modules/.prisma/client/" }],
});

return {
  api: api.url,
};

View the full example.

Puppeteer in Lambda

To use Puppeteer in a Lambda function you need:

puppeteer-core
Chromium
- In sst dev, we’ll use a locally installed Chromium version.
- In sst deploy, we’ll use the @sparticuz/chromium package. It comes with a pre-built binary for Lambda.

Chromium version

Since Puppeteer has a preferred version of Chromium, we’ll need to check the version of Chrome that a given version of Puppeteer supports. Head over to the Puppeteer’s Chromium Support page and check which versions work together.

For example, Puppeteer v23.1.1 supports Chrome for Testing 127.0.6533.119. So, we’ll use the v127 of @sparticuz/chromium.

npm install puppeteer-core@23.1.1 @sparticuz/chromium@127.0.0

Install Chromium locally

To use this locally, you’ll need to install Chromium.

npx @puppeteer/browsers install chromium@latest --path /tmp/localChromium

Once installed you’ll see the location of the Chromium binary, /tmp/localChromium/chromium/mac_arm-1350406/chrome-mac/Chromium.app/Contents/MacOS/Chromium.

Update this in your Lambda function.

// This is the path to the local Chromium binary
const YOUR_LOCAL_CHROMIUM_PATH = "/tmp/localChromium/chromium/mac_arm-1350406/chrome-mac/Chromium.app/Contents/MacOS/Chromium";

You’ll notice we are using the right binary with the SST_DEV environment variable.

const browser = await puppeteer.launch({
  args: chromium.args,
  defaultViewport: chromium.defaultViewport,
  executablePath: process.env.SST_DEV
    ? YOUR_LOCAL_CHROMIUM_PATH
    : await chromium.executablePath(),
  headless: chromium.headless,
});

Deploy

We don’t need a layer to deploy this because @sparticuz/chromium comes with a pre-built binary for Lambda.

We just need to set it in the nodejs.install.

{
  nodejs: {
    install: ["@sparticuz/chromium"]
  }
}

And on deploy, SST will use the right binary.

We are giving our function more memory and a longer timeout since running Puppeteer can take a while.

const api = new sst.aws.Function("MyFunction", {
  url: true,
  memory: "2 GB",
  timeout: "15 minutes",
  handler: "index.handler",
  nodejs: {
    install: ["@sparticuz/chromium"],
  },
});

return {
  url: api.url,
};

View the full example.

Create an SQS queue, subscribe to it, and publish to it from a function.

const queue = new sst.aws.Queue("MyQueue");
queue.subscribe("subscriber.handler");

const app = new sst.aws.Function("MyApp", {
  handler: "publisher.handler",
  link: [queue],
  url: true,
});

return {
  app: app.url,
  queue: queue.url,
};

View the full example.

AWS Redis local

In this example, we connect to a local Docker Redis instance for dev. While on deploy, we use Redis ElastiCache.

We use the docker run CLI to start a local Redis server. You don’t have to use Docker, you can run it locally any way you want.

docker run \
  --rm \
  -p 6379:6379 \
  -v $(pwd)/.sst/storage/redis:/data \
  redis:latest

The data is persisted to the .sst/storage directory. So if you restart the dev server, the data will still be there.

We then configure the dev property of the Redis component with the settings for the local Redis server.

dev: {
  host: "localhost",
  port: 6379
}

By providing the dev prop for Redis, SST will use the local Redis server and not deploy a new Redis ElastiCache cluster when running sst dev.

