Dockerfile

# Use the official vLLM image as our base
FROM vllm/vllm-openai:latest

# Set environment variables so vLLM binds to the correct HF port
ENV HOST="0.0.0.0"
ENV PORT="7860"

# HF Spaces runs as a non-root user. We need to tell vLLM to download 
# the model into a temporary folder where it has write permissions.
ENV HF_HOME="/tmp/.cache/huggingface"
RUN mkdir -p /tmp/.cache/huggingface && chmod 777 /tmp/.cache/huggingface

# Start the OpenAI-compatible vLLM server
ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]

# Pass the model configurations
CMD ["--model", "google/gemma-4-E4B-it", "--max-model-len", "4096", "--dtype", "half", "--enforce-eager", "--host", "0.0.0.0", "--port", "7860"]