FROM vllm/vllm-openai:latest

# Upgrade transformers for Qwen3.6 support (model config requires >= 4.57.1)
RUN pip install --no-cache-dir "transformers>=4.57.1"

# Copy serving entrypoint
COPY serve.sh /serve.sh
RUN chmod +x /serve.sh

EXPOSE 8000

ENTRYPOINT ["/serve.sh"]