# docker-compose.yml
# Used for "clone & run" deployment scenario
# Started by ./start.sh with auto-generated .env
# NOTE: Run ./start.sh first to populate .env with generated secrets

services:
  postgres:
    container_name: airweave-db
    image: postgres:16
    # Uses variables from .env with fallbacks
    environment:
      POSTGRES_DB: ${POSTGRES_DB:-airweave}
      POSTGRES_USER: ${POSTGRES_USER}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
    command:
      - "postgres"
      - "-c"
      - "max_connections=200"
      - "-c"
      - "shared_buffers=256MB"
      - "-c"
      - "effective_cache_size=1GB"
    ports:
      - "5432:5432"
    volumes:
      - postgres_data:/var/lib/postgresql/data
      - ./init-scripts:/docker-entrypoint-initdb.d:ro
    healthcheck:
      test: [ "CMD-SHELL", "pg_isready -U ${POSTGRES_USER}" ]
      interval: 5s
      timeout: 5s
      retries: 5

  redis:
    container_name: airweave-redis
    image: redis:7-alpine
    ports:
      - "6379:6379"
    volumes:
      - redis_data:/data
    healthcheck:
      test: [ "CMD", "redis-cli", "ping" ]
      interval: 5s
      timeout: 5s
      retries: 5

  backend:
    container_name: airweave-backend
    # Use pre-built image of latest release, or override with BACKEND_IMAGE env var for testing
    image: ${BACKEND_IMAGE:-ghcr.io/airweave-ai/airweave-backend:latest}
    # Import all variables from .env file
    env_file:
      - ../.env
    ports:
      - "8001:8001"
      - "9090:9090"
    volumes:
      - ../backend:/app
      - ../local_storage:/app/local_storage
    # Enable host.docker.internal on Linux (needed for webhook e2e tests)
    extra_hosts:
      - "host.docker.internal:host-gateway"
    # These variables override any from .env file
    # Required for container networking (services refer to container names, not localhost)
    environment:
      # Override any localhost references for container networking
      - POSTGRES_HOST=postgres
      - REDIS_HOST=redis
      - TEXT2VEC_INFERENCE_URL=http://text2vec-transformers:8080
      - VESPA_URL=http://vespa
      - VESPA_PORT=8081
      # Environment configuration (can be overridden in .env)
      - ENVIRONMENT=${ENVIRONMENT:-local}
      - LOCAL_DEVELOPMENT=${LOCAL_DEVELOPMENT:-true}
      # Temporal configuration
      - TEMPORAL_HOST=temporal
      - TEMPORAL_PORT=7233
      - TEMPORAL_NAMESPACE=default
      - AUTH_ENABLED=false
      # Svix configuration
      - SVIX_URL=http://svix:8071
      - SVIX_JWT_SECRET=${SVIX_JWT_SECRET}
      - SVIX_AUTH_TOKEN=${SVIX_AUTH_TOKEN:-}
      # Docling OCR (only active when DOCLING_BASE_URL is set in .env)
      - DOCLING_BASE_URL=${DOCLING_BASE_URL:+http://docling:5001}
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
      svix:
        condition: service_healthy
    healthcheck:
      test: [ "CMD", "curl", "-f", "http://localhost:8001/health/ready" ]
      interval: 10s
      timeout: 30s
      retries: 10
      start_period: 60s
    restart: on-failure

  frontend:
    container_name: airweave-frontend
    # Use pre-built image of latest release, or override with FRONTEND_IMAGE env var for testing
    image: ${FRONTEND_IMAGE:-ghcr.io/airweave-ai/airweave-frontend:latest}
    # Profile: only start if explicitly enabled with --profile frontend
    # Not needed for backend-only testing (API tests, connector tests)
    profiles:
      - frontend
    ports:
      - "${FRONTEND_LOCAL_DEVELOPMENT_PORT:-8080}:8080"
    environment:
      - API_URL=http://localhost:8001
      - ENABLE_AUTH=false
    depends_on:
      backend:
        condition: service_healthy
    healthcheck:
      test: [ "CMD", "wget", "--spider", "-q", "http://localhost:8080" ]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: on-failure

