Dockerfile

# 1. Base image
FROM python:3.10-slim

# 2. Noninteractive + timezone
ENV DEBIAN_FRONTEND=noninteractive \
    TZ=Etc/UTC \
    XDG_CACHE_HOME=/tmp/.cache \
    HF_HOME=/tmp/.cache/huggingface

# 3. Use bash strict mode
SHELL ["/bin/bash", "-euxo", "pipefail", "-c"]

# 4. Install system deps and set timezone
RUN apt-get update \
 && apt-get install -y --no-install-recommends \
      tzdata \
      git \
      pkg-config \
      cmake \
      build-essential \
      python3-opencv \
      build-essential \
      cmake \
      libopenblas-dev \
      libopenblas-dev \
 && ln -sf /usr/share/zoneinfo/$TZ /etc/localtime \
 && echo $TZ > /etc/timezone \
 && rm -rf /var/lib/apt/lists/*

# 5. Prepare cache dirs
RUN mkdir -p "$XDG_CACHE_HOME" "$HF_HOME" \
 && chmod -R a+rwX "$XDG_CACHE_HOME" "$HF_HOME"

# 6. Build llama-cpp-python with OpenBLAS
RUN git clone --depth 1 --recurse-submodules \
      https://github.com/abetlen/llama-cpp-python.git /tmp/llama-cpp-python \
 && cd /tmp/llama-cpp-python \
 && CMAKE_ARGS="-DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS" pip install . \
 && rm -rf /tmp/llama-cpp-python

# 7. (Optional) Your other Python deps
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# 8. Copy app code
COPY . .

# 9. Expose & run
EXPOSE 7860
CMD ["python3", "app.py"]