examples/speech_planner/docker-compose.yml

name: voice-agent

services:
  riva-tts-magpie:
    image: nvcr.io/nim/nvidia/magpie-tts-multilingual:latest
    environment:
      - NGC_API_KEY=${NVIDIA_API_KEY}
      - NIM_HTTP_API_PORT=9000
      - NIM_GRPC_API_PORT=50051
    ports:
      - "19000:9000"
      - "50151:50051"
    volumes:
      - nim_cache:/opt/nim/.cache
    shm_size: 8GB
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['0']
              capabilities: [gpu]

  riva-asr-parakeet:
    image: nvcr.io/nvstaging/riva/riva-speech:main_decoder_partials.26684117-linux-x86_64
    environment:
      - NGC_API_KEY=${NVIDIA_API_KEY}
      - NIM_HTTP_API_PORT=9001
      - NIM_GRPC_API_PORT=50052
      - NIM_TAGS_SELECTOR=mode=str,vad=silero
    ports:
      - "19001:9001"
      - "50152:50052"
    volumes:
      - nim_cache:/opt/nim/.cache
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['1']
              capabilities: [gpu]

  nvidia-llm:
    image: nvcr.io/nim/meta/llama-3.1-8b-instruct:latest
    environment:
      - NGC_API_KEY=${NVIDIA_API_KEY}
      - NIM_HTTP_API_PORT=8000
    ports:
      - "18000:8000"
    volumes:
      - nim_cache:/opt/nim/.cache
    shm_size: 16GB
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['2']
              capabilities: [gpu]

  nvidia-slm:
    image: nvcr.io/nim/google/gemma-2-2b-instruct:1.4.0
    environment:
      - NGC_API_KEY=${NVIDIA_API_KEY}
      - NIM_HTTP_API_PORT=8000
    ports:
      - "8080:8000"
    volumes:
      - nim_cache:/opt/nim/.cache
    shm_size: 8GB
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ['3']
              capabilities: [gpu]

  python-app:
    build:
      context: ../../
      dockerfile: examples/speech_planner/Dockerfile
    
    ports:
      - "8100:8100"
    environment:
      - NVIDIA_API_KEY=${NVIDIA_API_KEY}

      - RIVA_ASR_URL=riva-asr-parakeet:50052
      - RIVA_TTS_URL=riva-tts-magpie:50051
      - NVIDIA_LLM_URL=http://nvidia-llm:8000/v1
      - NVIDIA_SLM_URL=http://nvidia-slm:8000/v1

      - RIVA_ASR_MODEL=parakeet-1.1b-en-US-asr-streaming-silero-vad-asr-bls-ensemble
      - RIVA_TTS_MODEL=magpie_tts_ensemble-Magpie-Multilingual
      - NVIDIA_LLM_MODEL=meta/llama-3.1-8b-instruct
      - NVIDIA_SLM_MODEL=google/gemma-2-2b-instruct

      - RIVA_ASR_LANGUAGE=en-US
      - RIVA_TTS_LANGUAGE=en-US
      - RIVA_TTS_VOICE_ID=Magpie-Multilingual.EN-US.Sofia 
      - ZERO_SHOT_AUDIO_PROMPT= # set this only if using a zero-shot TTS model with a custom audio prompt

      - STATIC_DIR=/app/static
    volumes:
      - ../static:/app/static
    depends_on:
      - riva-tts-magpie
      - riva-asr-parakeet
      - nvidia-llm
      - nvidia-slm

volumes:
  nim_cache: