Dockerfile

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

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 -U pip huggingface_hub

# Model downloads
RUN python3 -c 'from huggingface_hub import hf_hub_download; \
    repo="HauhauCS/Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced"; \
    hf_hub_download(repo_id=repo, filename="Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf", local_dir="/app"); \
    hf_hub_download(repo_id=repo, filename="mmproj-Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf", local_dir="/app")'

# Blue-Purple CSS injection
RUN find / -name "index.html" 2>/dev/null | xargs -I {} sed -i 's|</head>|<style>body, html { background: linear-gradient(135deg, #0f172a, #2e1065) !important; color: #f3e8ff !important; } .chat-message, textarea, input, select { background-color: #1e1b4b !important; border: 1px solid #7e22ce !important; color: #ffffff !important; } button { background: linear-gradient(90deg, #4f46e5, #9333ea) !important; color: #ffffff !important; border: none !important; }</style></head>|g' {} || true

CMD ["--server", \
     "-m", "/app/Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-Q4_K_M.gguf", \
     "--mmproj", "/app/mmproj-Gemma4-12B-QAT-Uncensored-HauhauCS-Balanced-BF16.gguf", \
     "-r", "<start_of_turn>user", \
     "-r", "<end_of_turn>", \
     "--host", "0.0.0.0", \
     "--port", "7860", \
     "-t", "2", \
     "--cache-type-k", "q8_0", \
     "--cache-type-v", "iq4_nl", \
     "-c", "8192", \
     "-n", "4096"]