LangBot/deploy/prod/docker-compose.yml

services:
  postgres:
    image: pgvector/pgvector:pg17
    container_name: langbot-cloud-postgres
    restart: unless-stopped
    environment:
      POSTGRES_DB: langbot
      POSTGRES_USER: langbot_operator
      POSTGRES_PASSWORD: ${POSTGRES_OPERATOR_PASSWORD}
    volumes:
      - postgres-data:/var/lib/postgresql/data
    healthcheck:
      test: [CMD-SHELL, "pg_isready -U langbot_operator -d langbot"]
      interval: 5s
      timeout: 5s
      retries: 30
    networks: [internal]

  redis:
    image: redis:7.4-alpine
    container_name: langbot-cloud-redis
    restart: unless-stopped
    command: [redis-server, --appendonly, "yes", --requirepass, "${REDIS_PASSWORD}"]
    volumes:
      - redis-data:/data
    healthcheck:
      test: [CMD-SHELL, "redis-cli -a \"$${REDIS_PASSWORD}\" ping | grep PONG"]
      interval: 5s
      timeout: 5s
      retries: 20
    environment:
      REDIS_PASSWORD: ${REDIS_PASSWORD}
    networks: [internal]

  migrate:
    image: rockchin/langbot-cloud-core:${LANGBOT_IMAGE_TAG}
    profiles: [tools]
    command: [uv, run, langbot, migrate, --cloud]
    environment: &core-env
      TZ: Asia/Shanghai
      SYSTEM__INSTANCE_ID: ${CLOUD_V2_INSTANCE_UUID}
      SYSTEM__EDITION: cloud
      SYSTEM__RECOVERY_KEY: ${SYSTEM_RECOVERY_KEY}
      SYSTEM__JWT__SECRET: ${JWT_SECRET}
      SYSTEM__LIMITATION__MAX_BOTS: "2"
      SYSTEM__LIMITATION__MAX_PIPELINES: "3"
      SYSTEM__LIMITATION__MAX_EXTENSIONS: "3"
      SYSTEM__LIMITATION__MAX_KNOWLEDGE_BASES: "2"
      API__WEBHOOK_PREFIX: https://cloud.langbot.app
      API__WEBUI_URL: https://cloud.langbot.app
      WORKSPACE__INVITATIONS__PUBLIC_WEB_URL: https://cloud.langbot.app
      DATABASE__USE: postgresql
      DATABASE__POSTGRESQL__URL: postgresql+asyncpg://langbot_runtime:${POSTGRES_RUNTIME_PASSWORD}@postgres:5432/langbot
      DATABASE__CLOUD_MIGRATION__OPERATOR_DSN_ENV: LANGBOT_CLOUD_MIGRATION_DSN
      LANGBOT_CLOUD_MIGRATION_DSN: postgresql://langbot_operator:${POSTGRES_OPERATOR_PASSWORD}@postgres:5432/langbot
      VDB__USE: pgvector
      VDB__PGVECTOR__USE_BUSINESS_DATABASE: "true"
      VDB__PGVECTOR__ALLOWED_DIMENSIONS: "384,512,768,1024,1536"
      PLUGIN__ENABLE: "true"
      PLUGIN__RUNTIME_WS_URL: ws://plugin-runtime:5400/control/ws
      PLUGIN__DISPLAY_PLUGIN_DEBUG_URL: wss://cloud.langbot.app/plugin/debug/ws
      PLUGIN__WORKER__MAX_CPUS: "0.25"
      PLUGIN__WORKER__MAX_MEMORY_MB: "256"
      PLUGIN__WORKER__MAX_PIDS: "128"
      PLUGIN__WORKER__MAX_WORKERS: "16"
      PLUGIN__WORKER__MAX_TOTAL_CPUS: "4.0"
      PLUGIN__WORKER__MAX_TOTAL_MEMORY_MB: "4096"
      PLUGIN__WORKER__REQUIRE_HARD_LIMITS: "true"
      LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN: ${PLUGIN_RUNTIME_CONTROL_TOKEN}
      # Cloud v2 currently grants no managed Box capability. Keep the shared
      # runtime deployed but disable Core integration until a hard-quota-capable
      # backend can satisfy the fail-closed Cloud readiness contract.
      BOX__ENABLED: "false"
      BOX__BACKEND: nsjail
      BOX__RUNTIME__ENDPOINT: ws://box:5410
      BOX__ADMISSION__REQUIRED: "true"
      BOX__ADMISSION__LOGICAL_SESSION_ID: global
      BOX__ADMISSION__REQUIRED_BACKEND: nsjail
      BOX__ADMISSION__MAX_SESSIONS: "1"
      BOX__ADMISSION__MAX_MANAGED_PROCESSES: "0"
      BOX__ADMISSION__CPUS: "0.25"
      BOX__ADMISSION__MEMORY_MB: "256"
      BOX__ADMISSION__WORKSPACE_QUOTA_MB: "256"
      BOX__LOCAL__HOST_ROOT: /app/data/box
      BOX__LOCAL__DEFAULT_WORKSPACE: /app/data/box
      BOX__LOCAL__ALLOWED_MOUNT_ROOTS: /app/data/box
      LANGBOT_BOX_CONTROL_TOKEN: ${BOX_CONTROL_TOKEN}
      MCP__STDIO__ENABLED: "false"
      LANGBOT_SPACE_CONTROL_PLANE_URL: https://space.langbot.app
      LANGBOT_SPACE_CONTROL_PLANE_TOKEN: ${CLOUD_V2_CONTROL_PLANE_TOKEN}
      LANGBOT_TELEMETRY_INGEST_TOKEN: ${CLOUD_V2_CONTROL_PLANE_TOKEN}
      LANGBOT_SPACE_CONTROL_PLANE_PUBLIC_KEY: ${CLOUD_V2_MANIFEST_PUBLIC_KEY}
      LANGBOT_SPACE_CONTROL_PLANE_KEY_ID: ${CLOUD_V2_MANIFEST_KEY_ID}
      SPACE__URL: https://space.langbot.app
    depends_on:
      postgres: {condition: service_healthy}
    networks: [internal]

