Dockerfile

FROM python:3.9-slim

WORKDIR /app

# 1. Install only essential system dependencies
RUN apt-get update && \
    apt-get install -y --no-install-recommends \
    build-essential \
    && rm -rf /var/lib/apt/lists/*

# 2. Set cache directories (HF Spaces provides writable /tmp)
ENV TRANSFORMERS_CACHE=/tmp/model_cache \
    HF_HOME=/tmp/huggingface \
    XDG_CACHE_HOME=/tmp/xdg_cache

# 3. Create cache directories
RUN mkdir -p ${TRANSFORMERS_CACHE} ${HF_HOME} ${XDG_CACHE_HOME}

# 4. Copy requirements first (better layer caching)
COPY requirements.txt .
RUN pip install --no-cache-dir --upgrade pip && \
    pip install --no-cache-dir -r requirements.txt

# 5. Copy application code
COPY . .

# 6. Health check endpoint
HEALTHCHECK --interval=30s --timeout=30s --start-period=5s --retries=3 \
    CMD curl -f http://localhost:7860/health || exit 1

# 7. Run with optimized settings
CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]