Dockerfile

# 1. Start from a standard Python 3.10 image
FROM python:3.10-slim

# 2. Set env to non-interactive to prevent build hangs
ENV DEBIAN_FRONTEND=noninteractive

# 3. Install build tools for llama-cpp-python
# This fixes the "Failed building wheel" error.
RUN apt-get update && apt-get install -y --no-install-recommends \
    build-essential \
    cmake \
    && rm -rf /var/lib/apt/lists/*

# 4. Set up work directory
WORKDIR /code

# 5. Copy *only* requirements and install
# This caches the install layer, speeding up future builds
COPY ./requirements.txt /code/requirements.txt
RUN pip install --no-cache-dir --upgrade pip
RUN pip install --no-cache-dir -r /code/requirements.txt

# 6. Copy the rest of the app code
COPY ./app.py /code/app.py

# 7. Expose the port
EXPOSE 7860

# 8. Run the app
# The app.py script will now handle the model download on startup.
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]