#Docker Compose Deployment

Docker Compose is the quickest way to try Luna DevOps on a personal server, in a test environment, or with a small team. It starts every required service together, so PostgreSQL and Redis do not need to be installed separately.

If you want the platform itself to run in Kubernetes, start with Kubernetes (Helm).

#Before You Start

You need:

• A machine that can run Docker.
• Docker Compose.
• Network access to pull DockerHub images.
• Host port 8088 available.

#Choose A Version

The repository root docker-compose.yaml pulls these images by default:

liteyukistudio/luna-devops:nightly
liteyukistudio/luna-worker:nightly
liteyukistudio/luna-agent:nightly (AI profile only)

To verify a specific release, set the image tag before starting:

DEVOPS_IMAGE_TAG=v0.1.0-rc.1 docker compose up -d

#Start

Prepare production settings first:

cp .env.example .env

Edit .env, set SECRET_ENCRYPTION_KEY to a stable random key, and replace the placeholders in BOOTSTRAP_TOKEN and REDIS_PASSWORD. Use URL-safe letters and digits for the Redis password. Compose passes it directly to the built-in Redis server and assembles the complete URI for API and Worker. The complete stack defaults to production mode and does not expose a fixed development administrator.

Run this from the repository root:

docker compose up -d

This starts PostgreSQL, password-protected Redis, API, and Worker. API completes database migrations first, and Compose starts Worker only after /healthz passes, so Worker cannot access a fresh schema too early. The API image already embeds the web console, so you do not need to start Vite separately. On the first visit, open /bootstrap and use the BOOTSTRAP_TOKEN from .env to create the first administrator, then rotate or remove that one-time token.

#Enable The AI Assistant

The AI Agent uses an explicit profile and does not start by default. Generate one stable internal root and append it to .env:

printf 'AI_INTERNAL_SECRET=%s\n' "$(openssl rand -hex 32)" >> .env

API and Agent derive independent service identity, context signing, callback authentication, delegation signing, and Run Grant encryption keys with HKDF-SHA256. Do not reuse SECRET_ENCRYPTION_KEY, and do not rotate this value while durable Runs are active because those Runs would no longer be recoverable.

Then start the profile:

AI_ASSISTANT_AVAILABLE=true docker compose --profile ai up -d

Configure the Provider, model, access rules, and quotas under Global Settings → AI Assistant. The Provider API key is stored by the platform Secret Store and does not belong in .env. For diagnostics, run docker compose --profile ai logs -f agent.

To build images from the current source tree:

docker compose -f docker-compose-build.yaml up -d --build
# Include the Agent:
AI_ASSISTANT_AVAILABLE=true docker compose -f docker-compose-build.yaml --profile ai up -d --build

#Open The Console

Visit:

http://localhost:8088

The default Compose stack only exposes API on host port 8088. PostgreSQL and Redis stay inside the container network and do not occupy host ports 5432 and 6379.

#Check Services

docker compose ps
docker compose logs -f api
docker compose logs -f worker

When API is healthy, the console opens in the browser. Worker must also be healthy for builds, deployments, and status sync to run. If the page opens but tasks never start, check the Worker logs first.

#Next

1. Open First Sign-in and create or sign in as an administrator.
2. Open Connect Cluster and Registry to prepare runtime and image storage.
3. Follow Deploy a Web Project to complete the first delivery path.

#Stop

docker compose down

To remove data as well:

docker compose down -v