Universal AI Connector
One reusable connection layer for adding AI to Android, iOS, macOS, Windows, and Linux applications.
Most teams do not want to rebuild the same AI integration for every app and every provider. I created Universal AI Connector to give different products one shared way to send requests, receive answers, stream live responses, and handle errors.
The business value is simpler delivery and lower maintenance. Teams can support several operating systems and AI services while keeping most provider-specific complexity out of their product code.
What this says about my work: I can design a reusable cross-platform foundation, make it feel natural to both Swift and Kotlin developers, and verify it across mobile, desktop, and server environments.
Why This Project Matters
Build The Foundation Once
Mobile, desktop, and server teams can reuse the same core behavior instead of solving authentication, streaming, errors, and provider differences repeatedly.
Avoid Tight Provider Lock-In
The application works with a stable set of models while separate adapters translate requests for each AI service.
Predictable Behavior Everywhere
Every supported platform receives the same rules for responses, streaming, errors, cancellation, and structured data.
Keep Product Code Cleaner
Provider-specific formats stay inside the connector, leaving application teams with a smaller and easier-to-understand integration.
The Simple View
Mobile, desktop, and server products
One consistent way to request and receive AI results
Direct providers or compatible private infrastructure
Where It Runs
Android And iOS
Android uses the shared Kotlin client. iOS receives a Swift-friendly package designed to feel natural inside an Apple application.
macOS, Windows, And Linux
The Kotlin/JVM client is verified on all three desktop operating systems and can support desktop tools or background services.
Native Swift And Kotlin APIs
Swift developers use Swift concurrency and Swift errors. Kotlin developers use coroutines and Flow without learning a foreign programming model.
What Application Teams Get
Standard And Live Responses
Applications can wait for a complete answer or show the result progressively as the AI generates it.
Reliable Structured Results
Apps can request validated JSON when a feature needs predictable data instead of free-form text.
Safe Cancellation And Cleanup
When a user leaves a screen or stops a request, the connector cancels the work and releases its resources cleanly.
Credentials Stay With The Host App
The connector requests credentials from the application when needed and does not own the app’s storage or secret-management decisions.
AI Services Covered
OpenAI Responses
Supports normal answers, structured data, live streaming, usage information, cancellation, and safe error handling.
Anthropic Messages
Provides the same application-facing behavior through Anthropic without exposing Anthropic-specific data models to the app.
OpenRouter
Adds access to OpenRouter while preserving the shared response, streaming, error, and cancellation rules.
OpenAI-Compatible Gateways
Connects to compatible gateways through the same reusable path, including validation against my independent LLM Gateway project.
Skills This Project Demonstrates
Cross-Platform System Design
I designed one shared foundation without forcing every platform to expose the same implementation details.
Swift-Friendly SDK Design
I converted a Kotlin Multiplatform core into a clean Swift experience with async calls, streams, errors, and cancellation.
Complex Asynchronous Behavior
The implementation handles concurrent requests, cancellations, streaming events, resource ownership, and close races.
Testing And Release Discipline
The repository uses contract fixtures, platform samples, live-provider checks, CI across operating systems, and explicit proof limits.
Current Project Status
The Core And Provider Paths
The shared client, Swift layer, platform samples, AI provider adapters, and compatible Gateway validation are implemented and tested.
Easier Installation
The next milestone will publish installable packages and add a graphical desktop demonstration for macOS, Windows, and Linux.
First Alpha Release
The final roadmap milestone focuses on full-platform verification, API compatibility, documentation, security, and release hardening.
Not Yet A Production Release
This is a working and verified engineering foundation. Remote packages, the desktop demo, and the first public alpha are still planned.
Technology Behind It
Shared AI Foundation
Apple Integration And Quality
Explore The Work
The public repository shows the implementation, architecture decisions, platform examples, verification approach, and remaining roadmap.
See how I approach cross-platform architecture, developer-friendly APIs, AI provider integration, and evidence-based delivery.