Dockerfile

# Instead of FROM python:3.12, use a slim base image
FROM python:3.12-slim

# Set the working directory
WORKDIR /app

# Copy requirements.txt (llama-cpp-python is already included in the base image)
COPY ./requirements.txt /app/requirements.txt

# Install only project-specific dependencies
RUN pip install --no-cache-dir -r requirements.txt

# Copy the entire application source code
COPY . /app

# Expose port for Hugging Face Spaces
EXPOSE 7860

# Run FastAPI with Gunicorn - increased timeout for model loading
CMD ["gunicorn", "app.main:app", \
    "-k", "uvicorn.workers.UvicornWorker", \
    "--bind", "0.0.0.0:7860", \
    "--workers", "1", \
    "--timeout", "600", \
    "--graceful-timeout", "600", \
    "--log-level", "info"]