Deploy with Docker Compose

Got one beefy server and want AskAIs Customer Service running in 10 minutes? This is the path. Good for under ~500 concurrent visitors; for larger scale jump to the Kubernetes guide.

Prerequisites

• A Linux host with Docker 24+ and `docker compose` v2
• ≥ 2 vCPU, 4 GB RAM (8 GB recommended)
• A domain pointed at the host (for TLS via Caddy / Cloudflare / nginx)

One-line install

git clone https://github.com/singlink/singlink
cd singlink
APP_DOMAIN=chat.acme.com bash scripts/install.sh

The script:

1. Checks for docker + openssl
2. Generates .env.prod with strong random secrets (mode 600)
3. Builds the three images (web / ws / worker)
4. Brings the full stack up with docker compose -f docker-compose.prod.yml up -d
5. Runs prisma migrate deploy inside the web container

What runs

• postgres — pgvector PG16
• redis — pub/sub + queue (password-protected)
• minio — S3-compatible storage for attachments
• web / ws / worker — the three app processes
• nginx — reverse proxy on port 8080 with WS upgrade

Adding TLS

The bundled nginx terminates HTTP/1.1 on port 8080. Put a TLS-terminating proxy in front:

Caddy (easiest)

# /etc/caddy/Caddyfile
chat.acme.com {
  reverse_proxy localhost:8080
}

Cloudflare Tunnel (no inbound port at all)

cloudflared tunnel create singlink
cloudflared tunnel route dns singlink chat.acme.com
cloudflared tunnel run --url http://localhost:8080 singlink

Day-2 operations

# Logs
docker compose -f docker-compose.prod.yml logs -f

# Restart everything
docker compose -f docker-compose.prod.yml restart

# Stop (data persists in named volumes)
docker compose -f docker-compose.prod.yml down

# Upgrade
git pull
docker compose -f docker-compose.prod.yml pull
docker compose -f docker-compose.prod.yml up -d --build
docker compose -f docker-compose.prod.yml exec web pnpm db:deploy

# Backups (run from cron daily)
docker compose -f docker-compose.prod.yml exec -T postgres \
  pg_dump -U singlink singlink | gzip > /backups/singlink-$(date +%F).sql.gz

Scaling out workers

Run more AI worker replicas (Redis BullMQ handles distribution):

docker compose -f docker-compose.prod.yml up -d --scale worker=4

When to graduate to Kubernetes

• You need HA postgres / redis (managed RDS / ElastiCache)
• You need autoscaling under traffic spikes
• You're running it across multiple regions
• Single-host downtime is unacceptable for your SLA

Then jump to Kubernetes / Helm.