Dockerfile

FROM nvidia/cuda:12.1.1-runtime-ubuntu22.04

ENV DEBIAN_FRONTEND=noninteractive \
    PYTHONDONTWRITEBYTECODE=1 \
    PYTHONUNBUFFERED=1

RUN apt-get update && apt-get install -y --no-install-recommends \
    python3 python3-pip python3-venv wget ca-certificates && \
    ln -sf /usr/bin/python3 /usr/bin/python && \
    rm -rf /var/lib/apt/lists/*

WORKDIR /app

# Pin versions known to work with Piper import layout
RUN pip install --no-cache-dir fastapi==0.115.0 uvicorn[standard]==0.30.6 \
    piper-tts==1.2.0 onnxruntime==1.18.1
# Try GPU provider; remain functional on CPU if driver not present at runtime
RUN pip install --no-cache-dir onnxruntime-gpu==1.18.1 || true

# Bake model assets at build time (no runtime downloads)
RUN mkdir -p models && \
    wget -O models/en_US-ljspeech-medium.onnx \
      https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/ljspeech/medium/en_US-ljspeech-medium.onnx && \
    wget -O models/en_US-ljspeech-medium.onnx.json \
      https://huggingface.co/rhasspy/piper-voices/resolve/main/en/en_US/ljspeech/medium/en_US-ljspeech-medium.onnx.json

COPY app.py /app/app.py

ENV PIPER_MODEL=models/en_US-ljspeech-medium.onnx

EXPOSE 7860

# Bind to 0.0.0.0:7860 for Spaces ingress; CPU path works when CUDA not available
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--no-access-log"]