Add support for SpeechConfig

Author: daymxn

FirebaseAI/Sources/AILog.swift

@@ -72,7 +72,6 @@ enum AILog {
     case liveSessionFailedToSendClientMessage = 3021
     case liveSessionUnexpectedResponse = 3022
 
-
     // SDK State Errors
     case generateContentResponseNoCandidates = 4000
     case generateContentResponseNoText = 4001
@@ -81,7 +80,6 @@ enum AILog {
     case invalidWebsocketURL = 4004
     case duplicateLiveSessionSetupComplete = 4005
 
-
     // SDK Debugging
     case loadRequestStreamResponseLine = 5000
   }

FirebaseAI/Sources/Types/Internal/AppCheck.swift

@@ -15,7 +15,7 @@
 import FirebaseAppCheckInterop
 
 // TODO: document
-internal extension AppCheckInterop {
+extension AppCheckInterop {
   // TODO: Document
   func fetchAppCheckToken(limitedUse: Bool,
                           domain: String) async throws -> FIRAppCheckTokenResultInterop {

FirebaseAI/Sources/Types/Internal/Live/LiveSessionService.swift

@@ -100,7 +100,8 @@ actor LiveSessionService {
 
   /// Start a new connection to the backend.
   ///
-  /// Seperated into its own function to make it easier to surface a way to call it seperately when resuming the same session.
+  /// Seperated into its own function to make it easier to surface a way to call it seperately when
+  /// resuming the same session.
   public func connect() {
     setupTask.cancel()
     setupTask = Task { [weak self] in
@@ -197,8 +198,8 @@ actor LiveSessionService {
           } else if let liveMessage = LiveServerMessage.tryFrom(response) {
             responseContinuation.yield(liveMessage)
           } else {
-            // we don't raise an error, since this allows us to add support internally but not publicly
-            // we still log it in debug though, in case it's not expected
+            // we don't raise an error, since this allows us to add support internally but not
+            // publicly we still log it in debug though, in case it's not expected
             AILog.debug(
               code: .liveSessionUnsupportedMessage,
               "The server sent a message that we don't currently have a mapping for: \(response)"

FirebaseAI/Sources/Types/Internal/Live/SpeechConfig.swift

@@ -0,0 +1,29 @@
+// Copyright 2025 Google LLC
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+import Foundation
+
+/// Speech generation config.
+struct SpeechConfig: Encodable, Sendable {
+  /// The configuration for the speaker to use.
+  let voiceConfig: VoiceConfig
+
+  /// Language code (ISO 639. e.g. en-US) for the speech synthesization.
+  let languageCode: String?
+
+  init(voiceConfig: VoiceConfig, languageCode: String?) {
+    self.voiceConfig = voiceConfig
+    self.languageCode = languageCode
+  }
+}

FirebaseAI/Sources/Types/Internal/Live/VoiceConfig.swift

@@ -0,0 +1,66 @@
+// Copyright 2025 Google LLC
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+import Foundation
+
+/// Configuration for the speaker to use.
+enum VoiceConfig {
+  /// Configuration for the prebuilt voice to use.
+  case prebuiltVoiceConfig(PrebuiltVoiceConfig)
+
+  /// Configuration for the custom voice to use.
+  case customVoiceConfig(CustomVoiceConfig)
+}
+
+/// The configuration for the prebuilt speaker to use.
+///
+/// Not just a string on the parent proto, because there'll likely be a lot
+/// more options here.
+struct PrebuiltVoiceConfig: Encodable, Sendable {
+  /// The name of the preset voice to use.
+  let voiceName: String
+
+  init(voiceName: String) {
+    self.voiceName = voiceName
+  }
+}
+
+/// The configuration for the custom voice to use.
+struct CustomVoiceConfig: Encodable, Sendable {
+  /// The sample of the custom voice, in pcm16 s16e format.
+  let customVoiceSample: Data
+
+  init(customVoiceSample: Data) {
+    self.customVoiceSample = customVoiceSample
+  }
+}
+
+// MARK: - Encodable conformance
+
+extension VoiceConfig: Encodable {
+  enum CodingKeys: CodingKey {
+    case prebuiltVoiceConfig
+    case customVoiceConfig
+  }
+
+  func encode(to encoder: any Encoder) throws {
+    var container = encoder.container(keyedBy: CodingKeys.self)
+    switch self {
+    case let .prebuiltVoiceConfig(setup):
+      try container.encode(setup, forKey: .prebuiltVoiceConfig)
+    case let .customVoiceConfig(clientContent):
+      try container.encode(clientContent, forKey: .customVoiceConfig)
+    }
+  }
+}

FirebaseAI/Sources/Types/Public/Live/LiveGenerationConfig.swift

@@ -14,7 +14,6 @@
 
 import Foundation
 
-// TODO: add support for SpeechConfig
 /// A struct defining model parameters to be used when sending generative AI
 /// requests to the backend model.
 @available(iOS 15.0, macOS 12.0, macCatalyst 15.0, tvOS 15.0, watchOS 8.0, *)
@@ -43,6 +42,9 @@ public struct LiveGenerationConfig: Sendable {
   /// Supported modalities of the response.
   let responseModalities: [ResponseModality]?
 
+  /// Controls the voice of the model during conversation.
+  let speechConfig: SpeechConfig?
+
   /// Creates a new `GenerationConfig` value.
   ///
   /// See the
@@ -122,10 +124,13 @@ public struct LiveGenerationConfig: Sendable {
   ///     > Warning: Specifying response modalities is a **Public Preview** feature, which means
   ///     > that it is not subject to any SLA or deprecation policy and could change in
   ///     > backwards-incompatible ways.
+  ///   - speechConfig: Controls the voice of the model, when streaming `audio` via
+  ///     ``ResponseModality``.
   public init(temperature: Float? = nil, topP: Float? = nil, topK: Int? = nil,
               candidateCount: Int? = nil, maxOutputTokens: Int? = nil,
               presencePenalty: Float? = nil, frequencyPenalty: Float? = nil,
-              responseModalities: [ResponseModality]? = nil) {
+              responseModalities: [ResponseModality]? = nil,
+              speechConfig: LiveSpeechConfig? = nil) {
     // Explicit init because otherwise if we re-arrange the above variables it changes the API
     // surface.
     self.temperature = temperature
@@ -136,6 +141,7 @@ public struct LiveGenerationConfig: Sendable {
     self.presencePenalty = presencePenalty
     self.frequencyPenalty = frequencyPenalty
     self.responseModalities = responseModalities
+    self.speechConfig = speechConfig?.speechConfig
   }
 }
 
@@ -152,5 +158,6 @@ extension LiveGenerationConfig: Encodable {
     case presencePenalty
     case frequencyPenalty
     case responseModalities
+    case speechConfig
   }
 }

FirebaseAI/Sources/Types/Public/Live/LiveServerMessage.swift

@@ -25,7 +25,8 @@ public struct LiveServerMessage: Sendable {
     /// Request for the client to execute the provided functions.
     case toolCall(LiveServerToolCall)
 
-    /// Notification for the client that a previously issued ``LiveServerToolCall`` should be cancelled.
+    /// Notification for the client that a previously issued ``LiveServerToolCall`` should be
+    /// cancelled.
     case toolCallCancellation(LiveServerToolCallCancellation)
 
     /// Server will disconnect soon.
@@ -35,14 +36,14 @@ public struct LiveServerMessage: Sendable {
   /// The actual message sent from the server.
   public var messageType: MessageType
 
-  /// 
+  // TODO: document
   public var usageMetadata: GenerateContentResponse.UsageMetadata? { serverMessage.usageMetadata }
 }
 
 // MARK: - Internal parsing
 
 @available(iOS 15.0, macOS 12.0, macCatalyst 15.0, tvOS 15.0, watchOS 8.0, *)
-internal extension LiveServerMessage {
+extension LiveServerMessage {
   static func tryFrom(_ serverMessage: BidiGenerateContentServerMessage) -> Self? {
     guard let messageType = LiveServerMessage.MessageType.tryFrom(serverMessage.messageType) else {
       return nil
@@ -53,7 +54,7 @@ internal extension LiveServerMessage {
 }
 
 @available(iOS 15.0, macOS 12.0, macCatalyst 15.0, tvOS 15.0, watchOS 8.0, *)
-internal extension LiveServerMessage.MessageType {
+extension LiveServerMessage.MessageType {
   static func tryFrom(_ serverMessage: BidiGenerateContentServerMessage.MessageType) -> Self? {
     return switch serverMessage {
     case .setupComplete:

FirebaseAI/Sources/Types/Public/Live/LiveSpeechConfig.swift

@@ -0,0 +1,46 @@
+// Copyright 2025 Google LLC
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//      http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+import Foundation
+
+/// Configuration for controlling the voice of the model during conversation.
+@available(iOS 15.0, macOS 12.0, macCatalyst 15.0, tvOS 15.0, watchOS 8.0, *)
+public struct LiveSpeechConfig: Sendable {
+  let speechConfig: SpeechConfig
+
+  init(_ speechConfig: SpeechConfig) {
+    self.speechConfig = speechConfig
+  }
+
+  /// Creates a new `LiveSpeechConfig` value.
+  ///
+  /// - Parameters:
+  ///   - voiceName: The name of the prebuilt voice to be used for the model's speech response.
+  ///
+  ///     To learn more about the available voices, see the docs on
+  ///     [Voice options](https://ai.google.dev/gemini-api/docs/speech-generation#voices)\.
+  ///   - languageCode: ISO-639 language code to use when parsing text sent from the client, instead
+  ///     of audio. By default, the model will attempt to detect the input language automatically.
+  ///
+  ///     To learn which codes are supported, see the docs on
+  ///     [Supported languages](https://ai.google.dev/gemini-api/docs/speech-generation#languages)\.
+  public init(voiceName: String, languageCode: String? = nil) {
+    self.init(
+      SpeechConfig(
+        voiceConfig: .prebuiltVoiceConfig(.init(voiceName: voiceName)),
+        languageCode: languageCode
+      )
+    )
+  }
+}