LingmoOS AI Assistant

Multi-backend AI chat client with Material Design 3 UI built on Qt6.

Features

Feature	Description
Multi-backend	DeepSeek, Ollama (Local & Cloud), llama.cpp (GGUF)
Streaming	Real-time SSE / JSON-lines token streaming
Context management	Token-budget-based trimming, 8192 default history limit
Hardware-aware	Auto-detect CPU/GPU/VRAM for optimal local model config
Auto-update	GitHub release check with daily/weekly/monthly interval
Custom fonts	Outfit (UI), Noto Sans SC (CJK), JetBrains Mono (code)
Material Design 3	Light/dark theming with M3 color system
Debian packaging	Native .deb package for Debian Bookworm+ amd64

Architecture

src/
├── core/           ChatController, ContextManager, TokenEstimator
├── providers/      AIProvider → DeepSeek, Ollama, LlamaCppProvider
├── network/        NetworkClient, StreamParser (SSE/JSON-lines)
├── models/         HardwareDetector, ModelManager (GGUF), ModelDownloader
└── utils/          Logger, Platform
resources/qml/      Qt6 QML UI (ChatView, Sidebar, Settings, SetupWizard)
scripts/            Shell install/setup helpers
debian/             dpkg-buildpackage configuration

Prerequisites

• Debian 12 (Bookworm) or later
• Qt 6.4+ development packages
• CMake 3.22+, Ninja
• GCC 12+ or Clang 15+

Quick Install

From .deb

sudo apt install ./lingmo-assistant_0.1.0-1_amd64.deb
sudo apt install -f
lingmo-assistant

Build from source

git clone https://github.com/arkimium/lingmo-assistant.git
cd lingmo-assistant

Install dependencies

# One-liner
make deps

# Or manually:
sudo apt install -y \
    build-essential cmake ninja-build pkg-config \
    qt6-base-dev qt6-declarative-dev qt6-tools-dev \
    qt6-tools-dev-tools qt6-svg-dev qt6-base-dev-tools \
    libqt6network6 libqt6core6 libqt6qml6 libqt6quick6 \
    qml6-module-qtquick-controls qml6-module-qtquick-layouts \
    qml6-module-qtquick-templates qml6-module-qtquick-dialogs \
    qml6-module-qtquick-window qml6-module-qtqml \
    qml6-module-qtqml-workerscript qt6-qpa-plugins \
    libgl1-mesa-dev libvulkan-dev libcurl4-openssl-dev \
    libssl-dev libfontconfig1-dev libfreetype6-dev

Build and run

make build
make run

# Or build .deb package
make deb-docker

Makefile targets

Command	Description
make build	Build with CMake (Release)
make run	Build and launch
make deb	Build .deb natively
make deb-docker	Build .deb in Docker container
make deps	Install system dependencies via apt
make test	Quick smoke test
make test-docker	Full test with Xvfb in Docker
make install	Install to /usr/local/bin
make clean	Clean build artifacts
make lint	Lintian quality check

Manual build

cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build -j$(nproc)
./build/lingmo-assistant

Configuration

First run

The app auto-runs an initialization wizard on first launch:

• Checks system environment
• Downloads fonts (Outfit, Noto Sans SC, JetBrains Mono)
• Detects Ollama and llama.cpp availability
• Configures API keys

Alternatively, run the terminal wizard:

bash scripts/first-run-wizard.sh

Provider setup

Provider	Type	Setup
DeepSeek	Cloud	Enter API key in Settings
Ollama Local	Local	bash scripts/setup-ollama.sh
Ollama Cloud	Cloud	Configure endpoint in Settings
llama.cpp	Local	bash scripts/setup-llama-cpp.sh

Context window

Default history limit: 8192 tokens. Adjust in Settings → Context. Uses adaptive Chinese/English token estimation with 85% safety margin.

License

GNU General Public License v3.0