version: '3.8'

services:
  # 1. GLM-5.3 High-Throughput vLLM Core
  glm-engine:
    image: vllm/vllm-openai:v0.6.3
    container_name: glm53_inference_core
    restart: always
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 8
              capabilities: [gpu]
    environment:
      - CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
      - VLLM_WORKER_MULTIPROC_METHOD=spawn
    volumes:
      - /data/models/zhipuai/glm-5.3:/model_weights:ro
    command: >
      --model /model_weights
      --tensor-parallel-size 8
      --quantization fp8
      --max-model-len 1000000
      --gpu-memory-utilization 0.95
      --trust-remote-code
      --port 8000
      --enable-chunked-prefill
    ports:
      - "8000:8000"
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 10s
      timeout: 5s
      retries: 5

  # 2. Reverse Proxy & RBAC Gateway
  api-gateway:
    image: nginx:alpine
    container_name: glm53_api_gateway
    restart: always
    ports:
      - "443:443"
      - "80:80"
    volumes:
      - ./nginx.conf:/etc/nginx/nginx.conf:ro
      - ./ssl:/etc/nginx/ssl:ro
    depends_on:
      glm-engine:
        condition: service_healthy

  # 3. Prometheus Telemetry
  prometheus:
    image: prom/prometheus:latest
    container_name: glm53_prometheus
    restart: always
    volumes:
      - ./prometheus.yml:/etc/prometheus/prometheus.yml:ro
    ports:
      - "9090:9090"