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 270M IT Q8_0 ---
RUN wget "https://huggingface.co/unsloth/gemma-3-270m-it-GGUF/resolve/main/gemma-3-270m-it-Q8_0.gguf" -O /models/gemma-3-270m-it-Q8_0.gguf

# --- MODEL 3: Gemma 3 1B Model ---
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

# --- MODEL 4: Ornith 1.0 9B Model ---
RUN wget "https://huggingface.co/deepreinforce-ai/Ornith-1.0-9B-GGUF/resolve/main/ornith-1.0-9b-Q4_K_M.gguf" -O /models/ornith-1.0-9b-Q4_K_M.gguf

# --- MODEL 5: Qwen 3.5 0.8B Multimodal (Vision) ---
RUN wget "https://huggingface.co/unsloth/Qwen3.5-0.8B-GGUF/resolve/main/Qwen3.5-0.8B-Q4_0.gguf" -O /models/Qwen3.5-0.8B-Q4_0.gguf
RUN wget "https://huggingface.co/unsloth/Qwen3.5-0.8B-GGUF/resolve/main/mmproj-BF16.gguf" -O /models/mmproj-BF16.gguf

# --- MODEL 6: GPT OSS 20B ---
RUN wget "https://huggingface.co/unsloth/gpt-oss-20b-GGUF/resolve/main/gpt-oss-20b-Q4_0.gguf" -O /models/gpt-oss-20b-Q4_0.gguf

# --- MODEL 7: Gemma 4 12B IT ---
RUN wget "https://huggingface.co/unsloth/gemma-4-12b-it-GGUF/resolve/main/gemma-4-12b-it-Q4_K_M.gguf" -O /models/gemma-4-12b-it-Q4_K_M.gguf

# --- MODEL 8: Gemma 4 E4B IT ---
RUN wget "https://huggingface.co/unsloth/gemma-4-E4B-it-GGUF/resolve/main/gemma-4-E4B-it-Q4_K_M.gguf" -O /models/gemma-4-E4B-it-Q4_K_M.gguf

# --- MODEL 9: MiniCPM5 1B ---
RUN wget "https://huggingface.co/openbmb/MiniCPM5-1B-GGUF/resolve/main/MiniCPM5-1B-Q4_K_M.gguf" -O /models/MiniCPM5-1B-Q4_K_M.gguf

# --- MODEL 10: Nanbeige4.1 3B ---
RUN wget "https://huggingface.co/Mungert/Nanbeige4.1-3B-GGUF/resolve/main/Nanbeige4.1-3B-q4_k_m.gguf?download=true" -O /models/Nanbeige4.1-3B-q4_k_m.gguf

# --- MODEL 11: TwIL-LM3 Q5_K_S ---
RUN wget "https://huggingface.co/holooo/TwIL-LM3-Q5_K_S-GGUF/resolve/main/twil-lm3-q5_k_s.gguf?download=true" -O /models/twil-lm3-q5_k_s.gguf

# --- CREATE ROUTER CONFIGURATION ---
RUN echo "[gemma-4-E2B-it-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 = 131072" >> /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 = 131072" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[gemma-3-270m]" >> /models.ini && \
    echo "model = /models/gemma-3-270m-it-Q8_0.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[ornith-9b]" >> /models.ini && \
    echo "model = /models/ornith-1.0-9b-Q4_K_M.gguf" >> /models.ini && \
    echo "ctx-size = 8192" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[qwen-3.5-0.8b-vision]" >> /models.ini && \
    echo "model = /models/Qwen3.5-0.8B-Q4_0.gguf" >> /models.ini && \
    echo "mmproj = /models/mmproj-BF16.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[gpt-oss-20b]" >> /models.ini && \
    echo "model = /models/gpt-oss-20b-Q4_0.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[gemma-4-12b]" >> /models.ini && \
    echo "model = /models/gemma-4-12b-it-Q4_K_M.gguf" >> /models.ini && \
    echo "ctx-size = 131072" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[gemma-4-E4B-it]" >> /models.ini && \
    echo "model = /models/gemma-4-E4B-it-Q4_K_M.gguf" >> /models.ini && \
    echo "ctx-size = 131072" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[minicpm5-1b]" >> /models.ini && \
    echo "model = /models/MiniCPM5-1B-Q4_K_M.gguf" >> /models.ini && \
    echo "ctx-size = 131072" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[nanbeige4.1-3b]" >> /models.ini && \
    echo "model = /models/Nanbeige4.1-3B-q4_k_m.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /models.ini && \
    echo "[twil-lm3]" >> /models.ini && \
    echo "model = /models/twil-lm3-q5_k_s.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini

# --- START SERVER IN ROUTER MODE WITH CORS, TOOLS & MCP ENABLED ---
CMD [ \
  "--models-preset", "/models.ini", \
  "--port", "7860", \
  "--host", "0.0.0.0", \
  "-t", "2", \
  "-tb", "2", \
  "--tools", "all", \
  "--jinja", \
  "--webui-mcp-proxy", \
  "--cors-origins", "https://we2app-v4.hf.space" \
]