Dockerfile

# FROM ollama/ollama

# RUN mkdir -p /.ollama && chmod 777 /.ollama

# ENV OLLAMA_MAX_LOADED_MODELS=20 --OLLAMA_NUM_PARALLEL=20
# ENV OLLAMA_HOST "0.0.0.0:7860"

# CMD ["serve"]
# EXPOSE 7860

FROM python:3

RUN useradd -m -u 1000 user
USER user
ENV HOME=/home/user \
    PATH=/home/user/.local/bin:$PATH

COPY --chown=user . $HOME/LLM_API

WORKDIR $HOME/LLM_API

RUN mkdir $HOME/.cache

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

CMD ["gunicorn", "-w", "5", "-b", "0.0.0.0:7860","main:app"]