Dockerfile

FROM python:3.12

#RUN pip install --no-cache-dir torch==2.2.2
#RUN pip install --no-cache-dir flash-attn

RUN apt-get update && \
    apt-get install -y \
      bash \
      git git-lfs \
      wget curl procps gnupg \
      build-essential cmake \
      htop vim nano && \
    rm -rf /var/lib/apt/lists/*

# NVIDIA CUDA Keyring installieren (offizielle Methode seit 2024)
RUN wget https://developer.download.nvidia.com/compute/cuda/repos/debian12/x86_64/cuda-keyring_1.1-1_all.deb && \
    dpkg -i cuda-keyring_1.1-1_all.deb && \
    apt-get update && \
    apt-get -y install cuda

# CUDA ENV-Variablen setzen
ENV PATH=/usr/local/cuda/bin:${PATH}
ENV LD_LIBRARY_PATH=/usr/local/cuda/lib64:${LD_LIBRARY_PATH}
ENV CUDAToolkit_ROOT=/usr/local/cuda
ENV CMAKE_ARGS="-DGGML_CUDA=on -DCMAKE_CUDA_ARCHITECTURES=86"

RUN useradd -m -u 1000 user
USER user
ENV PATH="/home/user/.local/bin:$PATH"
# ^ when run as `user`, pip installs executables there

WORKDIR /app

COPY --chown=user . /app

# Wichtig: Isolation deaktivieren für llama-cpp-python Build
RUN pip install --upgrade pip

# Manuell Build-Werkzeuge bereitstellen
RUN pip install --no-cache-dir \
    setuptools \
    wheel \
    scikit-build-core[pyproject]

# Optional: zuerst llama-cpp-python bauen (cachebar)
RUN pip install --no-cache-dir --no-build-isolation llama-cpp-python

# Danach: outetts (zieht llama-cpp-python nicht erneut)
RUN pip install --no-cache-dir --no-build-isolation outetts

RUN pip install --no-cache-dir -r requirements.txt

#CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", , "--ws", "auto", "--allow-websocket-origin", "*"]
CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]