Dockerfile

FROM python:3.9-slim

# Install build essentials for llama-cpp-python
RUN apt-get update && apt-get install -y \
    build-essential \
    wget \
    && rm -rf /var/lib/apt/lists/*

# Set environment variables for llama-cpp optimizations
ENV CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS -DLLAMA_AVX=ON -DLLAMA_AVX2=ON -DLLAMA_F16C=ON -DLLAMA_FMA=ON"
ENV FORCE_CMAKE=1

RUN useradd -m -u 1000 user
USER user
ENV PATH="/home/user/.local/bin:$PATH"
WORKDIR /app

# Download model
RUN wget https://huggingface.co/cigol123/YugoGPT-Q4_0-GGUF/resolve/main/yugogpt-q4_0.gguf

COPY --chown=user ./requirements.txt requirements.txt
RUN pip install --no-cache-dir --upgrade -r requirements.txt

COPY --chown=user . /app
CMD ["python", "app.py"]