It also allows us to access Redis through a Reosurce link.

const client = Resource.MyRedis.host === "localhost"
  ? new Redis({
      host: Resource.MyRedis.host,
      port: Resource.MyRedis.port,
    })
  : new Cluster(
      [{
        host: Resource.MyRedis.host,
        port: Resource.MyRedis.port,
      }],
      {
        redisOptions: {
          tls: { checkServerIdentity: () => undefined },
          username: Resource.MyRedis.username,
          password: Resource.MyRedis.password,
        },
      },
    );

The local Redis server is running in standalone mode, whereas on deploy it’ll be in cluster mode. So our Lambda function needs to connect using the right config.

const vpc = new sst.aws.Vpc("MyVpc", { nat: "managed" });

const redis = new sst.aws.Redis("MyRedis", {
  dev: {
    host: "localhost",
    port: 6379,
  },
  vpc,
});

new sst.aws.Function("MyApp", {
  vpc,
  url: true,
  link: [redis],
  handler: "index.handler",
});

View the full example.

AWS Remix container with Redis

Creates a hit counter app with Remix and Redis.

This deploys Remix as a Fargate service to ECS and it’s linked to Redis.

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

Since our Redis cluster is in a VPC, we’ll need a tunnel to connect to it from our local machine.

sudo npx sst tunnel install

This needs sudo to create a network interface on your machine. You’ll only need to do this once on your machine.

To start your app locally run.

npx sst dev

Now if you go to http://localhost:5173 you’ll see a counter update as you refresh the page.

Finally, you can deploy it by adding the Dockerfile that’s included in this example and running npx sst deploy --stage production.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true });
const redis = new sst.aws.Redis("MyRedis", { vpc });
const cluster = new sst.aws.Cluster("MyCluster", { vpc });

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

View the full example.

AWS Remix streaming

Follows the Remix Streaming guide to create an app that streams data.

Uses the defer utility to stream data through the loader function.

return defer({
  spongebob,
  friends: friendsPromise,
});

Then uses the the Suspense and Await components to render the data.

<Suspense fallback={<div>Loading...</div>}>
  <Await resolve={friends}>
    { /* ... */ }
  </Await>
</Suspense>

You should see the friends section load after a 3 second delay.

Safari uses a different heuristic to determine when to stream data. You need to render enough initial HTML to trigger streaming. This is typically only a problem for demo apps.

Streaming works out of the box with the Remix component.

new sst.aws.Remix("MyWeb");

View the full example.

Router and bucket

Creates a router that serves static files from the public folder of a given bucket.

// Create a bucket that CloudFront can access
const bucket = new sst.aws.Bucket("MyBucket", {
  access: "cloudfront",
});

// Upload the image to the `public` folder
new aws.s3.BucketObjectv2("MyImage", {
  bucket: bucket.name,
  key: "public/spongebob.svg",
  contentType: "image/svg+xml",
  source: $asset("spongebob.svg"),
});

const router = new sst.aws.Router("MyRouter", {
  routes: {
    "/*": {
      bucket,
      rewrite: { regex: "^/(.*)$", to: "/public/$1" },
    },
  },
});

return {
  image: $interpolate`${router.url}/spongebob.svg`,
};

View the full example.

Router and function URL

Creates a router that routes all requests to a function with a URL.

const api = new sst.aws.Function("MyApi", {
  handler: "api.handler",
  url: true,
});
const bucket = new sst.aws.Bucket("MyBucket", {
  access: "public",
});
const router = new sst.aws.Router("MyRouter", {
  domain: "router.ion.dev.sst.dev",
  routes: {
    "/api/*": api.url,
    "/*": $interpolate`https://${bucket.domain}`,
  },
});

return {
  router: router.url,
  bucket: bucket.domain,
};

View the full example.

AWS Cluster Service Discovery

In this example, we are connecting to a service running on a cluster using its AWS Cloud Map service host name. This is useful for service discovery.

We are deploying a service to a cluster in a VPC. And we can access it within the VPC using the service’s cloud map hostname.

const reponse = await fetch(`http://${Resource.MyService.service}`);

Here we are accessing it through a Lambda function that’s linked to the service and is deployed to the same VPC.

const vpc = new sst.aws.Vpc("MyVpc", { nat: "ec2" });

const cluster = new sst.aws.Cluster("MyCluster", { vpc });
const service = cluster.addService("MyService");

new sst.aws.Function("MyFunction", {
  vpc,
  url: true,
  link: [service],
  handler: "lambda.handler",
});

View the full example.

Sharp in Lambda

Uses the Sharp library to resize images. In this example, it resizes a logo.png local file to 100x100 pixels.

{
  nodejs: { install: ["sharp"] }
}

We don’t need a layer to deploy this because sharp comes with a pre-built binary for Lambda. This is handled by nodejs.install.

In dev, this uses the sharp npm package locally.

{
  "dependencies": {
    "sharp": "^0.33.5"
  }
}

On deploy, SST will use the right binary from the sharp package for the target Lambda architecture.

const func = new sst.aws.Function("MyFunction", {
  url: true,
  handler: "index.handler",
  nodejs: { install: ["sharp"] },
  copyFiles: [{ from: "logo.png" }],
});

return {
  url: func.url,
};

View the full example.

AWS SolidStart WebSocket endpoint

Deploys a SolidStart app with a WebSocket endpoint in a container to AWS.

Uses the experimental WebSocket support in Nitro.

export default defineConfig({
  server: {
    experimental: {
      websocket: true,
    },
  },
}).addRouter({
  name: "ws",
  type: "http",
  handler: "./src/ws.ts",
  target: "server",
  base: "/ws",
});

Once deployed you can test the /ws endpoint and it’ll send a message back after a 3s delay.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true });
const cluster = new sst.aws.Cluster("MyCluster", { vpc });

cluster.addService("MyService", {
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

View the full example.

AWS static site basic auth

This deploys a simple static site and adds basic auth to it.

This is useful for dev environments where you want to share a static site with your team but ensure that it’s not publicly accessible.

This works by injecting some code into a CloudFront function that checks the basic auth header and matches it against the USERNAME and PASSWORD secrets.

{
  injection: $interpolate`
    if (
        !event.request.headers.authorization
          || event.request.headers.authorization.value !== "Basic ${basicAuth}"
       ) {
      return {
        statusCode: 401,
        headers: {
          "www-authenticate": { value: "Basic" }
        }
      };
    }`,
}

To deploy this, you need to first set the USERNAME and PASSWORD secrets.

sst secret set USERNAME my-username
sst secret set PASSWORD my-password

If you are deploying this to preview environments, you might want to set the secrets using the --fallback flag.

const username = new sst.Secret("USERNAME");
const password = new sst.Secret("PASSWORD");
const basicAuth = $resolve([username.value, password.value]).apply(
  ([username, password]) =>
    Buffer.from(`${username}:${password}`).toString("base64")
);

new sst.aws.StaticSite("MySite", {
  path: "site",
  // Don't password protect prod
  edge: $app.stage !== "production"
    ? {
      viewerRequest: {
        injection: $interpolate`
          if (
              !event.request.headers.authorization
                || event.request.headers.authorization.value !== "Basic ${basicAuth}"
             ) {
            return {
              statusCode: 401,
              headers: {
                "www-authenticate": { value: "Basic" }
              }
            };
          }`,
      },
    }
    : undefined,
});

View the full example.

AWS static site

Deploy a simple HTML file as a static site with S3 and CloudFront. The website is stored in the site/ directory.

new sst.aws.StaticSite("MySite", {
  path: "site",
});

View the full example.

AWS SvelteKit container with Redis

Creates a hit counter app with SvelteKit and Redis.

This deploys SvelteKit as a Fargate service to ECS and it’s linked to Redis.

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

Since our Redis cluster is in a VPC, we’ll need a tunnel to connect to it from our local machine.

sudo npx sst tunnel install

This needs sudo to create a network interface on your machine. You’ll only need to do this once on your machine.

To start your app locally run.

npx sst dev

Now if you go to http://localhost:5173 you’ll see a counter update as you refresh the page.

