Dockerfile

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

WORKDIR /app

RUN apt update && apt install -y python3 python3-pip python3-venv
RUN python3 -m venv /opt/venv
ENV PATH="/opt/venv/bin:$PATH"


RUN pip install  --no-cache-dir   huggingface_hub

RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    repo="unsloth/gemma-4-E4B-it-GGUF"; \
    hf_hub_download(repo_id=repo, filename="gemma-4-E4B-it-Q4_K_M.gguf", local_dir="/app"); \
    hf_hub_download(repo_id=repo, filename="mmproj-BF16.gguf", local_dir="/app")'

CMD ["-m", "/app/gemma-4-E4B-it-Q4_K_M.gguf", \
     "--mmproj", "/app/mmproj-BF16.gguf", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "--log-verbose", \
     "-t", "2", \
     "--cache-type-k", "q8_0", \
     "--cache-type-v", "iq4_nl", \
     "--top-k", "64", \
     "-rea", "off", \
     "-c", "128000"]