GhostQuery
The Problem
Productivity tools often demand attention - they pop up, take over your screen, interrupt your flow. I wanted the opposite: an AI assistant you could summon instantly without breaking focus, hidden from view when not needed.
What I Built
GhostQuery is a hotkey-triggered stealth AI assistant that runs hidden from OS app lists. Press a global hotkey, ask your question, get an answer, and disappear - no window switching, no taskbar clutter, no evidence you were using AI at all.
Technical Implementation
Why Rust + Tauri?
The technical stack was chosen for stealth:
- Rust backend - handles LLM API queries with low latency, lightweight memory footprint
- Tauri shell - uses system webview instead of bundling Chromium (unlike Electron)
- Result: ~5MB app instead of ~150MB
Global Hotkey System
The global hotkey system required careful implementation. I needed something that works across all applications without conflicts. The system tray icon is optional and can be hidden entirely.
When activated, a minimal overlay appears at the cursor position - just an input field and response area, styled to blend with whatever you're working on.
API Key Security
The Express.js proxy handles API key management and request routing. It sits between the Tauri frontend and LLM providers, keeping API keys secure on the server side. This architecture also makes it easy to switch between different AI providers without changing the client.
Performance
The biggest challenge was making it feel instant. LLM responses have inherent latency, so I focused on the perception of speed:
- Immediate hotkey response
- Streaming text output
- Dismiss animation completes in under 200ms
The goal was to make it feel like a system feature, not an app.
Key Takeaway
Sometimes the best UX is invisibility. By optimizing for stealth and speed, GhostQuery becomes a tool you use without thinking about it - which is exactly what a productivity tool should be.