feat: 为语音服务添加模式参数支持

- 在 ASR 和 TTS 服务接口中添加模式参数，支持 normal、phone_call、push_to_talk 等模式
- 修改翻译控制器，根据当前翻译模式设置对应的 TTS 和 ASR 模式
- 重构 iOS 音频会话配置逻辑，优化不同模式下的音频路由和蓝牙设备处理
- 统一 TTS 模式设置方法名，将 setPhoneCallMode 重命名为 setTtsMode

lib/data/services/asr_service.dart

@@ -22,6 +22,7 @@ abstract class AsrService {
   Future<bool> startContinuousRecognition(
     bool audioSourceType, {
     bool isRemoveFirstPunctuation = true,
+    String mode = "normal",
   });
 
   /// 停止连续语音识别

lib/data/services/speech_impl/azure_asr_service.dart

@@ -207,6 +207,7 @@ class AzureAsrService extends GetxService implements AsrService {
   Future<bool> startContinuousRecognition(
     bool audioSourceType, {
     bool isRemoveFirstPunctuation = true,
+    String mode = "normal",
   }) async {
     // 防抖判断：短时间内重复调用直接拦截
     final now = DateTime.now();
@@ -229,6 +230,7 @@ class AzureAsrService extends GetxService implements AsrService {
           await _channel.invokeMethod('startContinuousRecognition', {
         'audioSourceType': audioSourceType,
         'isRemoveFirstPunctuation': isRemoveFirstPunctuation,
+        'mode': mode,
       });
 
       if (!result) {

lib/data/services/speech_impl/azure_tts_service.dart

@@ -182,6 +182,20 @@ class AzureTtsService extends GetxService implements TtsService {
     }
   }
 
+  @override
+  Future<bool> setTtsMode(String mod) async {
+    if (!_isInitialized) await initialize();
+    try {
+      final result = await _channel.invokeMethod('setTtsMode', {
+        'mod': mod,
+      });
+      return result;
+    } catch (e) {
+      Logger.error('设置TTS模式失败: ${e.toString()}');
+      return false;
+    }
+  }
+
   @override
   Future<bool> setVoiceFlocking(String speakerProfileId) async {
     final result = await _channel.invokeMethod('setVoiceFlocking', {

lib/data/services/speech_impl/voice_clone_tts_service.dart

@@ -94,6 +94,18 @@ class VoiceCloneTtsService extends GetxService implements TtsService {
     }
   }
 
+  @override
+  Future<bool> setTtsMode(String mod) async {
+    if (!_isInitialized) await initialize();
+    try {
+      Logger.info('设置TTS模式: $mod');
+      return true;
+    } catch (e) {
+      Logger.error('设置TTS模式失败: ${e.toString()}');
+      return false;
+    }
+  }
+
   @override
   Future<bool> setVoiceFlocking(String speakerProfileId) async {
     return true;

lib/data/services/speech_impl/volcano_asr_api_service.dart

@@ -865,6 +865,7 @@ class VolcanoAsrApiService implements AsrService {
   Future<bool> startContinuousRecognition(
     bool audioSourceType, {
     bool isRemoveFirstPunctuation = true,
+    String mode = "normal",
   }) {
     // TODO: implement startContinuousRecognition
     throw UnimplementedError();

lib/data/services/speech_impl/volcano_asr_service.dart

@@ -400,6 +400,7 @@ class VolcanoAsrService extends GetxService implements AsrService {
   Future<bool> startContinuousRecognition(
     bool audioSourceType, {
     bool isRemoveFirstPunctuation = true,
+    String mode = "normal",
   }) {
     // TODO: implement startContinuousRecognition
     throw UnimplementedError();

lib/data/services/speech_impl/volcano_tts_service.dart

@@ -185,6 +185,23 @@ class VolcanoTtsService extends GetxService implements TtsService {
     }
   }
 
+  @override
+  Future<bool> setTtsMode(String mod) async {
+    if (!await _ensureInitialized()) return false;
+
+    try {
+      final result = await _channel.invokeMethod<bool>('setTtsMode', {
+            'mod': mod,
+          }) ??
+          false;
+
+      return result;
+    } catch (e) {
+      debugPrint('设置TTS模式失败：$e');
+      return false;
+    }
+  }
+
   @override
   Future<bool> setVoiceFlocking(String speakerProfileId) async {
     return true;

lib/data/services/tts_service.dart

@@ -19,6 +19,9 @@ abstract class TtsService {
   /// 设置语音
   Future<bool> setVoice(String voiceName);
 
+  /// 设置TTS模式
+  Future<bool> setTtsMode(String mod);
+
   /// 设置语音复刻
   Future<bool> setVoiceFlocking(String speakerProfileId);
 

lib/modules/translation/controllers/translation_controller.dart

@@ -344,6 +344,15 @@ class TranslationController extends GetxController with WidgetsBindingObserver {
       } catch (_) {}
     }
 
