Smart Core, Simple Form

English | 中文

🎉 Coming Soon to Mac App Store!

Fully compliant with Apple's security & privacy standards. Currently under review. Your support and stars ⭐️ keep us going!

WitNote is a local-first AI writing companion for macOS, Windows, and Linux. Supports free switching between Ollama / WebLLM / Cloud API engines, paired with an ultra-minimalist native card interface, it works right out of the box. No continuous cloud dependency, no privacy concerns — intelligence made lightweight.

• Smart: Three Engines in One, Freedom to Choose
  • WebLLM: Lightweight model, requires download on first run, then works offline
  • Ollama: Powerful local model, strong performance, completely offline
  • Cloud API: Connect to cloud intelligence, infinite possibilities
• Simple: No complexity
  • iOS-style card management, drag to organize
  • Smart focus mode — window narrows, editor simplifies
• Secure: Data sovereignty
  • Privacy Policy: 100% local storage. Your thoughts belong only to you.

• ✨ Enhanced Autocomplete — 3 selectable levels (Lite/Standard/Full) for different model sizes.
• ⌨️ Smart Tab — Press Tab to accept sentence by sentence, or segments, for precise control.
• 🛠️ Customizable Prompts — Edit system prompts directly in settings with a one-click restore option.
• 🌓 Dual-Pane Preview — WYSIWYG, edit on the left, real-time preview on the right
• ⚡️ Quick Edit — Enhanced floating menu for styling selected text

• 📝 Pure Local Notes — Choose any folder as your notes vault, supports .txt and .md
• 🤖 Three-in-One Engine — Switch freely between WebLLM (Light), Ollama (Local Power), or Cloud API (Custom Connection)
• 🎭 Rich Role Library — Built-in 10+ selected role prompts (Writer, Translator, Polisher, etc.), one-click switch & custom support
• 🌍 Global Communication — Supports 8 Languages:
  • English, 简体中文, 繁體中文, 日本語, 한국어, Français, Deutsch, Español
  • Interface and AI responses automatically adapt to your language
• 🔒 Privacy First — All AI inference is local (when using local engines), no data upload, Apple Notarized
• 💬 Highly Customizable — Freely edit system prompts to craft your exclusive AI assistant
• 🎨 Multiple Themes — Light / Dark / Zen Tea, fully optimized dark mode
• 🗂️ Card Grid View — iOS-style with drag-and-drop sorting, polished context menus
• 🔍 Context Aware — AI can directly read your current article or folder contents
• 🎯 Focus Mode — Auto-switches to distraction-free editing when window narrows

Download the latest installer from Releases:

❌ Important Note for Intel Macs:

This application does not support Mac computers with Intel chips. Even if forced to run, the experience will be extremely poor due to the following reasons:

1. Architectural Incompatibility: The built-in local inference engines (WebLLM/Ollama) deeply rely on the ARM64 architecture and NPU/Metal hardware acceleration of Apple Silicon.
2. Lack of Hardware Acceleration: Intel Macs lack Unified Memory Architecture. Running quantized models is extremely slow (generating a single token may take seconds) and causes severe device heating.
3. Architectural Trade-off: To ensure the best experience and minimal package size, we have removed support for the x86_64 architecture.

We strongly recommend using Mac devices equipped with Apple Silicon (M-series) chips.

⚠️ Note: Windows ARM64 devices (e.g. Surface Pro X) are now natively supported!

⚠️ Note: Windows version is newly released. Feedback welcome!

1. Download the .dmg file
2. Double-click to open the DMG
3. Drag the app to Applications folder
4. Launch from Applications

🎉 Great News!

This app is now Apple Notarized! No more "unverified developer" warnings!

😅 The developer bravely took out a loan to afford the $99 Apple Developer account... *(Yes, this actually happened. Thanks to all users for your support!)

1. Download the .exe installer
2. Run the setup wizard
3. Choose installation path (customizable)
4. Complete installation, launch from Desktop or Start Menu

AppImage (Universal):

1. Download .AppImage file
2. Right-click Properties -> Allow executing file as program (or chmod +x WitNote*.AppImage)
3. Double-click to run

Deb (Ubuntu/Debian):

1. Download .deb file (e.g., WitNote-1.3.1-amd64.deb)
2. Run installation via terminal (automatically handles dependencies):

   sudo apt install ./WitNote-1.3.2-amd64.deb

📝 Important Notes for Windows Users:

As an individual developer without an expensive EV Code Signing Certificate, you might encounter the following:

1. SmartScreen: If you see "Windows protected your PC" (Unknown Publisher), please click "More info" -> "Run anyway".
2. Antivirus Warning: Windows Defender or other AV software might flag the installer. The project is open-source and safe. If blocked, please try disabling AV temporarily.

The app includes a built-in WebLLM engine with qwen2.5:0.5b model (macOS only, Windows users are recommended to use Ollama).

• Pros: No extra software installation needed, works completely offline after initial model download.
• Best for: Quick Q&A, simple text polishing, low-end devices.

Supports connecting to locally running Ollama service.

• Pros: Runs 7B, 14B or even larger models, powerful performance, completely offline.
• Usage: Install Ollama first, then download more models in Settings (e.g., qwen2.5:7b, llama3, etc).

Supports connecting to OpenAI-compatible Cloud APIs.

• Pros: Access the most powerful models on Earth with just an API Key.
• Best for: Top-tier logical reasoning, or when local hardware cannot support large models.
• Config: Enter API URL and Key in Settings (Supports OpenAI, Gemini, DeepSeek, Moonshot, etc).

# Clone the repository
git clone https://github.com/hooosberg/WitNote.git
cd WitNote

# Install dependencies
npm install

# Start development server
npm run dev

# Build macOS version
npm run build

# Build Windows version
npm run build -- --win

MIT License

hooosberg

📧 zikedece@proton.me

🔗 https://github.com/hooosberg/WitNote

📖 Development Diary

Smart Core, Simple Form