stack/docker/docker-compose.build.yml

services:
  # ======================
  # PostgreSQL Database
  # ======================
  postgres:
    build:
      context: ./docker/postgres
      dockerfile: Dockerfile
    container_name: mosaic-postgres
    restart: unless-stopped
    environment:
      POSTGRES_USER: ${POSTGRES_USER:-mosaic}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-mosaic_dev_password}
      POSTGRES_DB: ${POSTGRES_DB:-mosaic}
      # Performance tuning
      POSTGRES_SHARED_BUFFERS: ${POSTGRES_SHARED_BUFFERS:-256MB}
      POSTGRES_EFFECTIVE_CACHE_SIZE: ${POSTGRES_EFFECTIVE_CACHE_SIZE:-1GB}
      POSTGRES_MAX_CONNECTIONS: ${POSTGRES_MAX_CONNECTIONS:-100}
    ports:
      - "${POSTGRES_PORT:-5432}:5432"
    volumes:
      - postgres_data:/var/lib/postgresql/data
      - ./docker/postgres/init-scripts:/docker-entrypoint-initdb.d:ro
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-mosaic} -d ${POSTGRES_DB:-mosaic}"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 30s
    networks:
      - mosaic-internal
    profiles:
      - database
      - full
    labels:
      - "com.mosaic.service=database"
      - "com.mosaic.description=PostgreSQL 17 with pgvector"

  # ======================
  # Valkey Cache
  # ======================
  valkey:
    image: valkey/valkey:8-alpine
    container_name: mosaic-valkey
    restart: unless-stopped
    command:
      - valkey-server
      - --maxmemory ${VALKEY_MAXMEMORY:-256mb}
      - --maxmemory-policy noeviction
      - --appendonly yes
    ports:
      - "${VALKEY_PORT:-6379}:6379"
    volumes:
      - valkey_data:/data
    healthcheck:
      test: ["CMD", "valkey-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    networks:
      - mosaic-internal
    profiles:
      - cache
      - full
    labels:
      - "com.mosaic.service=cache"
      - "com.mosaic.description=Valkey Redis-compatible cache"

  # ======================
  # Authentik PostgreSQL
  # ======================
  authentik-postgres:
    image: postgres:17.7-alpine3.22
    container_name: mosaic-authentik-postgres
    restart: unless-stopped
    environment:
      POSTGRES_USER: ${AUTHENTIK_POSTGRES_USER:-authentik}
      POSTGRES_PASSWORD: ${AUTHENTIK_POSTGRES_PASSWORD:-authentik_password}
      POSTGRES_DB: ${AUTHENTIK_POSTGRES_DB:-authentik}
    volumes:
      - authentik_postgres_data:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${AUTHENTIK_POSTGRES_USER:-authentik}"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 20s
    networks:
      - mosaic-internal
    profiles:
      - authentik
      - full
    labels:
      - "com.mosaic.service=auth-database"
      - "com.mosaic.description=Authentik PostgreSQL database"

  # ======================
  # Authentik Redis
  # ======================
  authentik-redis:
    image: valkey/valkey:8-alpine
    container_name: mosaic-authentik-redis
    restart: unless-stopped
    command: valkey-server --save 60 1 --loglevel warning
    volumes:
      - authentik_redis_data:/data
    healthcheck:
      test: ["CMD", "valkey-cli", "ping"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    networks:
      - mosaic-internal
    profiles:
      - authentik
      - full
    labels:
      - "com.mosaic.service=auth-cache"
      - "com.mosaic.description=Authentik Redis cache"

