docker-compose.prod.yml

# Production deployment — run with: docker compose -f docker-compose.prod.yml up -d
# Requires .env and .env.prod files — see .env.example for required variables.
# Images pulled from GHCR (built by GitHub Actions)
# GitHub repository: VitalPointAI/Bastion

services:
  # ==========================================
  # PostgreSQL Database with TimescaleDB
  # ==========================================
  postgres:
    image: timescale/timescaledb:latest-pg16
    container_name: bastion-postgres
    restart: always
    environment:
      POSTGRES_USER: postgres
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD}
      POSTGRES_DB: coalition_ops
    # NO host port mapping — only accessible on bastion-network
    # Databases must not be reachable from the internet
    volumes:
      - postgres_data:/var/lib/postgresql/data
      - ./backend/database/init.sql:/docker-entrypoint-initdb.d/01-init.sql
      - ./backend/database/schema.sql:/docker-entrypoint-initdb.d/02-schema.sql
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U postgres -d coalition_ops"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - bastion-network

  # ==========================================
  # Neo4j Graph Database
  # ==========================================
  neo4j:
    image: neo4j:2025-community
    container_name: bastion-neo4j
    restart: always
    environment:
      NEO4J_AUTH: neo4j/${NEO4J_PASSWORD}
      NEO4J_PLUGINS: '["apoc"]'
      # Tuned for CAX21 (8GB RAM shared with all services)
      # Reduced from dev (1g→512m) to leave headroom for PostgreSQL + Node.js
      NEO4J_server_memory_heap_initial__size: 256m
      NEO4J_server_memory_heap_max__size: 512m
      NEO4J_server_memory_pagecache_size: 256m
    # NO host port mapping — only accessible on bastion-network
    volumes:
      - neo4j_data:/data
      - neo4j_logs:/logs
    healthcheck:
      test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:7474 || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - bastion-network

  # ==========================================
  # Ironclaw PostgreSQL (pgvector) — isolated on ironclaw-network
  # ==========================================
  ironclaw-postgres:
    image: pgvector/pgvector:pg16
    container_name: bastion-ironclaw-postgres
    restart: always
    environment:
      POSTGRES_DB: ironclaw
      POSTGRES_USER: ironclaw
      POSTGRES_PASSWORD: ${IRONCLAW_DB_PASSWORD}
    volumes:
      - ironclaw_postgres_data:/var/lib/postgresql/data
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ironclaw -d ironclaw"]
      interval: 10s
      timeout: 5s
      retries: 5
    networks:
      - ironclaw-network

  # ==========================================
  # Ironclaw Agent Sidecar
  # ==========================================
  ironclaw:
    image: ghcr.io/vitalpointai/bastion/ironclaw:latest
    container_name: bastion-ironclaw
    # sleep infinity keeps stdin open so the repl channel doesn't EOF-shutdown.
    # --no-onboard skips the interactive setup wizard.
    entrypoint: ["/usr/local/bin/entrypoint.sh"]
    restart: unless-stopped
    environment:
      # LLM backend — Anthropic via OAuth (subscription only, no API key fallback)
      LLM_BACKEND: anthropic
      ANTHROPIC_OAUTH_TOKEN: ${ANTHROPIC_OAUTH_TOKEN:-}
      ANTHROPIC_MODEL: claude-haiku-4-5-20251001
      # Database (ironclaw's own pgvector instance)
      DATABASE_URL: postgresql://ironclaw:${IRONCLAW_DB_PASSWORD}@ironclaw-postgres:5432/ironclaw
      # HTTP webhook server
      HTTP_HOST: 0.0.0.0
      HTTP_PORT: "8080"
      HTTP_WEBHOOK_SECRET: ${IRONCLAW_SHARED_SECRET}
      # Secrets keychain master key (stable across restarts)
      IRONCLAW_SECRETS_MASTER_KEY: ${IRONCLAW_MASTER_KEY}
      # NEAR AI API key for embeddings
      NEARAI_API_KEY: ${NEARAI_API_KEY}
      # Agent settings
      AGENT_NAME: ironclaw
      AGENT_USE_PLANNING: "true"
      RUST_LOG: ironclaw=info
      # MCP server URL — entrypoint registers this with Ironclaw's REPL at startup
      MCP_BASTION_URL: http://bastion-mcp:3334/mcp
    volumes:
      - ironclaw-data:/home/ironclaw/.ironclaw
      - token-sync:/shared/tokens    # OAuth token file from backend
    depends_on:
      ironclaw-postgres:
        condition: service_healthy
    networks:
      # Ironclaw is on a separate network — NO access to postgres or neo4j.
      - ironclaw-network

  # ==========================================
  # SearXNG — Self-hosted meta-search engine for doc-intelligence researcher
  # ==========================================
  searxng:
    image: searxng/searxng:latest
    container_name: bastion-searxng
    restart: always
    volumes:
      - ./searxng/settings.yml:/etc/searxng/settings.yml:ro
      - ./searxng/limiter.toml:/etc/searxng/limiter.toml:ro
    environment:
      - SEARXNG_BASE_URL=http://searxng:8080/
    # NO host port mapping — only accessible on bastion-network
    networks:
      - bastion-network

