docker-compose.ghcr.yml

# ── Pre-built images from GitHub Container Registry ───────────────────────────
#
# Use this file to run the full Knowledge Graph stack locally without cloning
# the source code or building any image.
#
# Images are published automatically on every CI-green merge to main:
#   ghcr.io/agent-engineering-studio/kg-api:latest
#   ghcr.io/agent-engineering-studio/kg-ui:latest
#   ghcr.io/agent-engineering-studio/kg-mcp:latest
#   ghcr.io/agent-engineering-studio/kg-agents:latest
#
# Quick start:
#   1. Copy .env.example → .env  and set your passwords / Ollama URL
#   2. Pull and start (CPU, no GPU):
#        docker compose -f docker-compose.ghcr.yml --profile cpu up -d
#      Or if Ollama is already running on the host:
#        docker compose -f docker-compose.ghcr.yml up -d
#   3. docker compose -f docker-compose.ghcr.yml exec ollama-cpu \
#        ollama pull llama3 && ollama pull nomic-embed-text
#      (skip if OLLAMA_BASE_URL points to a host-side Ollama)
#
# Profiles:
#   (none)  → infra + app services; Ollama expected on the host at :11434
#   cpu     → same + Ollama CPU container
#   gpu     → same + Ollama GPU container (requires NVIDIA Container Toolkit)
# ──────────────────────────────────────────────────────────────────────────────

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

  # ── Redis Stack (Vector Store + RedisInsight) ─────────────────────────────────
  redis:
    container_name: kg-redis
    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
    restart: unless-stopped

  # ── Ollama CPU-only ───────────────────────────────────────────────────────────
  # Use profile "cpu" on machines without an NVIDIA GPU.
  # Skip both ollama profiles if Ollama is already running on the host and set
  # OLLAMA_BASE_URL=http://host.docker.internal:11434 in your .env.
  ollama-cpu:
    container_name: kg-ollama-cpu
    profiles: [cpu]
    image: ollama/ollama:latest
    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: 60s
    restart: unless-stopped

  # ── Ollama GPU (NVIDIA) ───────────────────────────────────────────────────────
  # Use profile "gpu" on machines with an NVIDIA GPU.
  ollama-gpu:
    container_name: kg-ollama-gpu
    profiles: [gpu]
    image: ollama/ollama:latest
    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: 60s
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    restart: unless-stopped

  # ── API — pre-built image from GHCR ──────────────────────────────────────────
  api:
    container_name: kg-api
    image: ghcr.io/agent-engineering-studio/kg-api:latest
    ports:
      - "8000:8000"
    env_file:
      - path: .env
        required: false
    environment:
      NEO4J_URI: "bolt://neo4j:7687"
      NEO4J_USER: "${NEO4J_USER:-neo4j}"
      NEO4J_PASSWORD: "${NEO4J_PASSWORD:-password}"
      REDIS_URL: "redis://redis:6379"
      OLLAMA_BASE_URL: "${OLLAMA_BASE_URL:-http://host.docker.internal:11434}"
      OLLAMA_LLM_MODEL: "${OLLAMA_LLM_MODEL:-llama3}"
      OLLAMA_EMBEDDING_MODEL: "${OLLAMA_EMBEDDING_MODEL:-nomic-embed-text}"
    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
    restart: unless-stopped

  # ── UI — pre-built image from GHCR ───────────────────────────────────────────
  ui:
    container_name: kg-ui
    image: ghcr.io/agent-engineering-studio/kg-ui:latest
    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
    restart: unless-stopped

  # ── MCP Server — pre-built image from GHCR ───────────────────────────────────
  mcp:
    container_name: kg-mcp
    image: ghcr.io/agent-engineering-studio/kg-mcp:latest
    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
    restart: unless-stopped

  # ── Agents — pre-built image from GHCR ───────────────────────────────────────
  agents:
    container_name: kg-agents
    image: ghcr.io/agent-engineering-studio/kg-agents:latest
    ports:
      - "8002:8001"
    environment:
      KG_API_URL: "http://api:8000"
      KG_MCP_URL: "http://mcp:8080"
      OLLAMA_BASE_URL: "${OLLAMA_BASE_URL:-http://host.docker.internal:11434}"
      OLLAMA_LLM_MODEL: "${OLLAMA_LLM_MODEL:-llama3}"
      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
    restart: unless-stopped

  # ── RedisInsight ──────────────────────────────────────────────────────────────
  redisinsight:
    container_name: kg-redisinsight
    image: redis/redisinsight:latest
    ports:
      - "5540:5540"
    volumes:
      - redisinsight_data:/data
    depends_on:
      redis:
        condition: service_healthy
    restart: unless-stopped

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

networks:
  default:
    name: kg_network