  # ======================
  # Authentik Server
  # ======================
  authentik-server:
    image: ghcr.io/goauthentik/server:2024.12.1
    container_name: mosaic-authentik-server
    restart: unless-stopped
    command: server
    environment:
      AUTHENTIK_SECRET_KEY: ${AUTHENTIK_SECRET_KEY:-change-this-to-a-random-secret}
      AUTHENTIK_ERROR_REPORTING__ENABLED: ${AUTHENTIK_ERROR_REPORTING:-false}
      AUTHENTIK_POSTGRESQL__HOST: authentik-postgres
      AUTHENTIK_POSTGRESQL__PORT: 5432
      AUTHENTIK_POSTGRESQL__NAME: ${AUTHENTIK_POSTGRES_DB:-authentik}
      AUTHENTIK_POSTGRESQL__USER: ${AUTHENTIK_POSTGRES_USER:-authentik}
      AUTHENTIK_POSTGRESQL__PASSWORD: ${AUTHENTIK_POSTGRES_PASSWORD:-authentik_password}
      AUTHENTIK_REDIS__HOST: authentik-redis
      AUTHENTIK_REDIS__PORT: 6379
      AUTHENTIK_BOOTSTRAP_PASSWORD: ${AUTHENTIK_BOOTSTRAP_PASSWORD:-admin}
      AUTHENTIK_BOOTSTRAP_EMAIL: ${AUTHENTIK_BOOTSTRAP_EMAIL:-admin@localhost}
      AUTHENTIK_COOKIE_DOMAIN: ${AUTHENTIK_COOKIE_DOMAIN:-.localhost}
    ports:
      - "${AUTHENTIK_PORT_HTTP:-9000}:9000"
      - "${AUTHENTIK_PORT_HTTPS:-9443}:9443"
    volumes:
      - authentik_media:/media
      - authentik_templates:/templates
    depends_on:
      authentik-postgres:
        condition: service_healthy
      authentik-redis:
        condition: service_healthy
    healthcheck:
      test:
        [
          "CMD",
          "wget",
          "--no-verbose",
          "--tries=1",
          "--spider",
          "http://localhost:9000/-/health/live/",
        ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 90s
    networks:
      - mosaic-internal
      - mosaic-public
    profiles:
      - authentik
      - full
    labels:
      - "com.mosaic.service=auth-server"
      - "com.mosaic.description=Authentik OIDC server"
      # Traefik labels (activated when TRAEFIK_MODE=bundled or upstream)
      - "traefik.enable=${TRAEFIK_ENABLE:-false}"
      - "traefik.http.routers.mosaic-auth.rule=Host(`${MOSAIC_AUTH_DOMAIN:-auth.mosaic.local}`)"
      - "traefik.http.routers.mosaic-auth.entrypoints=${TRAEFIK_ENTRYPOINT:-websecure}"
      - "traefik.http.routers.mosaic-auth.tls=${TRAEFIK_TLS_ENABLED:-true}"
      - "traefik.http.services.mosaic-auth.loadbalancer.server.port=9000"
      - "traefik.docker.network=${TRAEFIK_DOCKER_NETWORK:-mosaic-public}"
      # Let's Encrypt (if enabled)
      - "traefik.http.routers.mosaic-auth.tls.certresolver=${TRAEFIK_CERTRESOLVER:-}"

  # ======================
  # Authentik Worker
  # ======================
  authentik-worker:
    image: ghcr.io/goauthentik/server:2024.12.1
    container_name: mosaic-authentik-worker
    restart: unless-stopped
    command: worker
    environment:
      AUTHENTIK_SECRET_KEY: ${AUTHENTIK_SECRET_KEY:-change-this-to-a-random-secret}
      AUTHENTIK_ERROR_REPORTING__ENABLED: ${AUTHENTIK_ERROR_REPORTING:-false}
      AUTHENTIK_POSTGRESQL__HOST: authentik-postgres
      AUTHENTIK_POSTGRESQL__PORT: 5432
      AUTHENTIK_POSTGRESQL__NAME: ${AUTHENTIK_POSTGRES_DB:-authentik}
      AUTHENTIK_POSTGRESQL__USER: ${AUTHENTIK_POSTGRES_USER:-authentik}
      AUTHENTIK_POSTGRESQL__PASSWORD: ${AUTHENTIK_POSTGRES_PASSWORD:-authentik_password}
      AUTHENTIK_REDIS__HOST: authentik-redis
      AUTHENTIK_REDIS__PORT: 6379
    volumes:
      - authentik_media:/media
      - authentik_certs:/certs
      - authentik_templates:/templates
    depends_on:
      authentik-postgres:
        condition: service_healthy
      authentik-redis:
        condition: service_healthy
    networks:
      - mosaic-internal
    profiles:
      - authentik
      - full
    labels:
      - "com.mosaic.service=auth-worker"
      - "com.mosaic.description=Authentik background worker"