+    //设置播报的模式
+    if (mode == 'simultaneous') {
+      await _ttsService.setTtsMode('phone_call');
+    } else if (mode == 'faceToFace') {
+      await _ttsService.setTtsMode('push_to_talk');
+    } else {
+      await _ttsService.setTtsMode('normal');
+    }
+
     // 检查悬浮窗支持情况
     if (isFloatingWindowEnabled.value && !_isFloatingWindowSupportedMode()) {
       // 当前模式不支持悬浮窗，自动禁用
@@ -437,6 +446,7 @@ class TranslationController extends GetxController with WidgetsBindingObserver {
       await _asrService.startContinuousRecognition(
         _audioSourceType,
         isRemoveFirstPunctuation: false,
+        mode: 'push_to_talk',
       ); //开启识别
       _startAsrActiveTracking(); //开启识别活动跟踪
       if (isRecording.value) {
@@ -946,7 +956,13 @@ class TranslationController extends GetxController with WidgetsBindingObserver {
             !bleManager.isCodecActive) {
           return;
         }
-        await _startAsrService();
+        if (currentMode.value == 'simultaneous') {
+          await _startAsrService('push_to_talk');
+        } else if (currentMode.value == 'faceToFace') {
+          await _startAsrService('phone_call');
+        } else {
+          await _startAsrService('normal');
+        }
         await _finalizeRecognitionStart();
       }
     } catch (e) {
@@ -1045,8 +1061,12 @@ class TranslationController extends GetxController with WidgetsBindingObserver {
   }
 
   /// 启动语音识别
-  Future<void> _startAsrService() async {
-    await _asrService.startContinuousRecognition(_audioSourceType);
+  Future<void> _startAsrService(String mode) async {
+    await _asrService.startContinuousRecognition(
+      _audioSourceType,
+      isRemoveFirstPunctuation: false,
+      mode: mode,
+    );
   }
 
   /// 完成识别启动

local_plugins/agent_service/ios/agent_service/Sources/agent_service/AgentServiceImpl.swift

@@ -526,7 +526,7 @@ class AgentServiceImpl: NSObject {
        
         // 设置当前识别模式
         currentRecognitionMode = mode
-        os_log("设置语音识别模式: %{public}@", log: logger, type: .info, mode)
+        os_log("liwei--------------- 设置语音识别模式: %{public}@", log: logger, type: .info, mode)
 
         let audioSourceType: AzureAsrHelper.AudioSourceType = useBle ? .external : .microphone
 
@@ -724,7 +724,7 @@ class AgentServiceImpl: NSObject {
             return
         }
         _ = azureTtsHelper?.setSpeechParams(rate: ttsRatePercent, pitch: 0, volume: 100)
-        _ = azureTtsHelper?.setPhoneCallMode(mod: currentRecognitionMode)
+        _ = azureTtsHelper?.setTtsMode(mod: currentRecognitionMode)
         let userMessage = chatApiService.createUserMessage(content: text)
         processWithChatApiServiceInternal(sessionid: sessionid,userMessage: userMessage, displayText: text, speakResponse: speakResponse)
     }
@@ -1810,7 +1810,7 @@ class ChatApiStreamCallback: StreamCallback {
 
             responseBuilder += token
             if speakResponse && reply && broadcast{
-                _ = agentService.azureTtsHelper?.setPhoneCallMode(mod:agentService.currentRecognitionMode)
+//                _ = agentService.azureTtsHelper?.setTtsMode(mod:agentService.currentRecognitionMode)
                 try agentService.azureTtsHelper?.speakStream(sessionid:sessionid,token)
                 // 在开始流式TTS时立即停止气泡音
             }
@@ -1836,7 +1836,7 @@ class ChatApiStreamCallback: StreamCallback {
              }
             
             if speakResponse && reply && broadcast && sessionid == agentService.currsessionId{
-                _ = agentService.azureTtsHelper?.setPhoneCallMode(mod: agentService.currentRecognitionMode)
+                // _ = agentService.azureTtsHelper?.setTtsMode(mod: agentService.currentRecognitionMode)
                 agentService.azureTtsHelper?.flushStream(sessionid:sessionid)
             }
             
