Hi everyone, I’m running into a breaking issue with SwiftData automatic CloudKit syncing on iOS 26, and I'm trying to determine if this is a known regression or a new configuration requirement I missed. The Setup: My setup is extremely standard; I am using the default configuration exactly as described in Apple's documentation here: https://developer.apple.com/documentation/swiftdata/syncing-model-data-across-a-persons-devices The schema is very simple: A single @Model class. No relationships. The Issue: Prior to iOS 26, this exact app was successfully syncing data between devices and to iCloud without issues. Immediately after the iOS 26 update, syncing stopped completely. I haven't changed any code, but when I check the CloudKit Console, I am seeing some BAD_REQUEST errors during sync attempts. Since I am using the default SwiftData sync (and not manual CKRecord handling), I’m not sure how my client code could be triggering a bad request unless the schema requirements have changed under the hood. Questions: Has anyone else seen increased BAD_REQUEST errors with SwiftData on iOS 26? Are there new entitlements or strict schema requirements introduced in iOS 26 that might cause a previously valid model to be rejected by CloudKit? Any pointers or confirmations would be appreciated. Thanks!

This is probably super simple answer that I missed, but: I have an app that has a database; I'd like to create a second app (actually a CLI tool), and access the same database. Is that possible? And, if so, how? 😄

I have a single multiplatform application that I use NSPersistentCloudKitContainer on. This works great, except I noticed when I open two instances of the same process (not windows) on the same computer, which share the same store, data duplication and "Metadata Inconsistency" errors start appearing. This answer (https://stackoverflow.com/a/67243833) says this is not supported with NSPersistentCloudKitContainer. Is this indeed true? If it isn't allowed, is the only solution to disable multiple instances of the process via a lock file? I was thinking one could somehow coordinate a single "leader" process that syncs to the cloud, with the others using NSPersistentContainer, but this would be complicated when the "leader" process terminates. Currently, it seems iPad split views are new windows, not processes -- but overall I'm still curious :0 Thank you!

My iOS app uses CloudKit key-value storage. I have not updated the app in a few years but it works fine. Since it was last updated, I transferred the app from an old organization to my personal developer account. Now that I'm working on the app again I get an error: Provisioning profile "iOS Team Provisioning Profile: com.company.app" doesn't match the entitlements file's value for the com.apple.developer.ubiquity-kvstore-identifier entitlement. In the entitlement file, it has $(TeamIdentifierPrefix)$(CFBundleIdentifier) as the value for iCloud Key-Value Store. I've verified the variables resolve as expected. When I parse the provisioning profile there is no entitlement value for key-value storage. What am I getting wrong?

Hi, I was testing the new iOS 18 behavior where NSPersistentCloudKitContainer wipes the local Core Data store if the user logs out of iCloud, for privacy purposes. I ran the tests both with a Core Data + CloudKit app, and a simple one using SwiftData with CloudKit enabled. Results were identical in either case. In my testing, most of the time, the feature worked as expected. When I disabled iCloud for my app, the data was wiped (consistent with say the Notes app, except if you disable iCloud it warns you that it'll remove those notes). When I re-enabled iCloud, the data appeared. (all done through the Settings app) However, in scenarios when NSPersistentCloudKitContainer cannot immediately sync -- say due to rate throttling -- and one disables iCloud in Settings, this wipes the local data store and ultimately results in data loss. This occurs even if the changes to the managed objects are saved (to the local store) -- it's simply they aren't synced in time. It can be a little hard to reproduce the issue, especially since when you exit to the home screen from the app, it generally triggers a sync. To avoid this, I swiped up to the screen where you can choose which apps to close, and immediately closed mine. Then, you can disable iCloud, and run the app again (with a debugger is helpful). I once saw a message with something along the lines of export failed (for my record that wasn't synced), and unfortunately it was deleted (and never synced). Perhaps before NSPersistentCloudKitContainer wipes the local store it ought to force sync with the cloud first?

