Dockerfile

FROM python:3.9-slim
WORKDIR /app

# System deps: audio libs, phonemizer backend, compiler toolchain, healthcheck
RUN apt-get update && apt-get install -y \
    libsndfile1 \
    espeak-ng \
    espeak-ng-data \
    build-essential \
    python3-dev \
    curl \
    ffmpeg \
    && rm -rf /var/lib/apt/lists/*

# Modern pip for reliable resolution
RUN pip install --no-cache-dir --upgrade pip

# Force CPU-only PyTorch first to avoid CUDA wheels and heavy resolution
# (keeps downstream deps like spacy-curated-transformers from pulling GPU builds)
RUN pip install --no-cache-dir --extra-index-url https://download.pytorch.org/whl/cpu "torch==2.3.0+cpu"

# Prefer wheels to avoid compiling spaCy/thinc/blis from source on slim images
ENV PIP_ONLY_BINARY=:all: PIP_PREFER_BINARY=1

# Python deps
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# App code
COPY app.py .

# Ensure phonemizer finds eSpeak NG on Debian
ENV PHONEMIZER_ESPEAK_LIBRARY=/usr/lib/x86_64-linux-gnu/libespeak-ng.so.1

# IMPORTANT: avoid model prefetch during build; load at runtime instead
# (prevents HF Hub/network hiccups from failing the build)
# RUN python -c "from kittentts import KittenTTS; KittenTTS('KittenML/kitten-tts-nano-0.1')"

# Non-root
RUN useradd -m -u 1000 appuser && chown -R appuser /app
USER appuser

# Expose and healthcheck on 7860 to match the uvicorn port below
EXPOSE 7860
HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
  CMD curl -f http://localhost:7860/ || exit 1

CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]