AI Instruments

AI Instruments is a macOS app that analyzes an iOS .app bundle and produces an "Instruments-style" diagnostic report focused on:

• Memory leaks
• Swift concurrency safety
• Memory allocations

This repository also contains TestApp, an iOS app intentionally packed with issue patterns so you can validate and iterate on the analyzer.

Repository Layout

• AIInstruments/ - main macOS analyzer app (SwiftUI)
• TestApp/ - iOS app used as analysis input

What AIInstruments Does

AIInstruments performs static binary analysis of a provided .app bundle:

1. Loads app metadata (Info.plist, executable, linked frameworks).
2. Parses Mach-O binary data (including debug dylib slices when present).
3. Runs three instrument analyzers:
   • LeaksAnalyzer
   • ConcurrencyAnalyzer
   • AllocationsAnalyzer
4. Aggregates findings into a report with:
   • Severity-tagged issues
   • Confidence values
   • Per-instrument score
   • Overall score and grade

The app supports drag-and-drop and file picker import for .app bundles and allows exporting a text report.

What "AI-driven" Means Here

The current implementation is an intelligent, rule-based analyzer over symbols, selectors, strings, sections, and linked libraries extracted from Mach-O binaries.

It is not a runtime profiler replacement for Apple Instruments. Instead, it is a fast static analysis assistant that can highlight likely risk patterns before runtime testing.

AIInstruments Architecture

Key modules:

• Analysis/AnalysisEngine.swift - analysis pipeline orchestration
• Analysis/MachOParser.swift - Mach-O/fat binary parsing
• Analysis/BinaryAnalyzer.swift - shared evidence and pattern helpers
• Analysis/LeaksAnalyzer.swift - leak risk detection
• Analysis/ConcurrencyAnalyzer.swift - Swift concurrency risk detection
• Analysis/AllocationsAnalyzer.swift - allocation and memory footprint risks
• Models/ - issue/result/report domain models
• DI/ - dependency injection with Factory (FactoryKit)

TestApp Purpose

TestApp is a companion iOS app that intentionally demonstrates problematic patterns from the same three categories:

• Memory leak demos (delegates, closures, notifications, timers, circular refs)
• Concurrency demos (non-Sendable, detached tasks, mixed models, unsafe continuations)
• Allocation demos (image loading, data buffers, collection growth, string churn, file I/O)

Use it to generate realistic .app inputs and verify whether AIInstruments reports expected findings.

Typical Workflow

1. Build TestApp for iOS Simulator in Debug.
2. Locate the built TestApp.app product.
3. Launch AIInstruments (macOS app).
4. Drop TestApp.app into AIInstruments.
5. Review issues and scores for Leaks, Swift Concurrency, and Allocations.
6. Export report if needed.

Getting a .app from TestApp

From Xcode:

• Open TestApp/TestApp.xcodeproj
• Select a Simulator destination
• Build (Cmd+B)
• In Products, right-click TestApp.app -> "Show in Finder"

From command line (example):

cd TestApp
xcodebuild -project "TestApp.xcodeproj" -scheme "TestApp" -configuration Debug -destination "platform=iOS Simulator,name=iPhone 16"

Then find the app under DerivedData Build/Products/Debug-iphonesimulator/TestApp.app.

Running AIInstruments

• Open AIInstruments/AIInstruments.xcodeproj
• Run the AIInstruments scheme on macOS
• Import a .app bundle via drag-and-drop or "Browse Files"

Tests

AIInstruments includes unit tests for parser, analyzers, models, and analysis engine under AIInstruments/AIInstrumentsTests/.

Run from command line:

cd AIInstruments
xcodebuild test -project "AIInstruments.xcodeproj" -scheme "AIInstruments" -destination "platform=macOS" -only-testing:AIInstrumentsTests

Notes and Limitations

• Input format: .app bundles (no .ipa required).
• Analysis quality depends on symbol/string availability in the binary.
• Encrypted App Store binaries are not analyzable until decrypted.
• Findings are risk indicators, not guaranteed runtime bugs.

Feedback and contribution

I'm always open to receiving feedback and contribution to my project so please feel free to open Pull Requests, raise Issues or reach out to me on X https://x.com/dammarkowski. Thanks!