  # ======================
  # Ollama (Optional AI Service)
  # ======================
  ollama:
    image: ollama/ollama:latest
    container_name: mosaic-ollama
    restart: unless-stopped
    ports:
      - "${OLLAMA_PORT:-11434}:11434"
    volumes:
      - ollama_data:/root/.ollama
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:11434/api/tags"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 60s
    networks:
      - mosaic-internal
    profiles:
      - ollama
      - full
    labels:
      - "com.mosaic.service=ai"
      - "com.mosaic.description=Ollama LLM service"
    # Uncomment if you have GPU support
    # deploy:
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: 1
    #           capabilities: [gpu]

  # ======================
  # OpenBao Secrets Management (Optional)
  # ======================
  openbao:
    build:
      context: ./docker/openbao
      dockerfile: Dockerfile
    container_name: mosaic-openbao
    restart: unless-stopped
    user: root
    ports:
      - "127.0.0.1:${OPENBAO_PORT:-8200}:8200"
    volumes:
      - openbao_data:/openbao/data
      - openbao_init:/openbao/init
    environment:
      VAULT_ADDR: http://0.0.0.0:8200
      SKIP_SETCAP: "true"
    entrypoint: ["dumb-init", "--"]
    command: ["bao", "server", "-config=/openbao/config/config.hcl"]
    cap_add:
      - IPC_LOCK
    healthcheck:
      test: ["CMD-SHELL", "nc -z 127.0.0.1 8200 || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    networks:
      - mosaic-internal
    profiles:
      - openbao
      - full
    labels:
      - "com.mosaic.service=secrets"
      - "com.mosaic.description=OpenBao secrets management"

  openbao-init:
    build:
      context: ./docker/openbao
      dockerfile: Dockerfile
    container_name: mosaic-openbao-init
    restart: unless-stopped
    user: root
    volumes:
      - openbao_init:/openbao/init
    environment:
      VAULT_ADDR: http://openbao:8200
    command: ["/openbao/init.sh"]
    depends_on:
      openbao:
        condition: service_healthy
    networks:
      - mosaic-internal
    profiles:
      - openbao
      - full
    labels:
      - "com.mosaic.service=secrets-init"
      - "com.mosaic.description=OpenBao auto-initialization sidecar"

  # ======================
  # Traefik Reverse Proxy (Optional - Bundled Mode)
  # ======================
  # Enable with: COMPOSE_PROFILES=traefik-bundled or --profile traefik-bundled
  # Set TRAEFIK_MODE=bundled in .env
  traefik:
    image: traefik:v3.2
    container_name: mosaic-traefik
    restart: unless-stopped
    command:
      - "--configFile=/etc/traefik/traefik.yml"
    ports:
      - "${TRAEFIK_HTTP_PORT:-80}:80"
      - "${TRAEFIK_HTTPS_PORT:-443}:443"
      - "${TRAEFIK_DASHBOARD_PORT:-8080}:8080"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - ./docker/traefik/traefik.yml:/etc/traefik/traefik.yml:ro
      - ./docker/traefik/dynamic:/etc/traefik/dynamic:ro
      - traefik_letsencrypt:/letsencrypt
    environment:
      - TRAEFIK_ACME_EMAIL=${TRAEFIK_ACME_EMAIL:-}
    networks:
      - mosaic-public
    profiles:
      - traefik-bundled
      - full
    labels:
      - "com.mosaic.service=reverse-proxy"
      - "com.mosaic.description=Traefik reverse proxy and load balancer"
    healthcheck:
      test: ["CMD", "traefik", "healthcheck", "--ping"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 20s

