> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kuest.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Terraform

> Provision supported deployment targets declaratively with Terraform.

## Available environments

* `production/gke`
* `production/kubernetes`
* `production/cloud-run`
* `production/fly`
* `production/digital-ocean`

## Typical workflow

```bash theme={null}
cd infra/terraform/environments/production/<target>
cp terraform.tfvars.example terraform.tfvars
terraform init
terraform plan
terraform apply
```

## Target notes

<AccordionGroup>
  <Accordion title="GKE Autopilot">
    Creates the Kubernetes cluster itself. After apply, fetch credentials with `gcloud container clusters get-credentials`.
  </Accordion>

  <Accordion title="Kubernetes">
    Deploys app resources into an existing cluster, such as one created by the GKE target.
  </Accordion>

  <Accordion title="Cloud Run">
    Requires Secret Manager secrets to exist before apply.
  </Accordion>

  <Accordion title="Fly.io">
    Uses `flyctl deploy` via Terraform `local-exec`. Secrets are still managed manually in Fly.io.
  </Accordion>

  <Accordion title="DigitalOcean">
    Requires `DIGITALOCEAN_TOKEN` and repository access configured in App Platform.
  </Accordion>
</AccordionGroup>

## General rules

* Use immutable image refs whenever possible.
* Point `SITE_URL` to the public canonical endpoint.
* For Cloud Run, Kubernetes, and DigitalOcean, provide core secrets plus exactly one storage profile.
* Keep scheduler behavior aligned with [Scheduled Jobs](/manual-installation/scheduler-jobs).