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

# Deploy

> Deploy Greenlight with Docker Compose, standalone, or on Kubernetes.

## Prerequisites

* Docker (for the compose path) or Go 1.22+ and Node 20+ (for source builds).
* Redis instance reachable by every worker.
* S3-compatible object store (AWS S3, MinIO, or `noop` for stateless runs).
* The `screenjson` binary — either baked into the Greenlight image (the default Dockerfile copies it in) or mounted from `SCREENJSON_BIN`.

## Docker Compose (fastest)

```bash theme={null}
git clone https://github.com/screenjson/greenlight.git
cd greenlight
cp env.example backend/.env
docker build -f deployment/Dockerfile -t screenjson/greenlight:latest .
cd deployment
docker-compose up
```

The provided `docker-compose.yml` boots:

* Greenlight on `:19000` (maps internal `:8080`)
* Redis
* MinIO on `:9000` (S3 API) and `:9001` (console)

Drop files into the MinIO `greenlight/screenplays/input/` prefix — the worker picks them up.

## Standalone backend

```bash theme={null}
cd backend
cp ../env.example .env   # or backend/.env
go run cmd/greenlight/main.go
```

Watches `backend/.env`, `config/tasks.yml`, and any environment variables set before startup.

## Frontend dev server

```bash theme={null}
cd frontend
npm install
npm run dev
```

Runs on port `5173` (Vite default). The production Svelte build lives at `backend/greenlight/` (the `UI_DIST_DIR`) and is served by the Go binary in one process.

## Kubernetes

Greenlight scales horizontally. Each replica pulls from the shared Redis queue. Recommended pattern:

```
Deployment: greenlight (replicas: 3-8)
  env:
    - QUEUE_BACKEND=redis
    - REDIS_ADDR=redis.default.svc:6379
    - STORAGE_BACKEND=s3
    - S3_INPUT_BUCKET=...
    - S3_OUTPUT_BUCKET=...
  command: ["/app/greenlight"]

StatefulSet: redis
Service: greenlight  (ClusterIP on 8080)
Ingress: docs.example.com → greenlight
```

Sizing: each worker runs one conversion at a time. `MAX_WORKERS=4` per replica is a sensible starting point (matches the default). Scale replicas horizontally.

## Health endpoints

| Path           | Purpose                                                  |
| -------------- | -------------------------------------------------------- |
| `GET /up`      | Liveness — returns 200 as long as the HTTP server is up. |
| `GET /setup`   | Configuration snapshot.                                  |
| `GET /workers` | Per-worker live status.                                  |
| `GET /tail`    | Recent log lines.                                        |