FROM python:3.11-slim

WORKDIR /app

# Install llama-cpp-python (CPU, pre-built)
RUN pip install --no-cache-dir \
    llama-cpp-python \
    huggingface_hub \
    fastapi \
    uvicorn \
    pydantic

# Download Qwen 7B Q4 on build
RUN python -c "from huggingface_hub import hf_hub_download; hf_hub_download('Qwen/Qwen2.5-Coder-7B-Instruct-GGUF', 'qwen2.5-coder-7b-instruct-q4_k_m.gguf', local_dir='/app/models')"

COPY server.py .

EXPOSE 7860

CMD ["python", "server.py"]