laranjabeta

What gets deployed

laranja maps each thing it finds in your code to a small, predictable set of resources — all in your own account. The mapping depends on your provider:

What you writeAWSAzure
http() appProxy Lambda + Function URLHTTP-triggered function in a Function App
@Cron / cron()Lambda + EventBridge ruleTimer-triggered function
@Queue / queue()SQS queue + consumer LambdaStorage Queue + queue-triggered function
workers() rootOne consolidated worker LambdaIts own Function App
LogsCloudWatch log group per functionApplication Insights + Log Analytics

The rest of this page details AWS. For the Azure resource list — a single Function App per workload, its storage account, and its monitoring stack — see Deploying to Azure → What gets deployed.

HTTP app → one function behind a public URL

Your entire app is deployed as one Lambda function fronted by a Lambda Function URL. There is no API Gateway.

  • All routes are served by this single proxy Lambda.
  • The Function URL is public (authType: NONE); cross-origin access is off unless you opt in with cors, and your app handles auth as it sees fit.
  • Memory and timeout come from compute in your config, overridable per-resource under the http key in resources.
  • The public HTTPS URL is emitted as the HttpUrl stack output.
GET  https://‹id›.lambda-url.‹region›.on.aws/

@Cron → a scheduled function

Each cron handler becomes its own Lambda plus an EventBridge rule that invokes it on schedule.

@Queue → a queue + a consumer function

Each queue handler becomes an SQS queue and a consumer Lambda wired to it.

  • Encryption: SQS-managed (SSE-SQS).
  • FIFO queues are created when the name ends in .fifo or fifo: true is set (content-based deduplication is enabled for FIFO). FIFO is AWS-only.
  • Default batch size: 10. Your handler is invoked per message with the JSON-parsed body.
  • Partial-batch failures are enabled: throwing for one message fails only that message; the rest of the batch still succeeds.
  • Consumer memory/timeout come from compute. The queue's visibility timeout is derived to satisfy AWS's rule that it be ≥ the consumer timeout, and can be set explicitly per queue via resources.
  • The queue URL is emitted as a stack output.

Naming

Everything is named deterministically — no random suffixes:

ResourcePatternExample
CloudFormation stack‹name›-‹stage›my-api-prod
Lambda functions‹name›-‹fn›-‹stage›my-api-app-prod, my-api-sendEmail-prod

‹fn› is the handler's name (the method/function name), or the explicit id you set on @Cron/@Queue. The HTTP proxy uses app. Names are truncated to AWS's 64-character limit and sanitized to allowed characters. Azure follows the same ‹name›-‹fn›-‹stage› shape within its own naming rules.

Because the stage is part of the names, multiple stages can live in one account without colliding — see Stages & environments.

Supporting resources

Each Lambda gets an IAM execution role (and the queue consumers get the permissions to read their queue / EventBridge to invoke crons). The first deploy to an account/region also creates the one-time CDK bootstrap resources (an S3 asset bucket and roles).