  # ======================
  # Mosaic API
  # ======================
  api:
    build:
      context: .
      dockerfile: ./apps/api/Dockerfile
      args:
        - NODE_ENV=production
    container_name: mosaic-api
    restart: unless-stopped
    environment:
      NODE_ENV: production
      # API Configuration - PORT is what NestJS reads
      PORT: ${API_PORT:-3001}
      API_HOST: ${API_HOST:-0.0.0.0}
      # Database
      DATABASE_URL: postgresql://${POSTGRES_USER:-mosaic}:${POSTGRES_PASSWORD:-mosaic_dev_password}@postgres:5432/${POSTGRES_DB:-mosaic}
      # Cache
      VALKEY_URL: redis://valkey:6379
      # Authentication
      OIDC_ENABLED: ${OIDC_ENABLED:-false}
      OIDC_ISSUER: ${OIDC_ISSUER}
      OIDC_CLIENT_ID: ${OIDC_CLIENT_ID}
      OIDC_CLIENT_SECRET: ${OIDC_CLIENT_SECRET}
      OIDC_REDIRECT_URI: ${OIDC_REDIRECT_URI:-http://localhost:3001/auth/callback}
      # JWT
      JWT_SECRET: ${JWT_SECRET:-change-this-to-a-random-secret}
      JWT_EXPIRATION: ${JWT_EXPIRATION:-24h}
      # Better Auth
      BETTER_AUTH_SECRET: ${BETTER_AUTH_SECRET}
      # Security
      CSRF_SECRET: ${CSRF_SECRET}
      ENCRYPTION_KEY: ${ENCRYPTION_KEY}
      # Ollama (optional)
      OLLAMA_ENDPOINT: ${OLLAMA_ENDPOINT:-http://ollama:11434}
      # OpenBao (optional)
      OPENBAO_ADDR: ${OPENBAO_ADDR:-http://openbao:8200}
    volumes:
      - openbao_init:/openbao/init:ro
    ports:
      - "${API_PORT:-3001}:${API_PORT:-3001}"
    depends_on:
      postgres:
        condition: service_healthy
      valkey:
        condition: service_healthy
    healthcheck:
      test:
        [
          "CMD-SHELL",
          'node -e "require(''http'').get(''http://localhost:${API_PORT:-3001}/health'', (r) => {process.exit(r.statusCode === 200 ? 0 : 1)})"',
        ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s
    networks:
      - mosaic-internal
      - mosaic-public
    labels:
      - "com.mosaic.service=api"
      - "com.mosaic.description=Mosaic NestJS API"
      # Traefik labels (activated when TRAEFIK_MODE=bundled or upstream)
      - "traefik.enable=${TRAEFIK_ENABLE:-false}"
      - "traefik.http.routers.mosaic-api.rule=Host(`${MOSAIC_API_DOMAIN:-api.mosaic.local}`)"
      - "traefik.http.routers.mosaic-api.entrypoints=${TRAEFIK_ENTRYPOINT:-websecure}"
      - "traefik.http.routers.mosaic-api.tls=${TRAEFIK_TLS_ENABLED:-true}"
      - "traefik.http.services.mosaic-api.loadbalancer.server.port=${API_PORT:-3001}"
      - "traefik.docker.network=${TRAEFIK_DOCKER_NETWORK:-mosaic-public}"
      # Let's Encrypt (if enabled)
      - "traefik.http.routers.mosaic-api.tls.certresolver=${TRAEFIK_CERTRESOLVER:-}"

  # ======================
  # Mosaic Orchestrator
  # ======================
  orchestrator:
    build:
      context: .
      dockerfile: ./apps/orchestrator/Dockerfile
    container_name: mosaic-orchestrator
    restart: unless-stopped
    # Run as non-root user (node:node, UID 1000)
    user: "1000:1000"
    environment:
      NODE_ENV: production
      # Orchestrator Configuration
      ORCHESTRATOR_PORT: 3001
      # Valkey
      VALKEY_URL: redis://valkey:6379
      # Claude API
      CLAUDE_API_KEY: ${CLAUDE_API_KEY}
      # Docker
      DOCKER_SOCKET: /var/run/docker.sock
      # Git
      GIT_USER_NAME: "Mosaic Orchestrator"
      GIT_USER_EMAIL: "orchestrator@mosaicstack.dev"
      # Security
      KILLSWITCH_ENABLED: true
      SANDBOX_ENABLED: true
    ports:
      - "3002:3001"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock:ro
      - orchestrator_workspace:/workspace
    depends_on:
      valkey:
        condition: service_healthy
      api:
        condition: service_healthy
    healthcheck:
      test:
        ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:3001/health || exit 1"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s
    networks:
      - mosaic-internal
    # Security hardening
    security_opt:
      - no-new-privileges:true
    cap_drop:
      - ALL
    cap_add:
      - NET_BIND_SERVICE
    read_only: false # Cannot be read-only due to workspace writes
    tmpfs:
      - /tmp:noexec,nosuid,size=100m
    labels:
      - "com.mosaic.service=orchestrator"
      - "com.mosaic.description=Mosaic Agent Orchestrator"
      - "com.mosaic.security=hardened"
      - "com.mosaic.security.non-root=true"
      - "com.mosaic.security.capabilities=minimal"

  # ======================
  # Mosaic Web
  # ======================
  web:
    build:
      context: .
      dockerfile: ./apps/web/Dockerfile
      args:
        - NEXT_PUBLIC_API_URL=${NEXT_PUBLIC_API_URL:-http://localhost:3001}
    container_name: mosaic-web
    restart: unless-stopped
    environment:
      NODE_ENV: production
      PORT: ${WEB_PORT:-3000}
      NEXT_PUBLIC_API_URL: ${NEXT_PUBLIC_API_URL:-http://localhost:3001}
    ports:
      - "${WEB_PORT:-3000}:${WEB_PORT:-3000}"
    depends_on:
      api:
        condition: service_healthy
    healthcheck:
      test:
        [
          "CMD-SHELL",
          'node -e "require(''http'').get(''http://localhost:${WEB_PORT:-3000}'', (r) => {process.exit(r.statusCode === 200 ? 0 : 1)})"',
        ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s
    networks:
      - mosaic-public
    labels:
      - "com.mosaic.service=web"
      - "com.mosaic.description=Mosaic Next.js Web App"
      # Traefik labels (activated when TRAEFIK_MODE=bundled or upstream)
      - "traefik.enable=${TRAEFIK_ENABLE:-false}"
      - "traefik.http.routers.mosaic-web.rule=Host(`${MOSAIC_WEB_DOMAIN:-mosaic.local}`)"
      - "traefik.http.routers.mosaic-web.entrypoints=${TRAEFIK_ENTRYPOINT:-websecure}"
      - "traefik.http.routers.mosaic-web.tls=${TRAEFIK_TLS_ENABLED:-true}"
      - "traefik.http.services.mosaic-web.loadbalancer.server.port=${WEB_PORT:-3000}"
      - "traefik.docker.network=${TRAEFIK_DOCKER_NETWORK:-mosaic-public}"
      # Let's Encrypt (if enabled)
      - "traefik.http.routers.mosaic-web.tls.certresolver=${TRAEFIK_CERTRESOLVER:-}"

# ======================
# Volumes
# ======================
volumes:
  postgres_data:
    name: mosaic-postgres-data
    driver: local
  valkey_data:
    name: mosaic-valkey-data
    driver: local
  authentik_postgres_data:
    name: mosaic-authentik-postgres-data
    driver: local
  authentik_redis_data:
    name: mosaic-authentik-redis-data
    driver: local
  authentik_media:
    name: mosaic-authentik-media
    driver: local
  authentik_certs:
    name: mosaic-authentik-certs
    driver: local
  authentik_templates:
    name: mosaic-authentik-templates
    driver: local
  ollama_data:
    name: mosaic-ollama-data
    driver: local
  openbao_data:
    name: mosaic-openbao-data
    driver: local
  openbao_init:
    name: mosaic-openbao-init
    driver: local
  traefik_letsencrypt:
    name: mosaic-traefik-letsencrypt
    driver: local
  orchestrator_workspace:
    name: mosaic-orchestrator-workspace
    driver: local

# ======================
# Networks
# ======================
networks:
  # Internal network for database/cache isolation
  # Note: NOT marked as 'internal: true' because API needs external access
  # for Authentik OIDC and external Ollama services
  mosaic-internal:
    name: mosaic-internal
    driver: bridge
  # Public network for services that need external access
  mosaic-public:
    name: mosaic-public
    driver: bridge