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 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 13: LiquidAI LFM2.5 8B A1B ---
RUN wget "https://huggingface.co/LiquidAI/LFM2.5-8B-A1B-GGUF/resolve/main/LFM2.5-8B-A1B-Q4_0.gguf?download=true" -O /models/LFM2.5-8B-A1B-Q4_0.gguf

# --- CREATE ROUTER CONFIGURATION ---
RUN 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 "[lfm2.5-8b-a1b]" >> /models.ini && \
    echo "model = /models/LFM2.5-8B-A1B-Q4_0.gguf" >> /models.ini && \
    echo "ctx-size = 32768" >> /models.ini && \
    echo "flash-attn = on" >> /models.ini && \
    echo "" >> /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" \
]