# Extends the official RunPod Serverless vLLM worker. No GPU Pod or llama.cpp.
# Build the pinned upstream worker first, as described on the guide.
FROM rakazo-runpod-vllm-base:local
RUN python3 -m pip install --no-cache-dir "vllm==0.31.0" vllm-gguf-plugin \
    && python3 -c "import vllm; assert vllm.__version__ == '0.31.0'"
COPY start_orca.py /src/start_orca.py
ENTRYPOINT ["python3", "/src/start_orca.py"]
