one_click/docker-compose.all-local.yml

services:
  gateway:
    build:
      context: ..
      dockerfile: Dockerfile
    env_file:
      - .env.all-local
    environment:
      PUBLIC_MODEL_ID: glm-5.2-visual-runtime
      VISUAL_RUNTIME_MODE: local
      GLM_BASE_URL: http://glm52-vllm:8000/v1
      GLM_API_KEY: EMPTY
      GLM_MODEL: glm-5.2
      VISION_BASE_URL: http://vision-vllm:8000/v1
      VISION_API_KEY: EMPTY
      VISION_MODEL: qwen3-omni
      OCR_BASE_URL: http://ocr:8080
      DATABASE_URL: postgresql+psycopg://glm_vvr:glm_vvr@postgres:5432/glm_vvr
      OBJECT_STORE_ENDPOINT: http://minio:9000
      OBJECT_STORE_BUCKET: glm-vvr
      OBJECT_STORE_ACCESS_KEY: glm-vvr
      OBJECT_STORE_SECRET_KEY: glm-vvr-password
    ports:
      - "8000:8000"
    depends_on:
      - glm52-vllm
      - vision-vllm
      - ocr
      - postgres
      - minio

  glm52-vllm:
    image: vllm/vllm-openai:glm52
    command:
      - "${GLM_MODEL_PATH:-zai-org/GLM-5.2-FP8}"
      - --host
      - 0.0.0.0
      - --port
      - "8000"
      - --served-model-name
      - glm-5.2
      - --kv-cache-dtype
      - fp8
      - --tensor-parallel-size
      - "${GLM_TENSOR_PARALLEL_SIZE:-8}"
      - --speculative-config.method
      - mtp
      - --speculative-config.num_speculative_tokens
      - "5"
      - --tool-call-parser
      - glm47
      - --reasoning-parser
      - glm45
      - --enable-auto-tool-choice
    ports:
      - "8001:8000"
    ipc: host
    volumes:
      - "${HF_HOME:-${HOME}/.cache/huggingface}:/root/.cache/huggingface"
      - "../models:/models:ro"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  vision-vllm:
    image: vllm/vllm-openai:latest
    command:
      - "${QWEN_OMNI_MODEL_PATH:-Qwen/Qwen3-Omni-30B-A3B-Instruct}"
      - --omni
      - --host
      - 0.0.0.0
      - --port
      - "8000"
      - --served-model-name
      - qwen3-omni
      - --tensor-parallel-size
      - "${VISION_TENSOR_PARALLEL_SIZE:-2}"
      - --limit-mm-per-prompt
      - '{"image": 8, "video": 1, "audio": 1}'
    ports:
      - "8002:8000"
    ipc: host
    volumes:
      - "${HF_HOME:-${HOME}/.cache/huggingface}:/root/.cache/huggingface"
      - "../models:/models:ro"
      - visual-runtime-data:/data
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

  ocr:
    build:
      context: ./ocr
    ports:
      - "8081:8080"

  postgres:
    image: postgres:16
    environment:
      POSTGRES_DB: glm_vvr
      POSTGRES_USER: glm_vvr
      POSTGRES_PASSWORD: glm_vvr
    ports:
      - "5432:5432"
    volumes:
      - postgres-data:/var/lib/postgresql/data

  minio:
    image: minio/minio:latest
    command: server /data --console-address ":9001"
    environment:
      MINIO_ROOT_USER: glm-vvr
      MINIO_ROOT_PASSWORD: glm-vvr-password
    ports:
      - "9000:9000"
      - "9001:9001"
    volumes:
      - minio-data:/data

volumes:
  postgres-data:
  minio-data:
  visual-runtime-data: