glpi-neuroforge-mega/deployments/gpu-subagent/docker-compose.yml

name: neuroforge-gpu-subagent
services:
  ollama:
    image: ${OLLAMA_IMAGE:-ollama/ollama:0.33.2}
    restart: unless-stopped
    gpus: all
    environment:
      OLLAMA_HOST: 0.0.0.0:11434
      OLLAMA_KEEP_ALIVE: ${OLLAMA_KEEP_ALIVE:-10m}
    ports:
    - ${OLLAMA_BIND_ADDRESS:-0.0.0.0}:${OLLAMA_HOST_PORT:-11434}:11434
    volumes:
    - ollama-data:/root/.ollama
    security_opt:
    - no-new-privileges:true
    healthcheck:
      test:
      - CMD
      - ollama
      - list
      interval: 15s
      timeout: 5s
      retries: 20
      start_period: 15s
  gpu-subagent:
    image: git.send.nrw/sendnrw/glpi-neuroforge-mega-neuroforge-worker:${IMAGE_TAG:?Set
      immutable IMAGE_TAG}
    command:
    - -server
    - ${NEUROFORGE_MASTER_URL:?Set reachable NeuroForge master URL}
    - -id
    - ${NEUROFORGE_GPU_WORKER_ID:?Set GPU worker ID}
    restart: unless-stopped
    environment:
      NEUROFORGE_WORKER_TOKEN: ${NEUROFORGE_WORKER_TOKEN:?Set worker token shared
        with master}
      NEUROFORGE_WORKER_RESOURCE_CLASS: gpu
      NEUROFORGE_WORKER_CAPABILITIES: ${NEUROFORGE_GPU_WORKER_CAPABILITIES:-gpu,model.chat,model.embed}
      NEUROFORGE_WORKER_MAX_CONCURRENCY: ${NEUROFORGE_GPU_WORKER_CONCURRENCY:-1}
      NEUROFORGE_WORKER_HEARTBEAT_INTERVAL: ${NEUROFORGE_WORKER_HEARTBEAT_INTERVAL:-15s}
      NEUROFORGE_WORKER_OLLAMA_URL: http://ollama:11434
      NEUROFORGE_WORKER_OLLAMA_API_KEY: ${NEUROFORGE_WORKER_OLLAMA_API_KEY:-}
      OLLAMA_API_KEY: ${OLLAMA_API_KEY:-}
      NEUROFORGE_WORKER_OLLAMA_CHAT_MODEL: ${OLLAMA_MODEL:-gemma4}
      NEUROFORGE_WORKER_OLLAMA_EMBEDDING_MODEL: ${OLLAMA_EMBEDDING_MODEL:-embeddinggemma}
      NEUROFORGE_WORKER_OLLAMA_NUM_CTX: ${NEUROFORGE_OLLAMA_NUM_CTX:-8192}
      NEUROFORGE_WORKER_OLLAMA_KEEP_ALIVE: ${OLLAMA_KEEP_ALIVE:-10m}
    depends_on:
      ollama-model-chat-init:
        condition: service_completed_successfully
      ollama-model-embed-init:
        condition: service_completed_successfully
    read_only: true
    tmpfs:
    - /tmp:size=32m,mode=1777
    security_opt:
    - no-new-privileges:true
    cap_drop:
    - ALL
  node-exporter:
    profiles:
    - monitoring
    image: ${NODE_EXPORTER_IMAGE:-prom/node-exporter:v1.12.1}
    restart: unless-stopped
    command:
    - --path.procfs=/host/proc
    - --path.sysfs=/host/sys
    - --path.rootfs=/rootfs
    - --collector.filesystem.mount-points-exclude=^/(sys|proc|dev|host|etc)($$|/)
    ports:
    - ${EXPORTER_BIND_ADDRESS:-0.0.0.0}:${NODE_EXPORTER_HOST_PORT:-9100}:9100
    volumes:
    - /proc:/host/proc:ro
    - /sys:/host/sys:ro
    - /:/rootfs:ro,rslave
    read_only: true
    security_opt:
    - no-new-privileges:true
    cap_drop:
    - ALL
  dcgm-exporter:
    profiles:
    - monitoring
    image: ${DCGM_EXPORTER_IMAGE:-nvcr.io/nvidia/k8s/dcgm-exporter:4.6.0-4.8.3-distroless}
    restart: unless-stopped
    gpus: all
    cap_add:
    - SYS_ADMIN
    ports:
    - ${EXPORTER_BIND_ADDRESS:-0.0.0.0}:${DCGM_EXPORTER_HOST_PORT:-9400}:9400
    security_opt:
    - no-new-privileges:true
  ollama-model-chat-init:
    image: ${OLLAMA_IMAGE:-ollama/ollama:0.33.2}
    restart: 'no'
    command:
    - pull
    - ${OLLAMA_MODEL:-gemma4}
    environment:
      OLLAMA_HOST: http://ollama:11434
    depends_on:
      ollama:
        condition: service_healthy
    security_opt:
    - no-new-privileges:true
  ollama-model-embed-init:
    image: ${OLLAMA_IMAGE:-ollama/ollama:0.33.2}
    restart: 'no'
    command:
    - pull
    - ${OLLAMA_EMBEDDING_MODEL:-embeddinggemma}
    environment:
      OLLAMA_HOST: http://ollama:11434
    depends_on:
      ollama:
        condition: service_healthy
    security_opt:
    - no-new-privileges:true
volumes:
  ollama-data: null