I'm building a SwiftUI app with SwiftData and want to centralize both query logic and related actions in a manager class. For example, let's say I have a reading app where I need to track the currently reading book across multiple views. What I want to achieve: @Observable class ReadingManager { let modelContext: ModelContext // Ideally, I'd love to do this: @Query(filter: #Predicate<Book> { $0.isCurrentlyReading }) var currentBooks: [Book] // ❌ But @Query doesn't work here var currentBook: Book? { currentBooks.first } func startReading(_ book: Book) { // Stop current book if any if let current = currentBook { current.isCurrentlyReading = false } book.isCurrentlyReading = true try? modelContext.save() } func stopReading() { currentBook?.isCurrentlyReading = false try? modelContext.save() } } // Then use it cleanly in any view: struct BookRow: View { @Environment(ReadingManager.self) var manager let book: Book var body: some View { Text(book.title) Button("Start Reading") { manager.startReading(book) } if manager.currentBook == book { Text("Currently Reading") } } } The problem is @Query only works in SwiftUI views. Without the manager, I'd need to duplicate the same query in every view just to call these common actions. Is there a recommended pattern for this? Or should I just accept query duplication across views as the intended SwiftUI/SwiftData approach?

Hi all, I’m trying to understand SwiftData’s runtime semantics around optional to-many relationships, especially in the context of CloudKit-backed models. I ran into behavior that surprised me, and I’d like to confirm whether this is intended design or a potential issue / undocumented behavior. Minimal example import SwiftUI import SwiftData @Model class Node { var children: [Node]? = nil var parent: Node? = nil init(children: [Node]? = nil, parent: Node? = nil) { self.children = children self.parent = parent print(self.children == nil) } } struct ContentView: View { var body: some View { Button("Create") { _ = Node(children: nil) } } } Observed behavior If @Model is not used, children == nil prints true as expected. If @Model is used, children == nil prints false. Inspecting the macro expansion, it appears SwiftData initializes relationship storage using backing data placeholders and normalizes to-many relationships into empty collections at runtime, even when declared as optional. CloudKit context From the SwiftData + CloudKit documentation: “The iCloud servers don’t guarantee atomic processing of relationship changes, so CloudKit requires all relationships to be optional.” Because of this, modeling relationships as optional is required when syncing with CloudKit, even for to-many relationships. This is why I’m hesitant to simply switch the model to a non-optional [Node] = [], even though that would match the observed runtime behavior. Questions Is it intentional that optional to-many relationships in SwiftData are never nil at runtime, and instead materialize as empty collections? If so, is Optional<[Model]> effectively treated as [Model] for runtime access, despite being required for CloudKit compatibility? Is the defaultValue: nil in the generated Schema.PropertyMetadata intended only for schema/migration purposes rather than representing a possible runtime state? Is there a recommended modeling pattern for CloudKit-backed SwiftData models where relationships must be optional, but runtime semantics behave as non-optional? I’m mainly looking to ensure I’m aligning with SwiftData’s intended design and not relying on behavior that could change or break with CloudKit sync. Thanks in advance for any clarification!

I'm building a photo editing app with a token-based subscription system using RevenueCat and StoreKit. Users purchase subscriptions that grant tokens for AI generations. There are no user accounts, the app is fully anonymous. Currently, I generate an anonymous account ID via RevenueCat SDK and store it in iCloud Keychain. This allows users on the same iCloud account to restore both their subscription and token balance across devices. However, users on a different iCloud account can restore their subscription via Apple, but their token balance is lost because there's no way to link the anonymous IDs. The problem is that if a user switches iCloud accounts or gets a new device without the same iCloud, their purchased tokens are orphaned. The subscription restores fine through Apple, but the token balance tied to the old anonymous ID becomes inaccessible. I have a few constraints: no user accounts, no email or phone sign-in, must work across devices owned by the same person, and must comply with App Store guidelines. My questions are: Is iCloud Keychain the right tool for this, or is there a better approach? Would CloudKit with an anonymous record zone be more appropriate? Are there any recommended patterns for persisting consumable balances tied to anonymous users across device migrations? Any guidance would be appreciated.

I have a total of 100 asset packs associated with my app but I have archived 5 of them. Unfortunately I am now unable to upload any more asset packs (the reported error is "backgroundAsset limit reached -- This app has already reached the maximum number of active backgroundAssets. Maximum allowed is 100.") I assumed that archiving asset packs would make them inactive (and thus not count against the limit). This seems to not be the case and I'm not sure how I can upload new asset packs.

Hello, In my iOS/SwiftUI/SwiftData app, I want the user to be able to hit [Cancel] from editing in a detail screen and return to the previous screen without changes being saved. I believed that setting autosaveEnabled to false and/or calling .rollback would prevent changes from being saved, unless/until I call .save() when the user clicks [Save], but this does not seem to be correct. I set modelContext.autosaveEnabled = false and I call modelContext.rollback() when the user hits [Cancel], but any changes they made are not rolled back, but saved even if I don’t call save(). I have tried setting autosaveEnabled to false when I create the ModelContainer on a @MainActor function when the App starts, and in the detail/edit screen’s .onAppear(). I can see that .rollback is being called when the [Cancel] button is tapped. In all cases, any changes the user made before hitting [Cancel] are saved. The Developer Documentation on autosaveEnabled includes this: “The default value is false. SwiftData automatically sets this property to true for the model container’s mainContext." I am working on the mainContext, but it appears that setting autosaveEnabled to false has no effect no matter where in the code I set it. If someone sees what I am doing wrong, I’d sure appreciate the input. If this description doesn’t explain the problem well enough, I’ll develop a minimal focused example.

Greetings my fellow engineers, I use SwiftData in my iOS app. The schema is unversioned and consists of a single model. I've been modifying the model for almost two years now and relying on automatic database migrations. I had no problems for all that time, but now trying to add a property to the model or even remove a property from the model results in an error which seems like SwiftData is no longer capable of performing an automatic migration. The log console has things like the following: CoreData: error: NSUnderlyingError : Error Domain=NSCocoaErrorDomain Code=134190 "(null)" UserInfo={reason=Each property must have a unique renaming identifier} CoreData: error: reason : Can't find or automatically infer mapping model for migration CoreData: error: storeType: SQLite CoreData: error: configuration: default CoreData: annotation: options: CoreData: annotation: NSMigratePersistentStoresAutomaticallyOption : 1 CoreData: annotation: NSInferMappingModelAutomaticallyOption : 1 CoreData: annotation: NSPersistentStoreRemoteChangeNotificationOptionKey : 1 CoreData: annotation: NSPersistentHistoryTrackingKey : 1 CoreData: error: <NSPersistentStoreCoordinator: 0x7547b5480>: Attempting recovery from error encountered during addPersistentStore: 0x753f8d800 Error Domain=NSCocoaErrorDomain Code=134140 "Persistent store migration failed, missing mapping model." Have you ever encountered such an issue? What are my options?

Trying to support undo & redo in an app that utilizes Swift Data and as with anything other than provided simplistic Apple demo examples the experience is not great. The problem: Im trying to build functionality that allows users to add items to an item group, where item and item group have a many-to-many relationship e.g. item group can hold many items and items can appear in multiple groups. When trying to do so with relatively simple setup of either adding or removing item group from items relationship array, I am pretty consistently met with a hard crash after performing undo & redo. Sometimes it works the first few undo & redos but 95% of the time would crash on the first one. Could not cast value of type 'Swift.Optional<Any>' (0x20a676be0) to 'Swift.Array<App.CodableStructModel>' (0x207a2bc08). Where CodableStructModel is a Codable Value type inside Item. Adding and removing this relationship should be undoable & redoable as typical for Mac interaction and is "supported" by SwiftData by default, meaning that the developer has to actively either wholly opt out of undo support in their modelContainer setup or do it on a per action scale with the only thing I know of: modelContext.processPendingChanges() modelContext.undoManager?.disableUndoRegistration() ..... modelContext.processPendingChanges() modelContext.undoManager?.enableUndoRegistration() General rant on SwiftData: Random crashes, inconsistencies, random cryptic errors thrown by the debugger and general lack of production level stability. Each update breaks something new and there is very little guidance and communication from the Swift Data team on how to adapt and more importantly consideration for developers that have adopted Swift Data. If SwiftData is not ready for production, it would go a long way to clearly communicate that and mark it as Beta product.

CloudKit CKRecordZone Deletion Issue Problem: CloudKit record zones deleted via CKDatabase.modifyRecordZones(deleting:) or CKModifyRecordZonesOperation are successfully removed but then reappear. I suspect they are automatically reinstated by CloudKit sync, despite successful deletion confirmation. Environment: SwiftData with CloudKit integration Custom CloudKit zones created for legacy zone-based sharing Observed Behavior: Create custom zone (e.g., "TestZone1") via CKDatabase.modifyRecordZones(saving:) Copy records to zone for sharing purposes Delete zone using any CloudKit deletion API - returns success, no errors Immediate verification: Zone is gone from database.allRecordZones() After SwiftData/CloudKit sync or app restart: Zone reappears Reproduction: Tested with three different deletion methods - all exhibit same behaviour: modifyRecordZones(deleting:) async API CKModifyRecordZonesOperation (fire-and-forget) CKModifyRecordZonesOperation with result callbacks Zone deletion succeeds, change tokens (used to track updates to shared records) cleaned up But zones are restored presumably by CloudKit background sync Expected: Deleted zones should remain deleted Actual: Zones are reinstated, creating orphaned zones

Running Tahoe 26.1 in a virtual machine, I can't sign into my Apple account. There is an error message saying "Could not communicate with the server." Internet access otherwise seems to be working in the VM. I tried both UTM and VirtualBuddy. Is this supposed to work?

I am using SwiftData with CloudKit to synchronize data across multiple devices, and I have encountered an issue: occasionally, abnormal sync behavior occurs between two devices (it does not happen 100% of the time—only some users have reported this problem). It seems as if synchronization between the two devices completely stops; no matter what operations are performed on one end, the other end shows no response. After investigating, I suspect the issue might be caused by both devices simultaneously modifying the same field, which could lead to CloudKit's logic being unable to handle such conflicts and causing the sync to stall. Are there any methods to avoid or resolve this situation? Of course, I’m not entirely sure if this is the root cause. Has anyone encountered a similar issue?

Hi, I am implementing a premium feature in my app where CloudKit syncing is available only for "Pro" users. The Workflow: Free Users: I initialize the ModelContainer with cloudKitDatabase: .none so their data stays local. Pro Upgrade: When a user purchases a subscription, I restart the container with cloudKitDatabase: .automatic to enable syncing. The Problem: If a user starts as "Free" (creates local data) and later upgrades to "Pro", the app crashes immediately upon launch with the following error: Fatal error: Failed to create ModelContainer: SwiftDataError(_error: SwiftData.SwiftDataError._Error.loadIssueModelContainer, _explanation: nil) It seems that SwiftData fails to load the existing data once the configuration changes to expect a CloudKit-backed store. My Question: Is there a supported way to "toggle" CloudKit on for an existing local dataset without causing this crash? I want the user's existing local data to start syncing once they pay, but currently, it just crashes. My code: import Foundation import SwiftData public enum DataModelEnum: String { case task, calendar public static let container: ModelContainer = { let isSyncEnabled = UserDefaults.isProUser let config = ModelConfiguration( groupContainer: .identifier("group.com.yourcompany.myApp"), cloudKitDatabase: isSyncEnabled ? .automatic : .none ) do { return try ModelContainer(for: TaskModel.self, CalendarModel.self, configurations: config) } catch { fatalError("Failed to create ModelContainer: \(error)") } }() }

We are currently implementing a custom iCloud sync for our macOS and iOS apps using CloudKit. Syncing works fine as long as the number of record sends is relatively small. But when we test with a large number of changes ( 80,000+ CKRecords ) we start running into problems. Our sending strategy is very conservative to avoid rate limits: We send records sequentially in batches of 250 records With about 2 seconds pause between operations Records are small and contain no assets (assets are uploaded separately) At some point we start receiving: “Database commit size exceeds limit” After that, CloudKit begins returning rate-limit errors with retryAfter-Information in the error. We wait for the retry time and try again, but from this moment on, nothing progresses anymore. Every subsequent attempt fails. We could not find anything in the official documentation regarding such a “commit size” limit or what triggers this failure state. So my questions are: Are there undocumented limits on the total number of records that can exist in an iCloud database (private or shared)? Is there a maximum volume of record modifications a container can accept within a certain timeframe, even if operations are split into small batches with pauses? Is it possible that sending large numbers of records in a row can temporarily or permanently “stall” a CloudKit container? Any insights or experiences would be greatly appreciated. Thank you!

We have an unreleased SwiftData app for iOS18+. While we were testing I saw reports on the forum about unexpected database migrations for codable arrays on iOS26.1. I'd like to ask a couple of questions: 1- Does this issue originate from the new Xcode version, or is it specific to iOS 26.1? 2- Is it possible to change our attribute so that users on older iOS versions receive the same model, preventing a migration from being triggered when they upgrade to iOS 26.1? One of our models looks like this: struct Point: Codable, Hashable { let x: Int let y: Int } @Model class Grid { private(set) var gridId: String = "" var points: [Point] = [] var updatedAt: Date = Date() private(set) var createdAt: Date = Date() #Index<Grid>([\.gridId]) ... } I can think of some options like: // 1 @Attribute(.transformable(by: CustomJsonTransformer.self)) var points: [Point] = [] // 2 @Attribute(.externalStorage) var points: [Point] = [] // 3 var points: Data = Data() // store points as data However, I'm not sure which one to use. What would you recommend to handle this, or is there a better strategy you would suggest?

I'm experiencing the following error with my SwiftData container when running a build: Code=134504 "Cannot use staged migration with an unknown model version." Code Structure - Summary I am using a versionedSchema to store multiple models in SwiftData. I started experiencing this issue when adding two new models in the newest Schema version. Starting from the current public version, V4.4.6, there are two migrations. Migration Summary The first migration is to V4.4.7. This is a lightweight migration removing one attribute from one of the models. This was tested and worked successfully. The second migration is to V5.0.0. This is a custom migration adding two new models, and instantiating instances of the two new models based on data from instances of the existing models. In the initial testing of this version, no issues were observed. Issue and Steps to Reproduce Reproduction of issue: Starting from a fresh build of the publicly released V4.4.6, I run a new build that contains both Schema Versions (V4.4.7 and V5.0.0), and their associated migration stages. This builds successfully, and the container successfully migrates to V5.0.0. Checking the default.store file, all values appear to migrate and instantiate correctly. The second step in reproduction of the issue is to simply stop running the build, and then rebuild, without any code changes. This fails to initialize the model container every time afterwards. Going back to the simulator after successive builds are stopped in Xcode, the app launches and accesses/modifies the model container as normal. Supplementary Issue: I have been putting up with the same, persistent issue in the Xcode Preview Canvas of "Failed to Initialize Model Container" This is a 5 in 6 build issue, where builds will work at random. In the case of previews, I have cleared all data associated with all previews multiple times. The only difference being that the simulator is a 100% failure rate after the initial, successful initialization. I assume this is due to the different build structure of previews. Lastly, of note, the Xcode previews fail at the same line in instantiating the model container as the simulator does. From my research into this issue, people say that the Xcode preview is instantiating from elsewhere. I do have a separate model container set up specifically for canvas previews, but the error does not occur in that container, but rather the app's main container. Possible Contributing Factors & Tested Facts iOS: While I have experienced issues with SwiftData and the complier in iOS 26, I can rule that out as the issue here. This has been tested on simulators running iOS 18.6, 26.0.1, and 26.1, all encountering failures to initialize model container. While in iOS 18, subsequent builds after the successful migration did work, I did eventually encounter the same error and crash. In iOS 26.0.1 and 26.1, these errors come immediately on the second build. Container Initialization for V4.4.6 do { container = try ModelContainer( for: Job.self, JobTask.self, Day.self, Charge.self, Material.self, Person.self, TaskCategory.self, Service.self, migrationPlan: JobifyMigrationPlan.self ) } catch { fatalError("Failed to Initialize Model Container") } Versioned Schema Instance for V4.4.6 (V4.4.7 differs only by versionIdentifier) static var versionIdentifier = Schema.Version(4, 4, 6) static var models: [any PersistentModel.Type] { [Job.self, JobTask.self, Day.self, Charge.self, Material.self, Person.self, TaskCategory.self, Service.self] } Container Initialization for V5.0.0 do { let schema = Schema([Jobify.self, JobTask.self, Day.self, Charge.self, MaterialItem.self, Person.self, TaskCategory.self, Service.self, ServiceJob.self, RecurerRule.self]) container = try ModelContainer( for: schema, migrationPlan: JobifyMigrationPlan.self ) } catch { fatalError("Failed to Initialize Model Container") } Versioned Schema Instance for V5.0.0 static var versionIdentifier = Schema.Version(5, 0, 0) static var models: [any PersistentModel.Type] { [ JobifySchemaV500.Job.self, JobifySchemaV500.JobTask.self, JobifySchemaV500.Day.self, JobifySchemaV500.Charge.self, JobifySchemaV500.Material.self, JobifySchemaV500.Person.self, JobifySchemaV500.TaskCategory.self, JobifySchemaV500.Service.self, JobifySchemaV500.ServiceJob.self, JobifySchemaV500.RecurerRule.self ] } Addressing Differences in Object Names Type-aliasing: All my model types are type-aliased for simplification in view components. All types are aliased as 'JobifySchemeV446.<#Name#>' in V.4.4.6, and 'JobifySchemaV500.<#Name#>' in V5.0.0 Issues with iOS 26: My type-aliases dating back to iOS 17 overlapped with lower level objects in Swift, including 'Job' and 'Material'. These started to be an issue with initializing the model container when running in iOS 26. The type aliases have been renamed since, however the V4.4.6 build with the old names runs and builds perfectly fine in iOS 26 If there is any other code that may be relevant in determining where this error is occurring, I would be happy to add it. My current best theory is simply that I have mistakenly omitted code relevant to the SwiftData Migration.

Hello 👋, I encounter the "This model instance was invalidated because its backing data could no longer be found the store" crash with SwiftData. Which from what I understood means I try to access a model after it has been removed from the store (makes sense). I made a quick sample to reproduce/better understand because there some case(s) I can't figure it out. Let's take a concrete example, we have Home model and a Home can have many Room(s). // Sample code @MainActor let foo = Foo() // A single reference let database = Database(modelContainer: sharedModelContainer) // A single reference @MainActor class Foo { // Properties to explicilty keep reference of model(s) for the purpose of the POC var _homes = [Home]() var _rooms = [Room]() func fetch() async { let homes = await database.fetch().map { sharedModelContainer.mainContext.model(for: $0) as! Home } print(ObjectIdentifier(homes[0]), homes[0].rooms?.map(\.id)) // This will crash here or not. } // Same version of a delete function with subtle changes. // Depending on the one you use calling delete then fetch will result in a crash or not. // Keep a reference to only homes == NO CRASH func deleteV1() async { self._homes = await database.fetch().map { sharedModelContainer.mainContext.model(for: $0) as! Home } await database.delete() } // Keep a reference to only rooms == NO CRASH func deleteV2() async { self._rooms = await database.fetch().map { sharedModelContainer.mainContext.model(for: $0) as! Home }[0].rooms ?? [] await database.delete() } // Keep a reference to homes & rooms == CRASH 💥 func deleteV3() async { self._homes = await database.fetch().map { sharedModelContainer.mainContext.model(for: $0) as! Home } self._rooms = _homes[0].rooms ?? [] // or even only retain reference to rooms that have NOT been deleted 🤔 like here "id: 2" make it crash // self._rooms = _homes[0].rooms?.filter { r in r.id == "2" } ?? [] await database.delete() } } Calling deleteV() then fetch() will result in a crash or not depending on the scenario. I guess I understand deleteV1, deleteV2. In those case an unsaved model is served by the model(for:) API and accessing properties later on will resolve correctly. The doc says: "The identified persistent model, if known to the context; otherwise, an unsaved model with its persistentModelID property set to persistentModelID." But I'm not sure about deleteV3. It seems the ModelContext is kind of "aware" there is still cyclic reference between my models that are retained in my code so it will serve these instances instead when calling model(for:) API ? I see my home still have 4 rooms (instead of 2). So I then try to access rooms that are deleted and it crash. Why of that ? I mean why not returning home with two room like in deleteV1 ? Because SwiftData heavily rely on CoreData may be I miss a very simple thing here. If someone read this and have a clue for me I would be extremely graceful. PS: If someone wants to run it on his machine here's some helpful code: // Database let sharedModelContainer: ModelContainer = { let schema = Schema([ Home.self, Room.self, ]) let modelConfiguration = ModelConfiguration(schema: schema, isStoredInMemoryOnly: false) debugPrint(modelConfiguration.url.absoluteString.replacing("%20", with: "\\ ")) return try! ModelContainer(for: schema, configurations: [modelConfiguration]) }() extension Database { static let shared = Database(modelContainer: sharedModelContainer) } @ModelActor actor Database { func insert() async { let r1 = Room(id: "1", name: "R1") let r2 = Room(id: "2", name: "R2") let r3 = Room(id: "3", name: "R3") let r4 = Room(id: "4", name: "R4") let home = Home(id: "1", name: "My Home") home.rooms = [r1, r2, r3, r4] modelContext.insert(home) try! modelContext.save() } func fetch() async -> [PersistentIdentifier] { try! modelContext.fetchIdentifiers(FetchDescriptor<Home>()) } @MainActor func delete() async { let mainContext = sharedModelContainer.mainContext try! mainContext.delete( model: Room.self, where: #Predicate { r in r.id == "1" || r.id == "4" } ) try! mainContext.save() // 🤔 Calling fetch here seems to solve crash too, force home relationship to be rebuild correctly ? // let _ = try! sharedModelContainer.mainContext.fetch(FetchDescriptor<Home>()) } } // Models @Model class Home: Identifiable { @Attribute(.unique) public var id: String var name: String @Relationship(deleteRule: .cascade, inverse: \Room.home) var rooms: [Room]? init(id: String, name: String, rooms: [Room]? = nil) { self.id = id self.name = name self.rooms = rooms } } @Model class Room: Identifiable { @Attribute(.unique) public var id: String var name: String var home: Home? init(id: String, name: String, home: Home? = nil) { self.id = id self.name = name self.home = home } }