docker-compose.yml

# Zakura — OSS quickstart（server + web + 独立 node）
#
# 默认 SQLite（持久化到命名卷），完成管理员初始化后 node 自动登记：
#   docker compose up -d
#
# 使用 Postgres（内置 pgvector 服务，需把 DATABASE_URL 指过去）：
#   DATABASE_URL=postgresql://zakura:zakura@postgres:5432/zakura \
#     docker compose --profile postgres up -d
#
# 本地源码构建：
#   docker compose -f docker-compose.yml -f docker-compose.build.yml up -d --build
#
# 健康检查：server 暴露 /livez（liveness）与 /readyz（readiness）。
# 迁移：默认 AUTO_MIGRATE=true（首启建表）；生产多副本请设 AUTO_MIGRATE=false，
#       并先用一次 release 任务应用迁移。

services:
  server:
    image: ${ZAKURA_SERVER_IMAGE:-sunwuyuan/zakura-server:latest}
    container_name: zakura-server
    ports:
      - "8787:8787"
    environment:
      LISTEN_ADDR: ":8787"
      # 默认 SQLite；改用 Postgres：postgresql://zakura:zakura@postgres:5432/zakura
      DATABASE_URL: ${DATABASE_URL:-file:/data/zakura.db}
      DATA_DIR: /data
      # 必填（>=32 字节）；quickstart 用占位默认值，部署时务必覆盖
      ZAKURA_SECRET: ${ZAKURA_SECRET:-zakura-local-dev-secret-change-me-0123456789abcdef}
      PUBLIC_BASE_URL: ${PUBLIC_BASE_URL:-http://localhost:8787}
      WEB_PUBLIC_URL: ${WEB_PUBLIC_URL:-http://localhost:3001}
      ZAKURA_EDITION: ${ZAKURA_EDITION:-oss}
      AUTO_MIGRATE: ${AUTO_MIGRATE:-true}
      ZAKURA_NODE_BOOTSTRAP_DIR: /node-bootstrap
      OTEL_EXPORTER_OTLP_LOGS_ENDPOINT: ${OTEL_EXPORTER_OTLP_LOGS_ENDPOINT:-}
      OTEL_EXPORTER_OTLP_HEADERS: ${OTEL_EXPORTER_OTLP_HEADERS:-}
    volumes:
      - zakura-data:/data
      - node-bootstrap:/node-bootstrap
    restart: unless-stopped
    extra_hosts:
      - "host.docker.internal:host-gateway"
    healthcheck:
      test: ["CMD", "curl", "-fsS", "http://127.0.0.1:8787/livez"]
      interval: 15s
      timeout: 3s
      retries: 5
      start_period: 60s
    depends_on:
      postgres:
        condition: service_healthy
        required: false
    networks:
      - zakura

  node:
    image: ${ZAKURA_NODE_IMAGE:-sunwuyuan/zakura-agent:latest}
    environment:
      ZAKURA_AGENT_SERVER: http://server:8787
      ZAKURA_AGENT_KIND: server
      ZAKURA_AGENT_STATE_DIR: /state
      ZAKURA_AGENT_DATA_DIR: /workspace-data
      ZAKURA_NODE_BOOTSTRAP_DIR: /node-bootstrap
      ZAKURA_WORKSPACE_VOLUME: ${COMPOSE_PROJECT_NAME:-recloud}_workspace-data
    volumes:
      - node-state:/state
      - workspace-data:/workspace-data
      - type: volume
        source: node-bootstrap
        target: /node-bootstrap
        read_only: true
        volume:
          nocopy: true
      # Docker Socket 赋予节点宿主管理权限；控制面不挂载此 Socket。
      - /var/run/docker.sock:/var/run/docker.sock
    depends_on:
      server:
        condition: service_healthy
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "zakura-agent", "-healthcheck"]
      interval: 20s
      timeout: 10s
      retries: 3
      start_period: 60s
    networks:
      - zakura

  web:
    image: ${ZAKURA_WEB_IMAGE:-sunwuyuan/zakura-web:latest}
    container_name: zakura-web
    ports:
      - "3001:3001"
    environment:
      PORT: 3001
      HOSTNAME: 0.0.0.0
    depends_on:
      server:
        condition: service_healthy
    restart: unless-stopped
    networks:
      - zakura

  postgres:
    profiles: ["postgres"]
    # 需带 pgvector（memory embedding / 迁移）
    image: ${POSTGRES_IMAGE:-pgvector/pgvector:pg16}
    container_name: zakura-postgres
    environment:
      POSTGRES_USER: zakura
      POSTGRES_PASSWORD: zakura
      POSTGRES_DB: zakura
    # 仅本机调试用；容器间走 zakura 网络
    ports:
      - "127.0.0.1:5432:5432"
    volumes:
      - postgres-data:/var/lib/postgresql/data
    networks:
      - zakura
    restart: unless-stopped
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U zakura -d zakura"]
      interval: 5s
      timeout: 5s
      retries: 10

  redis:
    # 预留：Go 后端当前不使用 Redis（会话/事件直接落 SQL），此项仅为兼容旧栈
    # 或未来演进保留，默认不启动：docker compose --profile redis up -d
    profiles: ["redis"]
    image: ${REDIS_IMAGE:-redis:7-alpine}
    container_name: zakura-redis
    ports:
      - "127.0.0.1:6379:6379"
    volumes:
      - redis-data:/data
    command: ["redis-server", "--appendonly", "yes"]
    networks:
      - zakura
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 5s
      timeout: 3s
      retries: 10

volumes:
  zakura-data:
  node-bootstrap:
  node-state:
  workspace-data:
    name: ${COMPOSE_PROJECT_NAME:-recloud}_workspace-data
  postgres-data:
  redis-data:

networks:
  zakura:
    name: zakura
    driver: bridge