  connect:
    container_name: airweave-connect
    # Use pre-built image of latest release, or override with CONNECT_IMAGE env var for testing
    image: ${CONNECT_IMAGE:-ghcr.io/airweave-ai/airweave-connect:latest}
    # Profile: only start if explicitly enabled with --profile connect
    # Connect widget for embedded iframe integrations
    profiles:
      - connect
    ports:
      - "${CONNECT_LOCAL_DEVELOPMENT_PORT:-8082}:8082"
    environment:
      - API_URL=http://localhost:8001
      - PORT=8082
    depends_on:
      backend:
        condition: service_healthy
    healthcheck:
      test: [ "CMD", "wget", "--spider", "-q", "http://localhost:8082" ]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: on-failure

  text2vec-transformers:
    container_name: airweave-embeddings
    image: semitechnologies/transformers-inference:sentence-transformers-all-MiniLM-L6-v2
    # Profile: only start if explicitly enabled with --profile local-embeddings
    # By default, backend uses OpenAI embeddings when OPENAI_API_KEY is set
    profiles:
      - local-embeddings
    ports:
      - "9878:8080"
    environment:
      ENABLE_CUDA: 0
      WORKERS_PER_NODE: 1
    healthcheck:
      test: [ "CMD", "wget", "--spider", "-q", "http://localhost:8080/health" ]
      interval: 30s
      timeout: 10s
      retries: 3
    restart: on-failure

  temporal:
    container_name: airweave-temporal
    image: temporalio/auto-setup:1.24.2
    ports:
      - "7233:7233"
      - "8233:8233"
    environment:
      - DB=postgres12
      - DB_PORT=5432
      - POSTGRES_USER=${POSTGRES_USER}
      - POSTGRES_PWD=${POSTGRES_PASSWORD}
      - POSTGRES_SEEDS=postgres
      - DYNAMIC_CONFIG_FILE_PATH=config/dynamicconfig/development-sql.yaml
    volumes:
      - ./temporal-config:/etc/temporal/config/dynamicconfig
    depends_on:
      postgres:
        condition: service_healthy
    healthcheck:
      test: [ "CMD", "tctl", "--address", "temporal:7233", "workflow", "list" ]
      interval: 30s
      timeout: 10s
      retries: 5
    restart: on-failure

  temporal-init:
    container_name: airweave-temporal-init
    image: temporalio/admin-tools:1.27.2-tctl-1.18.2-cli-1.3.0
    depends_on:
      temporal:
        condition: service_healthy
    entrypoint: ["/bin/sh", "-c"]
    command:
      - |
        temporal operator search-attribute create \
          --name SyncId --type Keyword \
          --namespace default \
          --address temporal:7233 || true
        echo "SyncId search attribute registered"
    restart: "no"

  temporal-ui:
    container_name: airweave-temporal-ui
    image: temporalio/ui:2.26.2
    ports:
      - "8088:8080"
    environment:
      - TEMPORAL_ADDRESS=temporal:7233
      - TEMPORAL_CORS_ORIGINS=http://localhost:3000
    depends_on:
      temporal:
        condition: service_healthy
    restart: on-failure

  temporal-worker:
    container_name: airweave-temporal-worker
    # Use pre-built image of latest release, or override with BACKEND_IMAGE env var for testing
    image: ${BACKEND_IMAGE:-ghcr.io/airweave-ai/airweave-backend:latest}
    # Import all variables from .env file (same as backend)
    env_file:
      - ../.env
    environment:
      # Override any localhost references for container networking
      - POSTGRES_HOST=postgres
      - REDIS_HOST=redis
      - TEXT2VEC_INFERENCE_URL=http://text2vec-transformers:8080
      - VESPA_URL=http://vespa
      - VESPA_PORT=8081
      # Environment configuration (can be overridden in .env)
      - ENVIRONMENT=${ENVIRONMENT:-local}
      - LOCAL_DEVELOPMENT=${LOCAL_DEVELOPMENT:-true}
      # Temporal configuration
      - TEMPORAL_HOST=temporal
      - TEMPORAL_PORT=7233
      - TEMPORAL_NAMESPACE=default
      - TEMPORAL_TASK_QUEUE=airweave-sync-queue
      # Svix configuration
      - SVIX_URL=http://svix:8071
      - SVIX_JWT_SECRET=${SVIX_JWT_SECRET}
      - SVIX_AUTH_TOKEN=${SVIX_AUTH_TOKEN:-}
      # Docling OCR (only active when DOCLING_BASE_URL is set in .env)
      - DOCLING_BASE_URL=${DOCLING_BASE_URL:+http://docling:5001}
    volumes:
      - ../backend:/app
    entrypoint: [ "python", "-m", "airweave.domains.temporal.worker" ]
    depends_on:
      postgres:
        condition: service_healthy
      temporal:
        condition: service_healthy
      svix:
        condition: service_healthy
    restart: on-failure

  # Vespa single-node for development (config server + content + container combined)
  vespa:
    container_name: airweave-vespa
    image: vespaengine/vespa:8
    hostname: vespa
    ports:
      - "8081:8081" # Query/Document API
      - "19071:19071" # Config server (for deployment)
    volumes:
      - vespa_data:/opt/vespa/var
      - ../vespa/app:/app
    healthcheck:
      # Check config server is up (basic health for startup)
      test: [ "CMD", "curl", "-f", "http://localhost:19071/state/v1/health" ]
      interval: 10s
      timeout: 5s
      retries: 30
    restart: on-failure

  # Vespa init container - deploys the application package automatically
  # This runs once after vespa is healthy, then exits
  vespa-init:
    container_name: airweave-vespa-init
    image: alpine:3.19
    env_file:
      - ../.env
    volumes:
      - ../vespa/app:/app:ro
      - ../vespa/init-vespa.sh:/init-vespa.sh:ro
    entrypoint: [ "/bin/sh", "/init-vespa.sh" ]
    environment:
      - VESPA_CONFIG_SERVER=http://vespa:19071
    depends_on:
      vespa:
        condition: service_healthy
    restart: "no"

  docling:
    container_name: airweave-docling
    image: ghcr.io/docling-project/docling-serve
    profiles:
      - docling
    ports:
      - "5001:5001"
    healthcheck:
      test: [ "CMD", "curl", "-sf", "http://localhost:5001/health" ]
      interval: 10s
      timeout: 10s
      retries: 30
      start_period: 60s
    restart: on-failure

  svix:
    container_name: airweave-svix
    image: svix/svix-server
    ports:
      - "8071:8071"
    # Enable host.docker.internal on Linux (needed for webhook e2e tests —
    # Svix delivers webhooks to the test receiver running on the host)
    extra_hosts:
      - "host.docker.internal:host-gateway"
    environment:
      SVIX_WHITELIST_SUBNETS: "[0.0.0.0/0, ::/0]"
      WAIT_FOR: "true"
      SVIX_REDIS_DSN: "redis://redis:6379"
      SVIX_DB_DSN: "postgresql://${POSTGRES_USER}:${POSTGRES_PASSWORD}@postgres:5432/svix"
      SVIX_JWT_SECRET: ${SVIX_JWT_SECRET}
      SVIX_QUEUE_TYPE: "redis"
      SVIX_CACHE_TYPE: "none"         # "memory" cache causes stale endpoint lookups when messages arrive before endpoints exist
    depends_on:
      postgres:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test: [ "CMD-SHELL", "svix-server healthcheck http://localhost:8071" ]
      interval: 5s
      timeout: 5s
      retries: 30
    restart: on-failure

volumes:
  postgres_data:
  redis_data:
  vespa_data: