Dockerfile

FROM ghcr.io/ggml-org/llama.cpp:server

RUN apt update && apt install wget -y && rm -rf /var/lib/apt/lists/*

# 1. THE MAIN MODEL: Using Q4_K_M for balanced speed and performance
RUN wget "https://huggingface.co/openbmb/MiniCPM-V-4.6-Thinking-gguf/resolve/main/MiniCPM-V-4_6-Thinking-Q4_K_M.gguf" -O /model.gguf

# 2. THE PROJECTOR: Multimodal projector for vision capabilities
RUN wget "https://huggingface.co/openbmb/MiniCPM-V-4.6-Thinking-gguf/resolve/main/mmproj-model-f16.gguf" -O /mmproj.gguf

# Clear the entrypoint and use shell form CMD to silence all logs
ENTRYPOINT []
CMD /app/llama-server -m /model.gguf --mmproj /mmproj.gguf --port 7860 --host 0.0.0.0 -t 2 --mlock -c 4096 --log-disable > /dev/null 2>&1