  plugin-runtime:
    image: rockchin/langbot:${LANGBOT_IMAGE_TAG}
    container_name: langbot-cloud-plugin-runtime
    restart: unless-stopped
    command: [uv, run, python, -m, langbot_plugin.cli.__init__, rt]
    environment:
      LANGBOT_PLUGIN_RUNTIME_CONTROL_TOKEN: ${PLUGIN_RUNTIME_CONTROL_TOKEN}
    volumes:
      - plugin-data:/app/data
      - /sys/fs/cgroup:/sys/fs/cgroup:rw
    cgroup: host
    privileged: true
    expose: ["5400"]
    networks: [internal]

  box:
    image: rockchin/langbot:${LANGBOT_IMAGE_TAG}
    container_name: langbot-cloud-box
    restart: unless-stopped
    command: [uv, run, lbp, box, --host, 0.0.0.0, --ws-control-port, "5410"]
    environment:
      LANGBOT_BOX_CONTROL_TOKEN: ${BOX_CONTROL_TOKEN}
      LANGBOT_BOX_ROOT: /app/data/box
    volumes:
      - box-data:/app/data/box
      - /sys/fs/cgroup:/sys/fs/cgroup:rw
    cgroup: host
    privileged: true
    expose: ["5410"]
    networks: [internal]

  core:
    image: rockchin/langbot-cloud-core:${LANGBOT_IMAGE_TAG}
    container_name: langbot-cloud-core
    restart: unless-stopped
    environment: *core-env
    volumes:
      - core-data:/app/data
      - box-data:/app/data/box
    depends_on:
      postgres: {condition: service_healthy}
      redis: {condition: service_healthy}
      plugin-runtime: {condition: service_started}
      box: {condition: service_started}
    expose: ["5300"]
    healthcheck:
      test: [CMD-SHELL, "python -c 'import urllib.request; urllib.request.urlopen(\"http://127.0.0.1:5300/healthz\", timeout=3)'" ]
      interval: 10s
      timeout: 5s
      retries: 30
      start_period: 30s
    networks: [internal, shared-network]

networks:
  internal:
  shared-network:
    external: true

volumes:
  postgres-data:
  redis-data:
  plugin-data:
  box-data:
  core-data: