fix(ci): Update Live Session integration tests (#15664)

Author: daymxn

FirebaseAI/Tests/TestApp/Tests/Integration/LiveSessionTests.swift

@@ -48,10 +48,7 @@ struct LiveSessionTests {
       ),
     ]),
   ]
-  private let textConfig = LiveGenerationConfig(
-    responseModalities: [.text]
-  )
-  private let audioConfig = LiveGenerationConfig(
+  private let generationConfig = LiveGenerationConfig(
     responseModalities: [.audio],
     outputAudioTranscription: AudioTranscriptionConfig()
   )
@@ -76,8 +73,8 @@ struct LiveSessionTests {
       role: "system",
       parts: """
       When you receive a message, if the message is a single word, assume it's the first name of a \
-      person, and call the getLastName tool to get the last name of said person. Only respond with \
-      the last name.
+      person, and call the getLastName tool to get the last name of said person. Once you get the \
+      response, say the response.
       """.trimmingCharacters(in: .whitespacesAndNewlines)
     )
 
@@ -95,7 +92,7 @@ struct LiveSessionTests {
                                                        modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: audioConfig,
+      generationConfig: generationConfig,
       systemInstruction: SystemInstructions.helloGoodbye
     )
 
@@ -119,15 +116,12 @@ struct LiveSessionTests {
     #expect(modelResponse == "goodbye")
   }
 
-  @Test(
-    .disabled("Temporarily disabled"),
-    .bug("https://github.com/firebase/firebase-ios-sdk/issues/15640"),
-    arguments: arguments
-  )
-  func sendVideoRealtime_receiveText(_ config: InstanceConfig, modelName: String) async throws {
+  @Test(arguments: arguments)
+  func sendVideoRealtime_receiveAudioOutputTranscripts(_ config: InstanceConfig,
+                                                       modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: textConfig,
+      generationConfig: generationConfig,
       systemInstruction: SystemInstructions.animalInVideo
     )
 
@@ -152,7 +146,7 @@ struct LiveSessionTests {
     await session.sendAudioRealtime(audioFile.data)
     await session.sendAudioRealtime(Data(repeating: 0, count: audioFile.data.count))
 
-    let text = try await session.collectNextTextResponse()
+    let text = try await session.collectNextAudioOutputTranscript()
 
     await session.close()
     let modelResponse = text
@@ -164,15 +158,11 @@ struct LiveSessionTests {
     #expect(["kitten", "cat", "kitty"].contains(modelResponse))
   }
 
-  @Test(
-    .disabled("Temporarily disabled"),
-    .bug("https://github.com/firebase/firebase-ios-sdk/issues/15640"),
-    arguments: arguments
-  )
+  @Test(arguments: arguments)
   func realtime_functionCalling(_ config: InstanceConfig, modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: textConfig,
+      generationConfig: generationConfig,
       tools: tools,
       systemInstruction: SystemInstructions.lastNames
     )
@@ -200,11 +190,9 @@ struct LiveSessionTests {
         functionId: functionCall.functionId
       ),
     ])
-
-    var text = try await session.collectNextTextResponse()
+    var text = try await session.collectNextAudioOutputTranscript()
     if text.isEmpty {
-      // The model sometimes sends an empty text response first
-      text = try await session.collectNextTextResponse()
+      text = try await session.collectNextAudioOutputTranscript()
     }
 
     await session.close()
@@ -217,8 +205,6 @@ struct LiveSessionTests {
   }
 
   @Test(
-    .disabled("Temporarily disabled"),
-    .bug("https://github.com/firebase/firebase-ios-sdk/issues/15640"),
     arguments: arguments.filter {
       // TODO: (b/450982184) Remove when Vertex AI adds support for Function IDs and Cancellation
       switch $0.0.apiConfig.service {
@@ -233,7 +219,7 @@ struct LiveSessionTests {
                                              modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: textConfig,
+      generationConfig: generationConfig,
       tools: tools,
       systemInstruction: SystemInstructions.lastNames
     )
@@ -266,7 +252,7 @@ struct LiveSessionTests {
   func realtime_interruption(_ config: InstanceConfig, modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: audioConfig
+      generationConfig: generationConfig
     )
 
     let audioFile = try #require(
@@ -295,23 +281,23 @@ struct LiveSessionTests {
     }
   }
 
-  @Test(
-    .disabled("Temporarily disabled"),
-    .bug("https://github.com/firebase/firebase-ios-sdk/issues/15640"),
-    arguments: arguments
-  )
+  @Test(arguments: arguments)
   func incremental_works(_ config: InstanceConfig, modelName: String) async throws {
     let model = FirebaseAI.componentInstance(config).liveModel(
       modelName: modelName,
-      generationConfig: textConfig,
+      generationConfig: generationConfig,
       systemInstruction: SystemInstructions.yesOrNo
     )
 
     let session = try await model.connect()
     await session.sendContent("Does five plus")
     await session.sendContent(" five equal ten?", turnComplete: true)
 
-    let text = try await session.collectNextTextResponse()
+    var text = try await session.collectNextAudioOutputTranscript()
+    if text.isEmpty {
+      // The model sometimes sends an empty text response first
+      text = try await session.collectNextAudioOutputTranscript()
+    }
 
     await session.close()
     let modelResponse = text
@@ -339,26 +325,6 @@ struct LiveSessionTests {
 }
 
 private extension LiveSession {
-  /// Collects the text that the model sends for the next turn.
-  ///
-  /// Will listen for `LiveServerContent` messages from the model,
-  /// incrementally keeping track of any `TextPart`s it sends. Once
-  /// the model signals that its turn is complete, the function will return
-  /// a string concatenated of all the `TextPart`s.
-  func collectNextTextResponse() async throws -> String {
-    var text = ""
-
-    for try await content in responsesOf(LiveServerContent.self) {
-      text += content.modelTurn?.allText() ?? ""
-
-      if content.isTurnComplete {
-        break
-      }
-    }
-
-    return text
-  }
-
   /// Collects the audio output transcripts that the model sends for the next turn.
   ///
   /// Will listen for `LiveServerContent` messages from the model,
@@ -395,11 +361,7 @@ private extension LiveSession {
       case let .toolCall(toolCall):
         return toolCall
       case let .content(content):
-        if let text = content.modelTurn?.allText() {
-          error += text
-        } else {
-          error += content.outputAudioText()
-        }
+        error += content.outputAudioText()
 
         if content.isTurnComplete {
           Issue.record("The model didn't send a tool call. Text received: \(error)")
@@ -464,16 +426,6 @@ private struct NoInterruptionError: Error,
   var description: String { "The model never sent an interrupted message." }
 }
 
-private extension ModelContent {
-  /// A collection of text from all parts.
-  ///
-  /// If this doesn't contain any `TextPart`, then an empty
-  /// string will be returned instead.
-  func allText() -> String {
-    parts.compactMap { ($0 as? TextPart)?.text }.joined()
-  }
-}
-
 extension LiveServerContent {
   /// Text of the output `LiveAudioTranscript`, or an empty string if it's missing.
   func outputAudioText() -> String {