Custom Siri Intents: From INExtension to AppIntents

Users keep opening the app manually because Siri doesn't understand their commands. Custom Intents solve this: you define voice commands, Siri dictates parameters and executes the action in the background. We have implemented over 30 projects with SiriKit — average integration takes 2 weeks. Unlike basic Shortcuts, custom Intents support dialogs for parameter clarification. This reduces input errors and improves satisfaction. According to internal data, implementing intents increases app usage frequency by 15–20%. One client saved 2,000,000 rubles per year on support of old code after migrating to AppIntents.

How INExtension Works in Practice

INExtension is a process launched by Siri when a command is spoken. It contains handlers for each Intent defined in .intentdefinition. Xcode generates code for Intent, Response, and parameter classes. Consider a parameter structure:

Parameter: date
  Type: Date (INDateComponentsResolutionResult)
  Display Name: "date"
  Prompt: "What date?"
  Siri Dialog: "What date should the appointment be set for?"

Parameter types: String, Integer, Boolean, Date, CLPlacemark (geolocation), INPerson (contact from address book), custom INObject types for your app's entities. Siri receives parameters via INObject — a custom type describing entities. In INExtension, implement provide…Options…Collection which returns objects for selection.

Problems We Solve

Extension is not called — the most common reason. Ensure the Intent is added to NSExtension.NSExtensionAttributes.IntentsSupported in the extension's Info.plist, as well as in the main target. App Group not shared — extension and app are in different sandboxes. Use FileManager.default.containerURL(forSecurityApplicationGroupIdentifier: "group.com.yourapp") or UserDefaults(suiteName: "group.com.yourapp"). Siri doesn't understand specific terms — use INVocabulary.shared().setVocabularyStrings(projectNames, of: .organizationName) and a global AppIntentVocabulary.plist for static terms.

How to Solve Siri's Term Recognition Issues

Use INVocabulary and AppIntentVocabulary.plist to register names of projects, tasks, or other entities. This allows Siri to recognize them by voice without additional clarifications. In one project, we added over 200 terms, and recognition accuracy increased from 70% to 95%.

How We Develop Custom Intents

The first step is creating an .intentdefinition file with parameters and dialogs. Then we generate Swift classes and write handlers:

// Custom INObject
class ProjectObject: INObject {
    // identifier and displayString are mandatory
}

// In IntentHandler
func provideProjectOptionsCollection(
    for intent: CreateTaskIntent,
    with completion: @escaping (INObjectCollection<ProjectObject>?, Error?) -> Void
) {
    let projects = ProjectRepository().fetchAll()
    let objects = projects.map {
        ProjectObject(identifier: $0.id, display: $0.name)
    }
    completion(INObjectCollection(items: objects), nil)
}

Resolve, Confirm, Handle — three lifecycle methods:

Resolve — validates and clarifies each parameter:

func resolveTaskName(for intent: CreateTaskIntent,
                     with completion: @escaping (INStringResolutionResult) -> Void) {
    guard let name = intent.taskName, name.count >= 2 else {
        completion(.needsValue()) // Siri: "What should the task be named?"
        return
    }
    guard name.count <= 255 else {
        completion(.unsupported(forReason: .tooLong))
        return
    }
    completion(.success(with: name))
}

Confirm — final check:

func confirm(intent: CreateTaskIntent,
             completion: @escaping (CreateTaskIntentResponse) -> Void) {
    guard ProjectRepository().exists(id: intent.project?.identifier) else {
        let response = CreateTaskIntentResponse(code: .failure, userActivity: nil)
        response.failureReason = "Project not found"
        completion(response)
        return
    }
    completion(CreateTaskIntentResponse(code: .ready, userActivity: nil))
}

Handle — execution:

func handle(intent: CreateTaskIntent,
            completion: @escaping (CreateTaskIntentResponse) -> Void) {
    let store = TaskStore(appGroup: "group.com.yourapp")
    let task = store.create(
        name: intent.taskName!,
        projectId: intent.project?.identifier,
        dueDate: intent.dueDate?.dateComponents
    )
    let response = CreateTaskIntentResponse(code: .success, userActivity: nil)
    response.task = TaskObject(identifier: task.id, display: task.name)
    completion(response)
}

Response templates are defined in .intentdefinition: Siri will say "Task $(taskName) created in project $(project)".

