A fully offline AI chat desktop app built with Tauri v2 (Rust) and Vue 3. No cloud calls are made during inference — models run entirely on your machine.
- Fully offline inference — no data leaves your device
- Dual backend support — run models via llama.cpp (with Metal acceleration on macOS) or Cactus
- Model catalog — one-click download and install for curated models
- HuggingFace quant picker — browse and select quantizations directly from HF repos
- Streaming responses — tokens stream in real time as they are generated
- Session management — persistent chat history stored locally as JSON
- Multiple sessions — sidebar for creating, switching, and deleting conversations
| Model | Provider | Size (llama.cpp) | Backends |
|---|---|---|---|
| Gemma-3-1B | ~806 MB | llama.cpp, Cactus | |
| Gemma-4-E2B | ~3.1 GB | llama.cpp, Cactus | |
| Bonsai-8B | Prisma ML | ~1.2 GB | llama.cpp |
Any GGUF model from HuggingFace can also be installed by browsing quants directly in the app.
- macOS (primary target; Metal GPU acceleration on Apple Silicon and AMD/Intel Macs)
- Rust toolchain
- Node.js 18+
- Tauri CLI v2:
npm install -g @tauri-apps/cli
# Install frontend dependencies
npm install
# Start the full dev environment (Vite + Rust hot-reload)
npm run tauri dev
# Frontend only (http://localhost:1420)
npm run dev
# Type-check + build frontend bundle
npm run build
# Rust compilation only
cargo build --manifest-path src-tauri/Cargo.toml
# Rust tests
cargo test --manifest-path src-tauri/Cargo.toml
# Frontend tests
npm testnpm run tauri buildProduces a signed .app bundle (macOS) in src-tauri/target/release/bundle/.
localagent/
├── src/ # Vue 3 frontend
│ ├── views/ # SetupView, ChatView, SettingsView
│ ├── stores/ # Pinia stores (modelStore, chatStore)
│ ├── components/ # Reusable UI components
│ └── types/ # Shared TypeScript types
└── src-tauri/ # Rust backend
├── src/
│ ├── backends/ # CactusBackend, LlamaCppBackend
│ ├── commands/ # Tauri IPC commands (inference, models, sessions)
│ ├── catalog.rs # Hardcoded model catalog
│ └── state.rs # AppState (loaded model, inference guards, downloads)
└── Cargo.toml
Model files and session data are stored in the OS app local data directory (~/Library/Application Support/localagent/ on macOS).
MIT