| # Use an official Python runtime as a parent image | |
| FROM python:3.10-slim | |
| # Set the working directory in the container | |
| WORKDIR /code | |
| # --- FINAL FIX: Install system-level build tools --- | |
| # The 'slim' python image is minimal. Libraries like bitsandbytes and peft | |
| # need a C compiler (gcc) and other tools to build custom CUDA kernels. | |
| # 'build-essential' installs these required tools via the system package manager. | |
| RUN apt-get update && apt-get install -y build-essential | |
| # Create a writeable directory for caching models and other artifacts. | |
| RUN mkdir -p /code/data/.cache && \ | |
| chmod -R 777 /code/data | |
| # Redirect all cache directories to our writeable location. | |
| ENV HF_HOME="/code/data/.cache" | |
| ENV TRITON_CACHE_DIR="/code/data/.triton" | |
| # Copy and install all required python libraries | |
| COPY ./requirements.txt /code/requirements.txt | |
| RUN pip install --no-cache-dir --upgrade -r /code/requirements.txt | |
| # Copy the main application file | |
| COPY ./app.py /code/app.py | |
| # Command to run the app | |
| CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"] |