Dockerfile.gpu

# We use the lean NVIDIA CUDA runtime as the base so your GPU actually works
FROM nvidia/cuda:12.4.1-cudnn-runtime-ubuntu22.04

# Python environment variables
ENV PYTHONDONTWRITEBYTECODE=1
ENV PYTHONUNBUFFERED=1

# Magic trick: Copy the 'uv' binary directly from Astral's official image
COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/

WORKDIR /app

# Copy your dependency files first (for Docker layer caching)
COPY pyproject.toml uv.lock README.md .python-version ./

# uv will automatically download Python 3.12 (based on your pyproject.toml),
RUN uv venv

# and install dependencies into a standalone virtual environment at /app/.venv
RUN uv sync --locked --no-dev --no-install-project --extra gpu

# Copy your application code
COPY app ./app

# Run a final sync to install your actual project code
RUN uv sync --locked --no-dev --extra gpu

# Signal the app to use GPU logic
ENV USE_GPU=True

EXPOSE 8000

# Put the virtual environment on the PATH so it works natively
ENV PATH="/app/.venv/bin:$PATH"

# Run Uvicorn directly!
CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]