docker-compose.yml

# ── Profiles ──────────────────────────────────────────────────────────────────
#
#   dev   → infrastructure + dev tools (neo4j + redis + redisinsight)
#   prod  → full production stack (infra + api + ui + mcp + agents)
#   gpu   → Ollama with NVIDIA GPU support (use alongside dev or prod)
#   cpu   → Ollama CPU-only, no GPU required (use alongside dev or prod)
#
# Quick reference (see Makefile):
#   docker compose --profile dev  up -d              # infra only
#   docker compose --profile prod up -d              # full stack (Ollama on host)
#   docker compose --profile prod --profile gpu up   # full stack + Ollama GPU
#   docker compose --profile prod --profile cpu up   # full stack + Ollama CPU
#
# ──────────────────────────────────────────────────────────────────────────────

services:
  # ── Neo4j (Graph DB) ────────────────────────────────────────────────────────
  neo4j:
    container_name: kg-neo4j
    profiles: [dev, prod]
    image: neo4j:5.18
    ports:
      - "7474:7474"
      - "7687:7687"
    environment:
      NEO4J_AUTH: "neo4j/${NEO4J_PASSWORD}"
      NEO4J_PLUGINS: '["apoc"]'
    volumes:
      - neo4j_data:/data
      - neo4j_logs:/logs
    healthcheck:
      test:
        [
          "CMD",
          "cypher-shell",
          "-u",
          "neo4j",
          "-p",
          "${NEO4J_PASSWORD}",
          "RETURN 1",
        ]
      interval: 30s
      timeout: 10s
      retries: 5

  # ── Redis Stack (Vector Store) ───────────────────────────────────────────────
  redis:
    container_name: kg-redis
    profiles: [dev, prod]
    image: redis/redis-stack:latest
    ports:
      - "6379:6379"
      - "8001:8001"
    volumes:
      - redis_data:/data
    healthcheck:
      test: ["CMD", "redis-cli", "ping"]
      interval: 15s
      timeout: 5s
      retries: 3

  # ── Ollama GPU (NVIDIA) ──────────────────────────────────────────────────────
  # Use profile "gpu" on machines with an NVIDIA GPU.
  # If Ollama is already running on the host, omit both ollama profiles and
  # set OLLAMA_BASE_URL=http://host.docker.internal:11434 in your .env instead.
  ollama-gpu:
    container_name: kg-ollama-gpu
    profiles: [gpu]
    build:
      context: ./infra/ollama
      dockerfile: Dockerfile
    environment:
      - OLLAMA_MODELS=${OLLAMA_MODELS:-nomic-embed-text qwen2.5:14b}
    ports:
      - "127.0.0.1:11434:11434"
    volumes:
      - ollama_models:/root/.ollama
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://localhost:11434/api/tags"]
      interval: 30s
      timeout: 10s
      retries: 10
      start_period: 120s
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  # ── Ollama CPU-only (no GPU required) ────────────────────────────────────────
  # Use profile "cpu" on machines without NVIDIA GPU.
  ollama-cpu:
    container_name: kg-ollama-cpu
    profiles: [cpu]
    build:
      context: ./infra/ollama
      dockerfile: Dockerfile
    environment:
      - OLLAMA_MODELS=${OLLAMA_MODELS:-nomic-embed-text qwen2.5:14b}
    ports:
      - "127.0.0.1:11434:11434"
    volumes:
      - ollama_models:/root/.ollama
    healthcheck:
      test: ["CMD", "curl", "-sf", "http://localhost:11434/api/tags"]
      interval: 30s
      timeout: 10s
      retries: 10
      start_period: 120s

  # ── API (FastAPI / Python) ───────────────────────────────────────────────────
  api:
    container_name: kg-api
    profiles: [prod]
    build:
      context: ./knowledge-graph-api
      dockerfile: Dockerfile
    ports:
      - "8000:8000"
    # OLLAMA_BASE_URL is read from .env:
    #   default prod  → http://host.docker.internal:11434  (Ollama on host)
    #   profile gpu   → http://ollama-gpu:11434
    #   profile cpu   → http://ollama-cpu:11434
    env_file: .env
    depends_on:
      neo4j:
        condition: service_healthy
      redis:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 30s
      timeout: 10s
      retries: 3

  # ── UI (Next.js) ─────────────────────────────────────────────────────────────
  ui:
    container_name: kg-ui
    profiles: [prod]
    build:
      context: ./knowledge-graph-ui
      dockerfile: Dockerfile
      args:
        - NEXT_PUBLIC_API_URL=${NEXT_PUBLIC_API_URL:-http://localhost:8000}
    ports:
      - "3000:3000"
    environment:
      - NEXT_PUBLIC_API_URL=${NEXT_PUBLIC_API_URL:-http://localhost:8000}
      - NEXT_PUBLIC_ENABLE_STREAMING=${NEXT_PUBLIC_ENABLE_STREAMING:-true}
      - NEXT_PUBLIC_ENABLE_GRAPH_VIEW=${NEXT_PUBLIC_ENABLE_GRAPH_VIEW:-true}
    depends_on:
      api:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "wget", "-q", "--spider", "http://localhost:3000"]
      interval: 30s
      timeout: 10s
      retries: 3

  # ── MCP Server ───────────────────────────────────────────────────────────────
  mcp:
    container_name: kg-mcp
    profiles: [prod]
    build:
      context: ./knowledge-graph-mcp
      dockerfile: Dockerfile
    ports:
      - "8080:8080"
    environment:
      - KG_API_URL=http://api:8000
      - MCP_TRANSPORT=sse
      - MCP_HOST=0.0.0.0
      - MCP_PORT=8080
      - KG_API_TIMEOUT=600
    depends_on:
      api:
        condition: service_healthy
    healthcheck:
      test: ["CMD-SHELL", "python3 -c \"import socket; socket.create_connection(('localhost', 8080), 3).close()\""]
      interval: 15s
      timeout: 10s
      retries: 5
      start_period: 30s

  # ── Agents (Multi-Agent Orchestration) ──────────────────────────────────────
  agents:
    container_name: kg-agents
    profiles: [prod]
    build:
      context: ./knowledge-graph-agents
      dockerfile: Dockerfile
    ports:
      - "8002:8001"
    environment:
      - KG_MCP_URL=http://mcp:8080
      - KG_API_URL=http://api:8000
      - OLLAMA_BASE_URL=${OLLAMA_BASE_URL:-http://host.docker.internal:11434}
      - OLLAMA_LLM_MODEL=${OLLAMA_LLM_MODEL:-qwen2.5:14b}
      - KG_API_TIMEOUT=60
    depends_on:
      mcp:
        condition: service_healthy
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8001/agents/health"]
      interval: 30s
      timeout: 10s
      retries: 3

  # ── RedisInsight ─────────────────────────────────────────────────────────────
  # Web UI for browsing Redis keys, data structures and vectors.
  # Access: http://localhost:5540
  # First-time setup: add connection → host: redis  port: 6379
  redisinsight:
    container_name: kg-redisinsight
    profiles: [dev, prod]
    image: redis/redisinsight:latest
    ports:
      - "5540:5540"
    volumes:
      - redisinsight_data:/data
    depends_on:
      redis:
        condition: service_healthy

volumes:
  neo4j_data:
  neo4j_logs:
  redis_data:
  ollama_models:
  redisinsight_data:

networks:
  default:
    name: kg_network