M.I.M.I.R - Multi-agent Intelligent Memory & Insight Repository

# Build llama.cpp server for AMD64 with CUDA support and Qwen2.5-VL-7B # Supports vision-language models for image understanding with GPU acceleration FROM nvidia/cuda:12.3.1-devel-ubuntu22.04 AS builder # Install build dependencies RUN apt-get update && apt-get install -y \ build-essential \ cmake \ git \ curl \ libcurl4-openssl-dev \ && rm -rf /var/lib/apt/lists/* # Clone llama.cpp repository WORKDIR /build RUN git clone https://github.com/ggerganov/llama.cpp.git . # Build llama.cpp server with CUDA and multimodal support RUN mkdir build && cd build && \ cmake .. \ -DLLAMA_CUDA=ON \ -DLLAMA_CURL=ON \ -DLLAMA_BUILD_SERVER=ON \ -DBUILD_SHARED_LIBS=OFF \ -DCMAKE_BUILD_TYPE=Release && \ cmake --build . --config Release -j$(nproc) # Runtime stage with CUDA runtime FROM nvidia/cuda:12.3.1-runtime-ubuntu22.04 # Install runtime dependencies RUN apt-get update && apt-get install -y \ libcurl4 \ curl \ libgomp1 \ && rm -rf /var/lib/apt/lists/* # Copy built binaries COPY --from=builder /build/build/bin/llama-server /usr/local/bin/llama-server # Create models directory RUN mkdir -p /models # Copy Qwen2.5-VL-7B models COPY docker/llama-cpp/models/qwen2.5-vl-7b.gguf /models/qwen2.5-vl.gguf COPY docker/llama-cpp/models/qwen2.5-vl-7b-vision.gguf /models/qwen2.5-vl-vision.gguf # Expose port EXPOSE 8080 # Health check HEALTHCHECK --interval=30s --timeout=10s --start-period=60s --retries=3 \ CMD curl -f http://localhost:8080/health || exit 1 # Set working directory WORKDIR /app # Default command with Qwen2.5-VL-7B and GPU offloading # -ngl 99 = offload all layers to GPU ENTRYPOINT ["/usr/local/bin/llama-server"] CMD ["--host", "0.0.0.0", "--port", "8080", "--model", "/models/qwen2.5-vl.gguf", "--mmproj", "/models/qwen2.5-vl-vision.gguf", "--ctx-size", "131072", "--parallel", "4", "--threads", "-1", "-ngl", "99"]