Building _AugmentedIntelligence on Ubuntu Linux

This guide covers the Linux port of _AugmentedIntelligence: a CMake-based, console-first build for Ubuntu 22.04+ (and similar Debian-based distros).

The Windows Visual Studio project remains the primary, fully featured build (Dear ImGui + DirectX 11 GUI, Adobe/Microsoft integrations, etc.). The Linux port targets:

• x86_64 with AVX2 matrix acceleration (AVX-512 optional per CPU)
• CPU-only llama.cpp by default (optional CUDA build)
• Console REPL and web dashboard (web server start) instead of the Win32 GUI
• Excludes Windows-only Office/Adobe module sources

What works on Linux

Prerequisites

• Ubuntu 22.04 or 24.04 (64-bit x86)
• ~16 GB RAM recommended for linking the full binary
• gcc 11+ or clang 14+
• Python 3 (generates source list from .vcxproj)

Step 1 — Install dependencies

cd /path/to/_AugmentedIntelligence_v11.0\ -\ Typed\ Commands
chmod +x scripts/install_ubuntu_deps.sh scripts/build_ubuntu.sh scripts/build_llama_cpp_cpu.sh
./scripts/install_ubuntu_deps.sh

Step 2 — Build llama.cpp (CPU, AVX2)

./scripts/build_llama_cpp_cpu.sh avx2

Or manually:

export AI_LLAMA_SRC=$HOME/_AugmentedIntelligence/_src/llama.cpp
export AI_LLAMA_BUILD=$HOME/_AugmentedIntelligence/_src/llama_build_cpu_avx2
cmake -S "$AI_LLAMA_SRC" -B "$AI_LLAMA_BUILD" -DCMAKE_BUILD_TYPE=Release \
  -DGGML_CUDA=OFF -DGGML_AVX2=ON -DGGML_AVX=ON -DGGML_FMA=ON
cmake --build "$AI_LLAMA_BUILD" -j"$(nproc)"

Adjust AI_LLAMA_SRC / AI_LLAMA_BUILD if your trees live elsewhere. Update CMakeLists.txt AI_LLAMA_LIBS paths if your llama.cpp version places static libraries in different folders.

Step 3 — Configure and build the app

export AI_LLAMA_SRC=$HOME/_AugmentedIntelligence/_src/llama.cpp
export AI_LLAMA_BUILD=$HOME/_AugmentedIntelligence/_src/llama_build_cpu_avx2
export AI_SKIP_AD_CHECK=1

./scripts/build_ubuntu.sh

Binary output:

build/linux/_AugmentedIntelligence

Manual CMake:

cmake -S . -B build/linux -G Ninja -DCMAKE_BUILD_TYPE=Release
cmake --build build/linux -j"$(nproc)"

Step 4 — Run

export AI_SKIP_AD_CHECK=1
./build/linux/_AugmentedIntelligence --help
./build/linux/_AugmentedIntelligence

Useful environment variables:

Start the browser dashboard:

web server start

Then open http://127.0.0.1:8787/ (or your configured port).

Build variants

Optional CUDA app build:

ENABLE_CUDA=ON ./scripts/build_ubuntu.sh

Requires NVIDIA drivers, CUDA toolkit, and a CUDA-enabled llama.cpp build directory.

Project layout (Linux-specific)

Pairing with Pixel 4 / mobile

Run the Linux desktop agent, then use the Android companion app as documented in Pixel4-Install.md:

remote commands start 7710 YOUR_TOKEN
android audio start 7711 YOUR_TOKEN
web server start

Troubleshooting

Missing llama static libraries

CMake warns about missing libllama.a paths. Build llama.cpp first and verify files under $AI_LLAMA_BUILD.

mysql.h not found

sudo apt-get install libmysqlclient-dev

OpenCV / Tesseract errors

sudo apt-get install libopencv-dev libtesseract-dev libleptonica-dev

Active Directory / access denied on startup

export AI_SKIP_AD_CHECK=1

--gui does nothing useful

Expected on Linux. Use the web dashboard or console REPL.

First compile errors

The Linux port is Phase 1. Some translation units may still need small #ifdef _WIN32 guards as you iterate on your distro. Report or fix per-file platform issues; the scaffold (Platform.hpp, CMake, excluded sources) is the maintained baseline.

Related documents

• Pixel4-Install.md — Android companion
• llama_cpp_cpu_ram_cuda.md — llama memory tuning
• installer/install.ps1 — Windows installer (not used on Ubuntu)