Dockerfile

FROM python:3.11-alpine

WORKDIR /app

RUN apk add --no-cache libstdc++ libgomp

COPY requirements.txt .

RUN pip install --no-cache-dir -r requirements.txt \
    --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu \
    --only-binary llama-cpp-python

COPY . .

RUN mkdir -p /app/models && chmod -R 777 /app

EXPOSE 7860

CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]