Dockerfile

# Use Python base image
FROM python:3.11-slim

# Set working directory
WORKDIR /app

# Install system dependencies including cuda-toolkit for bitsandbytes
RUN apt-get update && apt-get install -y \
    build-essential \
    curl \
    git \
    && rm -rf /var/lib/apt/lists/*

# Create necessary directories with proper permissions
RUN mkdir -p /app/logs /app/src/static /home/user/.cache/huggingface /home/user/.local /app/offload \
    && chmod -R 777 /app/logs /home/user/.cache/huggingface /home/user/.local /app/offload

# Create non-root user
RUN useradd -m -u 1000 user \
    && chown -R user:user /app /home/user/.cache /home/user/.local

# Set environment variables
ENV PYTHONPATH=/app
ENV PYTHONUNBUFFERED=1
ENV PYTHONDONTWRITEBYTECODE=1
ENV PORT=7860
ENV PATH="/home/user/.local/bin:${PATH}"
ENV HF_HOME=/home/user/.cache/huggingface
# Memory optimizations
ENV MALLOC_ARENA_MAX=2
ENV MALLOC_TRIM_THRESHOLD_=100000
ENV MALLOC_MMAP_THRESHOLD_=100000
# Model optimizations
ENV OMP_NUM_THREADS=1
ENV MKL_NUM_THREADS=1
ENV NUMEXPR_NUM_THREADS=1
# Ensure offline mode is disabled
ENV HF_HUB_OFFLINE=0
ENV TRANSFORMERS_OFFLINE=0

# Switch to non-root user
USER user

# Upgrade pip and install numpy first
RUN pip install --user --no-cache-dir --upgrade pip
RUN pip install --user --no-cache-dir "numpy<2.0.0"

# Copy requirements first to leverage Docker cache
COPY --chown=user:user requirements.txt .

# Install Python dependencies with memory optimizations
RUN pip install --user --no-cache-dir -r requirements.txt

# Copy application code
COPY --chown=user:user . .

# Expose port for Hugging Face Spaces
EXPOSE 7860

# Run the application with memory optimizations
CMD ["python", "-u", "-m", "uvicorn", "src.api:app", "--host", "0.0.0.0", "--port", "7860", "--log-level", "debug", "--workers", "1", "--limit-concurrency", "1", "--timeout-keep-alive", "120"]