Finally, you can deploy it by adding the Dockerfile that’s included in this example and running npx sst deploy --stage production.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true });
const redis = new sst.aws.Redis("MyRedis", { vpc });
const cluster = new sst.aws.Cluster("MyCluster", { vpc });

cluster.addService("MyService", {
  link: [redis],
  loadBalancer: {
    ports: [{ listen: "80/http", forward: "3000/http" }],
  },
  dev: {
    command: "npm run dev",
  },
});

View the full example.

Swift in Lambda

Deploys a simple Swift application to Lambda using the al2023 runtime.

Check out the README in the repo for more details.

const swift = new sst.aws.Function("Swift", {
  runtime: "provided.al2023",
  architecture: process.arch === "arm64" ? "arm64" : "x86_64",
  bundle: build("app"),
  handler: "bootstrap",
  url: true,
});
const router = new sst.aws.Router("SwiftRouter", {
  routes: {
    "/*": swift.url,
  },
  domain: "swift.dev.sst.dev",
});
return {
  url: router.url,
};

View the full example.

T3 Stack in AWS

Deploy T3 stack with Drizzle and Postgres to AWS.

This example was created using create-t3-app and the following options: tRPC, Drizzle, no auth, Tailwind, Postgres, and the App Router.

Instead of a local database, we’ll be using an RDS Postgres database.

const pool = new Pool({
  host: Resource.MyPostgres.host,
  port: Resource.MyPostgres.port,
  user: Resource.MyPostgres.username,
  password: Resource.MyPostgres.password,
  database: Resource.MyPostgres.database,
});

Similarly, for Drizzle Kit.

export default {
  schema: "./src/server/db/schema.ts",
  dialect: "postgresql",
  dbCredentials: {
    ssl: {
      rejectUnauthorized: false,
    },
    host: Resource.MyPostgres.host,
    port: Resource.MyPostgres.port,
    user: Resource.MyPostgres.username,
    password: Resource.MyPostgres.password,
    database: Resource.MyPostgres.database,
  },
  tablesFilter: ["aws-t3_*"],
} satisfies Config;

In our Next.js app we can access our Postgres database because we link them both. We don’t need to use our .env files.

 const rds = new sst.aws.Postgres("MyPostgres", { vpc, proxy: true });

 new sst.aws.Nextjs("MyWeb", {
   vpc,
   link: [rds]
 });

To run this in dev mode run:

npm install
npx sst dev

It’ll take a few minutes to deploy the database and the VPC.

This also starts a tunnel to let your local machine connect to the RDS Postgres database. Make sure you have it installed, you only need to do this once for your local machine.

sudo npx sst tunnel install

Now in a new terminal you can run the database migrations.

npm run db:push

We also have the Drizzle Studio start automatically in dev mode under the Studio tab.

new sst.x.DevCommand("Studio", {
  link: [rds],
  dev: {
    command: "npx drizzle-kit studio",
  },
});

And to make sure our credentials are available, we update our package.json with the sst shell CLI.

"db:generate": "sst shell drizzle-kit generate",
"db:migrate": "sst shell drizzle-kit migrate",
"db:push": "sst shell drizzle-kit push",
"db:studio": "sst shell drizzle-kit studio",

So running npm run db:push will run Drizzle Kit with the right credentials.

To deploy this to production run:

npx sst deploy --stage production

Then run the migrations.

npx sst shell --stage production npx drizzle-kit push

If you are running this locally, you’ll need to have a tunnel running.

npx sst tunnel --stage production

If you are doing this in a CI/CD pipeline, you’d want your build containers to be in the same VPC.

const vpc = new sst.aws.Vpc("MyVpc", { bastion: true, nat: "ec2" });
const rds = new sst.aws.Postgres("MyPostgres", { vpc, proxy: true });

new sst.aws.Nextjs("MyWeb", {
  vpc,
  link: [rds]
});

new sst.x.DevCommand("Studio", {
  link: [rds],
  dev: {
    command: "npx drizzle-kit studio",
  },
});

View the full example.

Create an SNS topic, publish to it from a function, and subscribe to it with a function and a queue.

const queue = new sst.aws.Queue("MyQueue");
queue.subscribe("subscriber.handler");

const topic = new sst.aws.SnsTopic("MyTopic");
topic.subscribe("MySubscriber1", "subscriber.handler", {});
topic.subscribeQueue("MySubscriber2", queue.arn);

const app = new sst.aws.Function("MyApp", {
  handler: "publisher.handler",
  link: [topic],
  url: true,
});

return {
  app: app.url,
  topic: topic.name,
};

View the full example.

Vector search

Store and search for vector data using the Vector component. Includes a seeder API that uses an LLM to generate embeddings for some movies and optionally their posters.

Once seeded, you can call the search API to query the vector database.

const OpenAiApiKey = new sst.Secret("OpenAiApiKey");
const vector = new sst.aws.Vector("MyVectorDB", {
  dimension: 1536,
});

const seeder = new sst.aws.Function("Seeder", {
  handler: "index.seeder",
  link: [OpenAiApiKey, vector],
  copyFiles: [
    { from: "iron-man.jpg", to: "iron-man.jpg" },
    {
      from: "black-widow.jpg",
      to: "black-widow.jpg",
    },
    {
      from: "spider-man.jpg",
      to: "spider-man.jpg",
    },
    { from: "thor.jpg", to: "thor.jpg" },
    {
      from: "captain-america.jpg",
      to: "captain-america.jpg",
    },
  ],
  url: true,
});

const app = new sst.aws.Function("MyApp", {
  handler: "index.app",
  link: [OpenAiApiKey, vector],
  url: true,
});

return { seeder: seeder.url, app: app.url };

View the full example.

React SPA with Vite

Deploy a React single-page app (SPA) with Vite to S3 and CloudFront.

new sst.aws.StaticSite("Web", {
  build: {
    command: "pnpm run build",
    output: "dist",
  },
});

View the full example.

Cloudflare Cron

This example creates a Cloudflare Worker that runs on a schedule.

const cron = new sst.cloudflare.Cron("Cron", {
  job: "index.ts",
  schedules: ["* * * * *"]
});

return {};

View the full example.

Cloudflare KV

This example creates a Cloudflare KV namespace and links it to a worker. Now you can use the SDK to interact with the KV namespace in your worker.

const storage = new sst.cloudflare.Kv("MyStorage");
const worker = new sst.cloudflare.Worker("Worker", {
  url: true,
  link: [storage],
  handler: "index.ts",
});

return {
  url: worker.url,
};

View the full example.

Link multiple secrets

You might have multiple secrets that need to be used across your app. It can be tedious to create a new secret and link it to each function or resource.

A common pattern to addresses this is to create an object with all your secrets and then link them all at once. Now when you have a new secret, you can add it to the object and it will be automatically available to all your resources.

// Manage all secrets together
const secrets = {
  secret1: new sst.Secret("Secret1", "some-secret-value-1"),
  secret2: new sst.Secret("Secret2", "some-secret-value-2"),
};
const allSecrets = Object.values(secrets);

const bucket = new sst.aws.Bucket("MyBucket");

const api = new sst.aws.Function("MyApi", {
  link: [bucket, ...allSecrets],
  handler: "index.handler",
  url: true,
});

return {
  url: api.url,
};

View the full example.

Default function props

Set default props for all the functions in your app using the global $transform.

$transform(sst.aws.Function, (args) => {
  args.runtime = "nodejs14.x";
  args.environment = {
    FOO: "BAR",
  };
});
new sst.aws.Function("MyFunction", {
  handler: "index.ts",
});

View the full example.

Vercel domains

Creates a router that uses domains purchased through and hosted in your Vercel account. Ensure the VERCEL_API_TOKEN and VERCEL_TEAM_ID environment variables are set.

const router = new sst.aws.Router("MyRouter", {
  domain: {
    name: "ion.sst.moe",
    dns: sst.vercel.dns({ domain: "sst.moe" }),
  },
  routes: {
    "/*": "https://sst.dev",
  },
});
return {
  router: router.url,
};

View the full example.