  # ==========================================
  # BASTION MCP Server
  # Exposes BASTION tools to AI agents via Model Context Protocol (SSE transport)
  # ==========================================
  bastion-mcp:
    image: ghcr.io/vitalpointai/bastion/backend:latest
    container_name: bastion-mcp
    restart: always
    command: node dist/mcp/index.js
    # NO host port mapping — only accessible on internal networks
    environment:
      NODE_ENV: production
      DATABASE_URL: postgresql://postgres:${POSTGRES_PASSWORD}@postgres:5432/coalition_ops
      NEO4J_URI: bolt://neo4j:7687
      NEO4J_USER: neo4j
      NEO4J_PASSWORD: ${NEO4J_PASSWORD}
      MCP_PORT: "3334"
      SEARXNG_URL: http://searxng:8080
      TAVILY_API_KEY: ${TAVILY_API_KEY:-}
      ANTHROPIC_API_KEY: ${ANTHROPIC_API_KEY:-}
      OPENAI_API_KEY: ${OPENAI_API_KEY:-}
    healthcheck:
      test: ["CMD", "node", "-e", "fetch('http://localhost:3334/health').then(r => r.ok ? process.exit(0) : process.exit(1)).catch(() => process.exit(1))"]
      interval: 30s
      timeout: 10s
      start_period: 10s
      retries: 3
    depends_on:
      postgres:
        condition: service_healthy
    networks:
      - bastion-network
      - ironclaw-network

  # ==========================================
  # Backend API Server
  # ==========================================
  backend:
    image: ghcr.io/vitalpointai/bastion/backend:latest
    # NO build block — uses pre-built GHCR image pushed by GitHub Actions
    container_name: bastion-backend
    restart: always
    # NO host port mapping — only accessible via bastion-network
    # Frontend container's nginx proxies /api/ to http://backend:3001
    env_file:
      - .env.prod
    environment:
      NODE_ENV: production
      PORT: "3001"
      DATABASE_URL: postgresql://postgres:${POSTGRES_PASSWORD}@postgres:5432/coalition_ops
      NEO4J_URI: bolt://neo4j:7687
      NEO4J_USER: neo4j
      NEO4J_PASSWORD: ${NEO4J_PASSWORD}
      IRONCLAW_URL: http://ironclaw:8080
      # IRONCLAW_VERSION: auto-detected from health endpoint at startup
      IRONCLAW_SHARED_SECRET: ${IRONCLAW_SHARED_SECRET}
      DATABASE_URL_IRONCLAW: postgresql://ironclaw:${IRONCLAW_DB_PASSWORD}@ironclaw-postgres:5432/ironclaw
      MCP_BASTION_URL: http://bastion-mcp:3334/mcp
      SEARXNG_URL: http://searxng:8080
    depends_on:
      postgres:
        condition: service_healthy
      neo4j:
        condition: service_healthy
      ironclaw:
        condition: service_started
    healthcheck:
      test: ["CMD", "node", "-e", "fetch('http://localhost:3001/health').then(r => r.ok ? process.exit(0) : process.exit(1)).catch(() => process.exit(1))"]
      interval: 30s
      timeout: 10s
      start_period: 10s
      retries: 3
    group_add:
      - "988"  # Docker group GID — grants access to /var/run/docker.sock for Ironclaw self-update
    volumes:
      - token-sync:/shared/tokens    # Write refreshed OAuth tokens for Ironclaw
      - ${HOME}/.claude:/home/nodejs/.claude:ro  # Claude Code OAuth credentials (read by LLM factory)
      - /var/run/docker.sock:/var/run/docker.sock  # Docker socket for Ironclaw self-update
    networks:
      - bastion-network
      - ironclaw-network

  # ==========================================
  # Frontend (nginx:alpine serving static files)
  # ==========================================
  frontend:
    image: ghcr.io/vitalpointai/bastion/frontend:latest
    # NO build block — uses pre-built GHCR image pushed by GitHub Actions
    container_name: bastion-frontend
    restart: always
    ports:
      # Host nginx (port 443) proxies to this port
      # Container runs nginx:alpine serving pre-built static files on port 80
      - "8080:80"
    # NO source volume mounts — production serves pre-built static files from image
    depends_on:
      backend:
        condition: service_healthy
    healthcheck:
      test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://127.0.0.1/health || exit 1"]
      interval: 30s
      timeout: 10s
      start_period: 5s
      retries: 3
    networks:
      - bastion-network

# Named volumes for persistent data
volumes:
  postgres_data:
  neo4j_data:
  neo4j_logs:
  ironclaw_postgres_data:
  ironclaw-data:
  token-sync:       # Shared volume: backend writes OAuth tokens, Ironclaw reads them

# Internal Docker network — databases never exposed to host network
networks:
  bastion-network:
    driver: bridge
  # Isolated network for Ironclaw agent — only the backend can reach it.
  # Ironclaw has NO access to postgres, neo4j, or any other bastion-network service.
  ironclaw-network:
    driver: bridge