laranjabeta

How it works

laranja reads your source to find your HTTP app, cron jobs, queue consumers, and env vars, then synthesizes a deployment template — CloudFormation on AWS, ARM on Azure — that your local CLI applies with your own credentials. It never executes your code, and it never holds your cloud keys. Two things are worth knowing about that.

It reads your code — it never runs it

laranja discovers your infrastructure by reading your source: your HTTP app and its routes, your @Cron / cron() jobs and their schedules, your @Queue / queue() consumers, and the env vars you wrap with env(). It does this without executing your code, so planning a deploy is always safe — nothing of yours runs just to figure out what to deploy.

That's also why a few things must be written so laranja can see them: schedules use literal builders like rate(5, "minutes"), and env("…") takes a string literal.

It deploys into your own account

laranja turns what it found into real cloud resources — a function serving your app, a schedule per cron, a queue per consumer — and deploys them with your credentials into your account. Which resources depends on the provider you picked; see what gets deployed for the full mapping.

  • The provider toolchain is embedded (the AWS CDK toolkit, or ARM for Azure), so there's nothing extra to install — no CDK, no CLI, no Bicep.
  • plan previews what a deploy would change (created/changed/unchanged); destroy tears the deployment down.
  • Outputs (your HTTPS URL, queue URLs) are printed when the deploy finishes.
  • On AWS only, the first deploy to a new account/region runs a one-time bootstrap.

The template is built on the server — your code stays local

plan and deploy synthesize the deployment template on the laranja server. laranja scans and bundles your code locally, then sends only a description of your infrastructure (the internal model plus the asset hashes of your bundles) — your source code and bundles never leave your machine. The returned template is then applied to your cloud with your own credentials. This is why these commands need a LARANJA_API_KEY and a projectId (run laranja init once to set both up).