docker-compose.yml

# Docker Compose for PocketTTS OpenAI-Compatible Server
#
# Usage:
#   docker compose up -d          # Start the server
#   docker compose logs -f        # View logs
#   docker compose down           # Stop the server
#
# Custom voices:
#   Mount your own voices directory to /app/voices

services:
  pockettts:
    build:
      context: .
      dockerfile: Dockerfile
    image: pockettts-openai-server:latest
    container_name: pockettts-server

    ports:
      - '${POCKET_TTS_PORT:-49112}:49112'

    environment:
      - POCKET_TTS_HOST=0.0.0.0
      - POCKET_TTS_PORT=49112
      - POCKET_TTS_VOICES_DIR=/app/voices
      - POCKET_TTS_LOG_LEVEL=${POCKET_TTS_LOG_LEVEL:-INFO}
      - POCKET_TTS_STREAM_DEFAULT=${POCKET_TTS_STREAM_DEFAULT:-true}
      # Model language (e.g., english, french_24l, german_24l, portuguese, italian, spanish_24l)
      # Mutually exclusive with POCKET_TTS_MODEL_PATH. Requires pocket-tts>=2.0.0.
      - POCKET_TTS_LANGUAGE=${POCKET_TTS_LANGUAGE:-}
      # Enable int8 quantization for lower memory usage and improved speed
      - POCKET_TTS_QUANTIZE=${POCKET_TTS_QUANTIZE:-false}
      - POCKET_TTS_VOICE_CACHE_DIR=/app/voice_cache
      # Hugging Face token for voice cloning (optional)
      - HF_TOKEN=${HF_TOKEN:-}

    volumes:
      # Mount custom voices (optional - overrides bundled voices)
      - ${POCKET_TTS_VOICES_DIR:-./voices}:/app/voices:ro
      # Persist logs
      - ./logs:/app/logs
      # Cache HuggingFace models to avoid re-downloading
      - pockettts-cache:/home/pockettts/.cache/huggingface
      # Writable cache for per-model cloned voice safetensors
      - pockettts-voice-cache:/app/voice_cache

    restart: unless-stopped

    # Resource limits (adjust based on your hardware)
    deploy:
      resources:
        limits:
          memory: 4G
        reservations:
          memory: 2G

    healthcheck:
      test:
        [
          'CMD',
          'python',
          '-c',
          "import urllib.request; urllib.request.urlopen('http://localhost:49112/health')",
        ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 120s # Model loading takes time

volumes:
  pockettts-cache:
    name: pockettts-huggingface-cache
  pockettts-voice-cache:
    name: pockettts-voice-cache