Dockerfile.vllm

# vLLM-Omni Dockerfile for Qwen3-TTS
# Uses official vLLM-Omni image (requires Python 3.12)
# See: https://docs.vllm.ai/projects/vllm-omni/en/latest/getting_started/installation/gpu/

FROM vllm/vllm-omni:v0.14.0rc1

WORKDIR /app

# Fix stale GPG keys and install runtime deps for audio + optional mp3 conversion
RUN rm -rf /var/lib/apt/lists/* && \
    apt-get clean && \
    apt-get update --allow-releaseinfo-change -o Acquire::AllowInsecureRepositories=true -o Acquire::AllowDowngradeToInsecureRepositories=true 2>/dev/null || true && \
    apt-get install -y --allow-unauthenticated --no-install-recommends \
    ffmpeg \
    libsndfile1 \
    curl \
 && rm -rf /var/lib/apt/lists/*

# Install API dependencies
RUN pip install --no-cache-dir -U pip && \
    pip install --no-cache-dir \
    fastapi>=0.109.0 \
    uvicorn[standard]>=0.27.0 \
    python-multipart \
    pydantic>=2.0.0 \
    inflect \
    aiofiles \
    soundfile \
    numpy \
    scipy \
    pydub \
    librosa

# Install ninja and flash-attention 2 for optimized inference
RUN pip install --no-cache-dir ninja packaging wheel && \
    pip install --no-cache-dir flash-attn --no-build-isolation

# Copy application code
COPY . /app

# Install the package
RUN pip install --no-cache-dir -e .

# Environment configuration
ENV HOST=0.0.0.0
ENV PORT=8880
ENV WORKERS=1
ENV TTS_BACKEND=vllm_omni
ENV VLLM_WORKER_MULTIPROC_METHOD=spawn
ENV PYTHONUNBUFFERED=1

# Expose port
EXPOSE 8880

# Health check
HEALTHCHECK --interval=30s --timeout=10s --start-period=180s --retries=3 \
    CMD curl -f http://localhost:8880/health || exit 1

# Run the server
CMD ["python", "-m", "uvicorn", "api.main:app", "--host", "0.0.0.0", "--port", "8880"]