Nexus is an ambient, context-aware AI assistant that lives on your desktop. It's designed to be an "over-the-shoulder" partner, capable of seeing your screen, understanding your requests, and even performing actions on your computer, all while remaining invisible to screen recordings.
nexus-demo.online-video-cutter.com.online-video-cutter.com.mp4
- Context-Aware Vision: Can analyze your screen to understand the context of your questions. Perfect for debugging code, summarizing articles, or getting help with any application.
- Stealth Mode: Remains completely invisible during screen shares or recordings, so you can get help without distracting others.
- Always-On-Top & Transparent UI: The sleek, translucent interface floats above your other windows, ensuring the assistant is always available at a glance.
- Actionable Commands: Go beyond chat. Instruct Nexus to create files, open applications, or browse folders directly on your desktop.
- Live Internet Access: Integrated with real-time web search to provide up-to-the-minute answers on current events and breaking news.
- Personalized & Conversational: Remembers your self-described profile and the current conversation to provide tailored, follow-up responses.
- Desktop-Native Feel: Fully draggable window and movable with
Ctrl/Cmd + Arrow Keyhotkeys. - Developer-Friendly: Automatically formats and syntax-highlights code snippets in its responses.
- Framework: Electron
- UI: React & TypeScript
- Styling: Tailwind CSS
- Bundler: Vite
- Core AI Model: Google Gemini Pro & Vision
- Web Search: SerpAPI
- Node.js (v18 or later recommended)
npmoryarnpackage manager
First, clone the project to your local machine:
git clone <your-repository-url>
cd nexusInstall all the necessary packages for the project:
npm installThe assistant requires two API keys to function:
- Google Gemini API Key: For the core AI and vision capabilities. Get one from Google AI Studio.
- SerpAPI Key: For the real-time web search tool. Get one from your SerpAPI Dashboard. The free plan is sufficient.
Note: You do not need to put these keys in the code. The application will ask for them on the first run.
Launch the application in development mode. This will open the window and enable live-reloading for any code changes:
npm run dev-
Setup Page: On the very first launch, you'll be greeted by a Setup Page. Write a short paragraph describing yourself or what you do (e.g., "I'm a backend developer specializing in Go and Python."). This helps personalize the AI's responses.
-
Settings Page: Next, you'll be taken to the Settings Page. Enter your Google Gemini API key and your SerpAPI key into the respective fields and click "Save & Continue".
Simply type a question and get an answer. The AI will remember your profile and the previous turns in the conversation.
Example: "What do you know about me?"
Check the "Include Screenshot" box before sending your message. Nexus will analyze your screen along with your text.
Example: (While looking at a bug in your code) "Find the error in this function."
Ask a question about a recent event or a specific fact. The AI will automatically decide when to use its web search tool.
Example: "What were the top headlines in tech news yesterday?"
You can command Nexus to perform tasks. Try prompts like:
"Create a file on my desktop named 'test.txt' with the content 'Hello, World!'""Open the Notes app for me."
- Drag & Drop: Click and drag the header bar of the application to move it.
- Hotkeys: Use
Ctrl + Arrow Keys(orCmd + Arrow Keyson macOS) to nudge the window's position on your screen.
- Debug code by showing your screen and asking for help
- Get explanations of complex code patterns
- Create files and folders quickly
- Open development tools and applications
- Summarize articles or documentation
- Get help with writing and editing
- Research topics with live web search
- Format content in various styles
- Quick file operations
- Application launching
- Context-aware assistance with any task
- Always-available help without disrupting workflow
Nexus automatically enters stealth mode during screen recordings and shares, making it invisible to others while remaining functional for you.
The interface transparency can be adjusted to suit your preferences and desktop environment.
Ctrl/Cmd + Arrow Keys: Move window position- Default positioning keeps the assistant accessible but unobtrusive
-
Be Specific: When asking for help with screen content, be clear about what you want to focus on.
-
Use Action Commands: Take advantage of the ability to perform file operations and launch applications.
-
Leverage Context: Include screenshots when the visual context is important to your question.
-
Personalize: Update your profile in settings to get more tailored responses.
-
Combine Features: Use web search for current information while leveraging your screen context for specific tasks.
- API keys are stored locally and never transmitted except to the respective services
- Screen captures are processed locally and only sent to AI services when explicitly requested
- No conversation data is stored permanently without your consent
- Stealth mode ensures privacy during screen shares
Contributions to this project are welcomed.
If you encounter any issues or have questions:
- Check the troubleshooting section in our documentation
- Search existing issues in the repository
- Create a new issue with detailed information about your problem
Made with β€οΈ for productivity enthusiasts and developers who want an intelligent desktop companion.