Dockerfile

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

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

# Create a dedicated directory to organize the models
RUN mkdir -p /models

# --- MODEL 1: Gemma 4 Multimodal + Speculative Setup ---
RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf" -O /models/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf
RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mtp-gemma-4-E2B-it.gguf" -O /models/mtp-gemma-4-E2B-it.gguf
RUN wget "https://huggingface.co/unsloth/gemma-4-E2B-it-qat-GGUF/resolve/main/mmproj-F16.gguf" -O /models/gemma-4-E2B-it-mmproj.gguf

# --- MODEL 2: Gemma 3 1B Model (Corrected Direct Download Link) ---
RUN wget "https://huggingface.co/MaziyarPanahi/gemma-3-1b-it-GGUF/resolve/main/gemma-3-1b-it.Q3_K_M.gguf" -O /models/gemma-3-1b-it.Q3_K_M.gguf


# --- CREATE ROUTER CONFIGURATION ---
# Appending configuration step-by-step to avoid any multiline shell escaping issues
RUN echo "[gemma-4-vision]" > /models.ini && \
    echo "model = /models/gemma-4-E2B-it-qat-UD-Q4_K_XL.gguf" >> /models.ini && \
    echo "spec-draft-model = /models/mtp-gemma-4-E2B-it.gguf" >> /models.ini && \
    echo "mmproj = /models/gemma-4-E2B-it-mmproj.gguf" >> /models.ini && \
    echo "ctx-size = 4096" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "ubatch-size = 128" >> /models.ini && \
    echo "batch-size = 512" >> /models.ini && \
    echo "spec-type = draft-mtp" >> /models.ini && \
    echo "spec-draft-n-max = 3" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[gemma-3-1b]" >> /models.ini && \
    echo "model = /models/gemma-3-1b-it.Q3_K_M.gguf" >> /models.ini && \
    echo "ctx-size = 4096" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini


# --- START SERVER IN ROUTER MODE ---
CMD [ \
  "--models-preset", "/models.ini", \
  "--port", "7860", \
  "--host", "0.0.0.0", \
  "-t", "2", \
  "-tb", "2" \
]