# ODS — NVIDIA GPU Overlay (Core Services Only)
# ComfyUI GPU config moved to extensions/services/comfyui/compose.nvidia.yaml
# Whisper GPU config moved to extensions/services/whisper/compose.nvidia.yaml
# Use with: docker compose -f docker-compose.base.yml -f docker-compose.nvidia.yml up -d

services:
  llama-server:
    image: ${LLAMA_SERVER_IMAGE:-ghcr.io/ggml-org/llama.cpp:server-cuda-b9014}
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
        limits:
          cpus: '${LLAMA_CPU_LIMIT:-16.0}'
          memory: ${LLAMA_SERVER_MEMORY_LIMIT:-64G}

  dashboard-api:
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all  # nvidia-smi needs visibility into all GPUs for monitoring
              capabilities: [utility]

  open-webui:
    environment:
      AUDIO_STT_MODEL: "${AUDIO_STT_MODEL:-deepdml/faster-whisper-large-v3-turbo-ct2}"
      ENABLE_IMAGE_GENERATION: "${ENABLE_IMAGE_GENERATION:-true}"