Process

1. Analytics — study business logic and define commands.
2. Design — create .intentdefinition with parameters and dialogs.
3. Implementation — write handlers, INObject, integration with App Group and INVocabulary.
4. Testing — simulate Siri on a real device, check all failure scenarios.
5. Deployment — prepare for App Store Review, update config files if needed.

What's Included

• .intentdefinition file with parameters and dialogs.
• INExtension with full resolve/confirm/handle cycle.
• App Group setup for shared database.
• Integration with INVocabulary for voice recognition.
• (Optional) Migration to AppIntents for current iOS versions.
• Integration documentation.

Estimated Timelines

SiriKit vs AppIntents Comparison

AppIntents (current iOS versions) uses async/await, is more concise, supports SwiftUI, and works in widgets. Old code requires up to 150 lines per Intent, while new code takes about 40. AppIntents is 2–3 times faster to implement and easier to maintain. If your app already uses modern patterns, AppIntents is the obvious choice. More details: AppIntents.

Get a consultation from an experienced developer to evaluate your project. Contact us to discuss details and order an audit of your current SiriKit solution. Order custom Intents development — we guarantee App Store Review approval and full integration support.

Why is Native iOS Development the Best Choice for Complex Apps

The app crashes on cold start — EXC_BAD_ACCESS at the moment of initializing a singleton that accesses another singleton that hasn't been initialized yet. Or: a ViewController leaks memory because a closure captures self without [weak self], and that ViewController hangs in memory two transitions after the user left it. These are not hypothetical scenarios — they are the two most common classes of problems on iOS projects that come to us after another team.

We have been doing iOS development for over 5 years, delivered 40+ projects of varying complexity — from startups to enterprise solutions with millions of users. Each project undergoes 3 stages of Code Review, a custom set of UI tests (150+ test cases on average), and a mandatory run through Xcode Instruments before release.

Native iOS development with Swift means direct access to the platform. No middleware, no performance compromises, full control over what happens on every frame.

What Makes Native iOS Development on Swift the Choice for Enterprise Apps?

Native code guarantees compatibility with new Apple APIs on the day they are released, not after months of adaptation in cross-platform frameworks. For apps with latency-sensitive logic (financial terminals, medical monitors, AR navigation), this is critical. Swift with ARC and strict typing allows maintaining a crash-free rate of 99.9% with proper architecture.

SwiftUI or UIKit: What to Choose for Native iOS Development

By now, SwiftUI covers the vast majority of production tasks. But UIKit is not deprecated and will not disappear — Apple does not deprecate it but continues to add APIs. The real picture on large projects: a hybrid approach. SwiftUI for most screens, UIKit where SwiftUI hits limitations.

Which Scenarios Does SwiftUI Win Unconditionally

SwiftUI's declarative syntax reduces UI code by 3-5 times compared to UIKit. A settings screen with List, Toggle, Picker — that's 40 lines of SwiftUI versus 200 lines of UIKit with UITableViewDataSource delegates. Time savings on UI development reach 60%. Apple recommends starting new projects on SwiftUI (Human Interface Guidelines).

@State, @Binding, @ObservableObject (and with iOS 17, the @Observable macro) create a reactive link between data and UI without manual reloadData(). Changing a @State variable automatically redraws the affected part of the hierarchy. This works correctly if you understand how SwiftUI computes the diff — via Equatable and id in ForEach.

AsyncImage, NavigationStack with type-safe routing via NavigationPath, searchable, refreshable — these are ready-made patterns that UIKit requires implementing manually.

When UIKit Remains Necessary

UICollectionView with compositional layout and diffable data source — complex grids with different cell types, horizontal sections inside vertical scroll, dynamic cell sizes. SwiftUI LazyVGrid / LazyHGrid do not provide such control.

Custom transitions between screens. UIViewControllerAnimatedTransitioning and UIViewControllerInteractiveTransitioning — interactive pop gesture with partial progress, custom hero transition with precise frame control. SwiftUI matchedGeometryEffect covers some cases, but not all.

UITextView with TextKit 2. Rich text editor, custom attributes, custom rendering — TextKit 2 (available since iOS 16) switched to async layout, solving performance issues on long documents. SwiftUI TextEditor is a wrapper around UITextView without direct access to TextKit.

UIScrollView with custom behavior. scrollViewDidScroll, parallax effects, sticky headers with custom logic, pull-to-refresh with custom indicator. SwiftUI ScrollView with scrollPosition and onScrollGeometryChange (iOS 17) covers some cases, but not all.

How Do We Integrate SwiftUI and UIKit Step by Step

1. Identify screens where SwiftUI gives maximum gain (lists, forms, settings) — usually 70-80% of screens.
2. For performance-critical areas (complex collections, custom animations) leave UIKit.
3. Use UIHostingController to embed SwiftUI views into UIKit navigation stack.
4. For backward compatibility, wrap UIKit components via UIViewRepresentable.
5. Coordinator pattern (UIKit) manages navigation at the flow level, screens are implemented in SwiftUI.

One pattern we use on projects: UIKit coordinator manages navigation, while the screens themselves are in SwiftUI. The coordinator creates a UIHostingController, passes ViewModel via initializer or @EnvironmentObject, and manages transitions. This gives clean separation: SwiftUI handles UI, Coordinator handles navigation.

How async/await and Combine Work Together

Before Swift 5.5, asynchronous code on iOS was built on Combine or callback chains. With the advent of async/await and Actor, concurrency has become part of the language. On new projects we use async/await as the primary tool for network calls and business logic, and Combine for reactive UI state binding.

// Correct — @MainActor guarantees UI updates on main thread
@MainActor
class UserViewModel: ObservableObject {
    @Published var user: User?
    @Published var isLoading = false

    func loadUser(id: String) async {
        isLoading = true
        defer { isLoading = false }
        do {
            user = try await userService.fetch(id: id)
        } catch {
            // handle error
        }
    }
}

Combine remains indispensable for debouncing input, merging multiple Publishers (CombineLatest, Zip), and functional processing of value streams (map, flatMap, filter). In practice, 80% of projects use both approaches, choosing the tool for the task.

iOS App Architecture

MVVM — the basic pattern. ViewModel contains logic and @Published state, SwiftUI View subscribes via @ObservedObject or @StateObject. One rule: View knows nothing about URLSession, CoreData, UserDefaults.

Clean Architecture adds Repository and UseCase layers. UserRepository abstracts the data source (network vs cache). FetchUserUseCase contains business logic. UserViewModel calls UseCase and manages UI state.

TCA (The Composable Architecture) — a stricter pattern from Point-Free. State, Action, Reducer, Effect — everything explicit, testable, composable via Scope. Works well in large teams (5+ iOS developers) where predictability is important.

What's Included in iOS App Development

Tools Without Which No Release Is Complete

Xcode Instruments. Time Profiler shows where CPU spends time. Allocations — memory leaks and excessive allocations. Leaks — objects that are not freed. Before every release — a mandatory run.

Firebase Crashlytics. Crash-free rate, grouping by stack trace, breadcrumbs of events leading to crash. Set up in 30 minutes, provides visibility across the entire device fleet. On our projects, average crash-free rate is 99.8%.

Fastlane match. Manage certificates and provisioning profiles via an encrypted git repository. Eliminates the 'it builds locally but not on CI' issue once and for all. Saves up to 4 hours per build when signing manually.

XCTest + XCUITest. Unit tests for ViewModel and UseCase, UI tests for critical flows (onboarding, payment, authorization). On average, code coverage is 75%.

Process and Timelines

Cost is calculated individually after analyzing the technical specification and design. Typically, the first 2 weeks are spent on design, after which we finalize the timeline and budget.

Order turnkey development — we will evaluate your project in 2 business days and propose the optimal architecture. Contact us to discuss your task: we guarantee code quality, compliance with App Store Review Guidelines, and experience with projects of any scale. Get a consultation — we will help you choose the right stack and avoid common mistakes at the start.