Google MLKit Swift Package Manager Wrapper

The demo app uses a generic reusable view called BaseDetectionView to implement vision-based features. This pattern provides a consistent UI for:

• Photo library picking
• Camera capture
• Sample image navigation (previous/next)
• A detection button with loading states
• Error handling

When building your own vision feature views, you can follow this signature to ensure a consistent user experience.

BaseDetectionView(
  title: String,
  sampleImages: [String],
  detectionHandler: (UIImage) async throws -> Void,
  resultView: () -> Content
)

Due to limitations in the pre-built MLKit binaries regarding arm64 support for the iOS simulator, this project only supports the following architectures:

• iphoneos: arm64
• iphonesimulator: x86_64

Once XCFrameworks are built, you can upload the resulting zip files to an existing GitHub release using upload_release.sh.

Note: The release must already exist. You can create one using the GitHub CLI:

gh release create <version> --title "Release <version>" --notes "Updated to MLKit <version>"

./scripts/upload_release.sh <version>

# Example
./scripts/upload_release.sh 7.0.0

You can manually trigger the XCFramework build process through the GitHub UI:

1. Navigate to the Actions tab.
2. Select the "Build MLKit XCFrameworks" workflow.
3. Click "Run workflow".
4. Enter the desired MLKit version (e.g., 5.1.0).
5. Select whether to create a release automatically.
6. Click "Run workflow".

To use this wrapper, add the package dependency to your Package.swift.

Important for App Store submissions: The standard from: "9.0.0" version contains internal Info.plist values that cause App Store Connect rejections. If you are submitting to the App Store, you must pin the wrapper-only repackage using the exact: parameter to use the fixed version (e.g., 9.0.0-1).

After adding the package, add the specific ML Kit modules you need to your target dependencies.

// Standard installation
dependencies: [
    .package(url: "https://github.com/d-date/google-mlkit-swiftpm", from: "9.0.0")
]

// REQUIRED for App Store submissions to avoid Info.plist rejections
dependencies: [
    .package(url: "https://github.com/d-date/google-mlkit-swiftpm", exact: "9.0.0-1")
]

When preparing a new version of the MLKit SwiftPM wrapper, follow this sequence to ensure stability:

1. Build: Run the build script for the target version.
2. Analyze: Run static analysis on the resulting artifacts.
3. Verify: Open the Example app and test on a physical device.
4. Commit: If all tests pass, commit the updated Podfile, Package.swift, and Resources/.

# 1. Build new version
./scripts/build_all.sh <version>

# 2. Run static analysis
./scripts/verify_runtime.sh <version>

# 3. Test Example app
cd Example
open Example.xcworkspace
# Build and run on device
# Test all features
# Check console

# 4. If all tests pass, commit
cd ..
git add Podfile Podfile.lock Package.swift Resources/
git commit -m "Update to MLKit <version>"
git tag -a <version> -m "Release <version>"

After building a new version, you should run manual static analysis to check for common runtime issues such as incorrect architectures, missing Info.plist files, or invalid symbol tables. Use the verify_runtime.sh script and provide the version number as an argument.

./scripts/verify_runtime.sh <version>

The xcframework-maker tool is a custom utility used to convert MLKit frameworks into XCFrameworks. It handles static frameworks with FAT object files, adds arm64 simulator support, and manages Info.plist files. You must build this tool before attempting to create XCFrameworks.

make bootstrap-builder

To ensure the ML Kit features work correctly in your project, verify the following setup steps:

Permissions

Add the following keys to your Info.plist to allow access to hardware and data:

• Camera: Required for real-time vision features.
• Photo Library: Required for picking images from the gallery.

Resources

• Custom Models: If using custom object detection or image labeling, ensure your .tflite models (e.g., bird.tflite) are included in your app's target bundle resources.
• Sample Images: Ensure all test images used for navigation are included in the Resources folder and added to the target.

The most critical step in verifying a build is running the provided Example app. This ensures that the frameworks link correctly and that features like Face Detection or Barcode Scanning do not crash at runtime.

Device Testing (Required)

1. Build for a physical iOS device.
2. Ensure the app launches without crashing.
3. Navigate through all available features (Barcode Scanner, Face Detection, etc.).
4. Monitor the Xcode console for warnings like Could not load bundle... or symbol conflicts.

Simulator Testing (Optional)

1. Build for the iOS Simulator.
2. Note that camera-based features may not function in the simulator.
3. Verify basic UI navigation works.

cd Example
open Example.xcworkspace