Dockerfile

# Use the official prebuilt server image
FROM ghcr.io/ggml-org/llama.cpp:server

# Hugging Face Spaces port requirement
ENV HOST=0.0.0.0
ENV PORT=7860
EXPOSE 7860

# Pass ONLY the arguments via CMD. 
# The base image already knows where the binary is and will execute it.
CMD ["-hf", "unsloth/gemma-4-31B-it-GGUF:IQ4_XS", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "-c", "128000"]