services:
  wlk-gpu-sortformer:
    build:
      context: .
      dockerfile: Dockerfile
      args:
        EXTRAS: ${GPU_SORTFORMER_EXTRAS:-cu129,diarization-sortformer}
    image: wlk:gpu-sortformer
    gpus: all
    ports:
      - "8000:8000"
    volumes:
      - hf-cache:/root/.cache/huggingface/hub
      # - ${HF_TKN_FILE:-./token}:/root/.cache/huggingface/token:ro
    environment:
      - HF_TOKEN
    command: ["--model", "medium", "--diarization", "--pcm-input"]

  wlk-gpu-voxtral:
    build:
      context: .
      dockerfile: Dockerfile
      args:
        EXTRAS: ${GPU_VOXTRAL_EXTRAS:-cu129,voxtral-hf,translation}
    image: wlk:gpu-voxtral
    gpus: all
    ports:
      - "8001:8000"
    volumes:
      - hf-cache:/root/.cache/huggingface/hub
      # - ${HF_TKN_FILE:-./token}:/root/.cache/huggingface/token:ro
    environment:
      - HF_TOKEN
    command: ["--backend", "voxtral", "--pcm-input"]

  wlk-cpu:
    build:
      context: .
      dockerfile: Dockerfile.cpu
      args:
        EXTRAS: ${CPU_EXTRAS:-cpu,diarization-diart,translation}
    image: wlk:cpu
    ports:
      - "8000:8000"
    volumes:
      - hf-cache:/root/.cache/huggingface/hub
      # - ${HF_TKN_FILE:-./token}:/root/.cache/huggingface/token:ro
    environment:
      - HF_TOKEN

volumes:
  hf-cache: