[Enhancement]Keep CallKit answer actions alive during WebRTC audio setup (#1081)

Author: ipavlidakis

CHANGELOG.md

@@ -15,6 +15,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 - Fix local mediaAdapters not reacting to changed own capabilities. [#1070](https://github.com/GetStream/stream-video-swift/pull/1070)
 - Fix label color when presenting. [#1077](https://github.com/GetStream/stream-video-swift/pull/1077)
 - Ensure CallKit push token updates and invalidation mutate `deviceToken` on the main actor to avoid Swift concurrency/actor-isolation issues. [#1076](https://github.com/GetStream/stream-video-swift/pull/1076)
+- Ensure CallKit joins keep the answer action completion alive until WebRTC has configured the audio device module. [#1081](https://github.com/GetStream/stream-video-swift/pull/1081)
 
 # [1.43.0](https://github.com/GetStream/stream-video-swift/releases/tag/1.43.0)
 _February 27, 2026_

Sources/StreamVideo/CallKit/CallKitService.swift

@@ -473,9 +473,14 @@ open class CallKitService: NSObject, CXProviderDelegate, @unchecked Sendable {
             }
 
             do {
-                /// Mark join source as `.callKit` for audio session.
-                ///
-                callToJoinEntry.call.state.joinSource = .callKit
+                // Pass a CallKit completion hook through the join flow so the
+                // WebRTC layer can release CallKit's audio session ownership as
+                // soon as it has configured the audio device module.
+                callToJoinEntry.call.state.joinSource = .callKit(.init {
+                    // Allow CallKit to hand audio session activation back to
+                    // the app before we continue configuring audio locally.
+                    action.fulfill()
+                })
 
                 try await callToJoinEntry.call.join(callSettings: callSettings)
                 action.fulfill()

Sources/StreamVideo/Models/JoinSource.swift

@@ -10,10 +10,39 @@ import Foundation
 /// the app's own UI or through a system-level interface such as CallKit.
 /// This helps distinguish the user's entry point and can be used to customize
 /// behavior or analytics based on how the call was initiated.
-enum JoinSource {
+enum JoinSource: Sendable, Equatable {
+    /// Carries the completion hook CallKit expects us to invoke once the SDK is
+    /// ready for CallKit to hand audio session ownership back to the app.
+    struct ActionCompletion: @unchecked Sendable {
+        fileprivate let identifier: UUID = .init()
+        private let completion: () -> Void
+
+        init(_ completion: @escaping () -> Void) {
+            self.completion = completion
+        }
+
+        /// Invokes the stored completion callback.
+        func complete() {
+            completion()
+        }
+    }
+
     /// Indicates that the call was joined from within the app's UI.
     case inApp
 
     /// Indicates that the call was joined via CallKit integration.
-    case callKit
+    case callKit(ActionCompletion)
+
+    /// Compares `JoinSource` values while treating CallKit sources as distinct
+    /// whenever they wrap different completion hooks.
+    static func == (lhs: JoinSource, rhs: JoinSource) -> Bool {
+        switch (lhs, rhs) {
+        case (.inApp, .inApp):
+            return true
+        case (.callKit(let lhs), .callKit(let rhs)):
+            return lhs.identifier == rhs.identifier
+        default:
+            return false
+        }
+    }
 }

Sources/StreamVideo/WebRTC/v2/WebRTCStateAdapter.swift

@@ -785,15 +785,21 @@ actor WebRTCStateAdapter: ObservableObject, StreamAudioSessionAdapterDelegate, W
         try await audioStore.dispatch([
             .setAudioDeviceModule(peerConnectionFactory.audioDeviceModule)
         ]).result()
-        
+
+        if case let .callKit(completion) = source {
+            // Let CallKit release its audio session ownership once WebRTC has
+            // the audio device module it needs.
+            completion.complete()
+        }
+
         audioSession.activate(
             callSettingsPublisher: $callSettings.removeDuplicates().eraseToAnyPublisher(),
             ownCapabilitiesPublisher: $ownCapabilities.removeDuplicates().eraseToAnyPublisher(),
             delegate: self,
             statsAdapter: statsAdapter,
             /// If we are joining from CallKit the AudioSession will be activated from it and we
             /// shouldn't attempt another activation.
-            shouldSetActive: source != .callKit
+            shouldSetActive: source == .inApp
         )
     }
 

StreamVideoTests/Call/Call_Tests.swift

@@ -587,16 +587,17 @@ final class Call_Tests: StreamVideoTestCase, @unchecked Sendable {
         let call = MockCall(.dummy(callController: mockCallController))
         call.stub(for: \.state, with: .init(.dummy()))
         mockCallController.stub(for: .join, with: JoinCallResponse.dummy())
+        let expectedJoinSource = JoinSource.callKit(.init {})
 
-        call.state.joinSource = .callKit
+        call.state.joinSource = expectedJoinSource
         _ = try await call.join()
 
         XCTAssertEqual(
             mockCallController.recordedInputPayload(
                 (Bool, CallSettings?, CreateCallOptions?, Bool, Bool, JoinSource).self,
                 for: .join
             )?.first?.5,
-            .callKit
+            expectedJoinSource
         )
     }
 

StreamVideoTests/Controllers/CallController_Tests.swift

@@ -99,6 +99,7 @@ final class CallController_Tests: StreamVideoTestCase, @unchecked Sendable {
     func test_joinCall_coordinatorTransitionsToConnecting() async throws {
         let callSettings = CallSettings(cameraPosition: .back)
         let options = CreateCallOptions(team: .unique)
+        let expectedJoinSource = JoinSource.callKit(.init {})
 
         try await assertTransitionToStage(
             .connecting,
@@ -113,7 +114,7 @@ final class CallController_Tests: StreamVideoTestCase, @unchecked Sendable {
                             options: options,
                             ring: true,
                             notify: true,
-                            source: .callKit
+                            source: expectedJoinSource
                         )
                 }
             }
@@ -122,7 +123,7 @@ final class CallController_Tests: StreamVideoTestCase, @unchecked Sendable {
             XCTAssertEqual(expectedStage.options?.team, options.team)
             XCTAssertTrue(expectedStage.ring)
             XCTAssertTrue(expectedStage.notify)
-            XCTAssertEqual(expectedStage.context.joinSource, .callKit)
+            XCTAssertEqual(expectedStage.context.joinSource, expectedJoinSource)
             await self.assertEqualAsync(
                 await self
                     .mockWebRTCCoordinatorFactory

StreamVideoTests/WebRTC/v2/WebRTCCoorindator_Tests.swift

@@ -87,6 +87,7 @@ final class WebRTCCoordinator_Tests: XCTestCase, @unchecked Sendable {
             startsAt: .init(timeIntervalSince1970: 100),
             team: .unique
         )
+        let expectedJoinSource = JoinSource.callKit(.init {})
 
         try await assertTransitionToStage(
             .connecting,
@@ -98,7 +99,7 @@ final class WebRTCCoordinator_Tests: XCTestCase, @unchecked Sendable {
                         options: expectedOptions,
                         ring: true,
                         notify: true,
-                        source: .callKit
+                        source: expectedJoinSource
                     )
             }
         ) { stage in
@@ -111,7 +112,7 @@ final class WebRTCCoordinator_Tests: XCTestCase, @unchecked Sendable {
             XCTAssertEqual(expectedStage.options?.team, expectedOptions.team)
             XCTAssertTrue(expectedStage.ring)
             XCTAssertTrue(expectedStage.notify)
-            XCTAssertEqual(expectedStage.context.joinSource, .callKit)
+            XCTAssertEqual(expectedStage.context.joinSource, expectedJoinSource)
             await self.assertEqualAsync(
                 await self.subject.stateAdapter.initialCallSettings,
                 expectedCallSettings

StreamVideoTests/WebRTC/v2/WebRTCStateAdapter_Tests.swift

@@ -651,6 +651,25 @@ final class WebRTCStateAdapter_Tests: XCTestCase, @unchecked Sendable {
         }
     }
 
+    func test_configureAudioSession_callKitSource_completesActionCompletion() async throws {
+        let completionExpectation = expectation(
+            description: "CallKit action completion invoked."
+        )
+        let sfuStack = MockSFUStack()
+        sfuStack.setConnectionState(to: .connected(healthCheckInfo: .init()))
+        await subject.set(sfuAdapter: sfuStack.adapter)
+        let ownCapabilities = Set<OwnCapability>([OwnCapability.blockUsers])
+        await subject.enqueueOwnCapabilities { ownCapabilities }
+
+        try await subject.configureAudioSession(
+            source: .callKit(.init {
+                completionExpectation.fulfill()
+            })
+        )
+
+        await safeFulfillment(of: [completionExpectation], timeout: 1)
+    }
+
     // MARK: - cleanUp
 
     func test_cleanUp_shouldResetProperties() async throws {