Back to selected works
macOS Native App & Local AI Sandbox

OmniKey

A unified Swift/AppKit command hub utility for macOS, combining window snapping, custom gestures, clipboard history, dev server monitoring, and on-device offline LLMs.

SwiftAppKitCore DataLocalLLMAccessibility API

Project Overview & Objectives

OmniKey was conceived to solve utility fragmentation on macOS. Power users routinely run five or six distinct utility applications for simple window management, clipboard caching, short-keys, and dev port listening. Each background utility consumes separate system resources, requires independent macOS accessibility permissions, and frequently clashes on keyboard shortcut mappings. OmniKey consolidates these utility operations into a unified Swift daemon utilizing Apple's modern AppKit library. By writing the core event loops in native Swift and avoiding heavy Electron or Javascript wrappers, the background process runs continuously with less than 40MB of RAM. Additionally, we embedded a lightweight Local AI sandbox using the on-device Apple Neural Engine, enabling offline text explanation and code suggestion without transmitting sensitive clipboard data to external servers.

Core Tech Stack

  • Swift
  • AppKit
  • Core Data
  • LocalLLM
  • Accessibility API

The Challenge

Power users running separate tools for window snapping, clipboard history, hotkeys, gesture monitoring, and AI assistants experience massive CPU/RAM overhead, disjointed shortcuts, and privacy risks from cloud telemetry.

Our Solution

Tilda designed and built OmniKey in native Swift and AppKit. Running entirely offline with zero telemetry, it consolidates all productivity utility engines into a single lightweight background process with a footprint of less than 40MB RAM, utilizing macOS Accessibility APIs for window management and the local Apple Neural Engine for offline AI assistance.

< 40MB RAM Footprint
100% Offline Privacy
6-in-1 Utility Consolidation

Interactive Device Preview

https://omnikeyapp.com
OmniKey Menu Bar Control Panel
SYSTEM ARCHITECTURE

Data Flow & Pipeline Architecture

A technical overview of how data is ingested, processed, and routed through the system.

01

Event Listener

Captures system-wide keystrokes and gesture inputs via macOS Accessibility APIs.

02

Local Routing Engine

Decides whether to trigger window snapping, clipboard retrieval, port scanning, or the local AI context.

03

Apple Neural Engine

Loads local Llama weights entirely offline for immediate clipboard analysis and code autocomplete.

ENGINEERING SPRINT

Development & Delivery Timeline

Key phases of the development sprint from initial database design to production rollout.

Phase 1

Core Daemon & Window Snapping

Initializing the AppKit daemon process and registering accessibility event handlers for instant window resizing.

Phase 2

Port Scanning & Clipboard Cache

Coding the background thread port scanner and integrating Core Data to store clipboard history securely.

Phase 3

Neural Engine AI Sandbox

Optimizing the on-device ML model execution using CoreML and building the SwiftUI overlay UI.

COMPLIANCE & ARCHITECTURE

Sandboxing & Security Framework

Every digital product we engineer complies with modern security frameworks and operating system boundaries.

🛡️

Window Snap Engine

Resizes and positions application frames on-screen using macOS Accessibility APIs.

🛡️

TCP Port Listener

Scans and detects active local development servers and custom endpoints.

🛡️

Local AI Sandbox

Runs fully offline semantic intelligence and code suggestions via on-device LLMs.