@@ -1888,7 +1888,7 @@ class ChatApiStreamCallback: StreamCallback {
                 "message": message
             ])
             if(code == 2001){
-                _ = agentService.azureTtsHelper?.setPhoneCallMode(mod: agentService.currentRecognitionMode)
+                // _ = agentService.azureTtsHelper?.setTtsMode(mod: agentService.currentRecognitionMode)
                 agentService.azureTtsHelper?.speakStream(sessionid: sessionid,agentService.insufficientIntegralText)
             }
             agentService.isAiStreaming = false
@@ -1932,7 +1932,7 @@ class ChatApiStreamCallback: StreamCallback {
          
             if (iscallingTool) {
                 os_log("收到函数调用: callingToolText:%{public}@", log: agentService.logger, type: .info, agentService.callingToolText)
-                _ = agentService.azureTtsHelper?.setPhoneCallMode(mod: agentService.currentRecognitionMode)
+                // _ = agentService.azureTtsHelper?.setTtsMode(mod: agentService.currentRecognitionMode)
                 agentService.azureTtsHelper?.speakStream(sessionid: sessionid,agentService.callingToolText)
                 iscallingTool = false
             }

local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureSpeechPlugin.swift

@@ -382,6 +382,7 @@ private func sendAstEvent(_ event: [String: Any]) {
             }
             
             let removeFirstPunctuation = args["isRemoveFirstPunctuation"] as? Bool ?? true
+            let mode = args["mode"] as? String ?? "normal"
             
             // 根据参数确定音频源类型
             let audioSourceType = isExternalActive ? 
@@ -392,6 +393,7 @@ private func sendAstEvent(_ event: [String: Any]) {
             
             // 启动连续识别
             let success = azureAsrHelper.startContinuousRecognition(
+                mod: mode,
                 audioSourceType: audioSourceType,
                 isRemoveFirstPunctuation: removeFirstPunctuation
             )
@@ -631,19 +633,14 @@ private func sendAstEvent(_ event: [String: Any]) {
             
             let success = azureTtsHelper.setVoice(voiceName)
             result(success)
-
-        case "setSpeechParams":
-            guard let args = call.arguments as? [String: Any] else {
+        case "setTtsMode":
+            guard let args = call.arguments as? [String: Any],
+                  let mod = args["mod"] as? String else {
                 result(FlutterError(code: "INVALID_ARGUMENTS", message: "参数不能为空", details: nil))
                 return
             }
-            
-            let rate = args["rate"] as? Int ?? 0
-            let pitch = args["pitch"] as? Int ?? 0
-            let volume = args["volume"] as? Int ?? 100
-            let success = azureTtsHelper.setSpeechParams(rate: rate, pitch: pitch, volume: volume)
+            let success = azureTtsHelper.setTtsMode(mod: mod)
             result(success)
-            
         case "speakText":
             guard let args = call.arguments as? [String: Any],
                   let sessionid = args["sessionid"] as? String,

local_plugins/azure_speech/ios/azure_speech/Sources/azure_speech/AzureTtsHelper.swift

@@ -415,14 +415,14 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
     }
 
 
-    public func setPhoneCallMode(mod: String) -> Bool {
-        // os_log("[liwei] AzureTtsHelper: setPhoneCallMode: %{public}@", log: log, type: .debug, mod)
+    public func setTtsMode(mod: String) -> Bool {
+        os_log("[liwei] AzureTtsHelper: setTtsMode: %{public}@", log: log, type: .debug, mod)
         self.previousRecognitionMode = self.currentRecognitionMode
         self.currentRecognitionMode = mod
         self.isHeadphones = detectHeadphonesOutput()
-        if isInitialized {
-            safeConfigureAudioSessionForTTS()
-        }
+        // if isInitialized {
+        //     safeConfigureAudioSessionForTTS()
+        // }
         return true
     }
     /**
@@ -438,6 +438,7 @@ public class AzureTtsHelper: NSObject, ITtsService, AVAudioPlayerDelegate {
             out.portType == .bluetoothA2DP ||
             out.portType == .bluetoothHFP ||
             out.portType == .bluetoothLE ||
+            out.portType == .headsetMic ||
             out.portType == .headphones
         }
     }
@@ -1376,6 +1377,18 @@ private var audioSessionObserversAdded = false
     private var routeChangeDebounceWorkItem: DispatchWorkItem?
     private var isApplyingAudioSessionConfig = false
 
+private func isTtsBusy() -> Bool {
+    if speaking { return true }
+    if isPlaying { return true }
+    if !playbackQueue.isEmpty { return true }
+    if pendingTextCount > 0 { return true }
+    if !pendingTasks.isEmpty { return true }
+    if activeStreamSynthesisCount > 0 { return true }
+    if !pcmPendingData.isEmpty { return true }
+    if scheduledBufferCount > 0 { return true }
+    return false
+}
+
 /**
  * 确保音频会话队列已设置 specific key（用于避免同队列 sync 死锁）
  * - Parameters: 无
@@ -1388,6 +1401,12 @@ private func ensureAudioSessionQueueSpecificKeySet() {
     audioSessionQueue.setSpecific(key: audioSessionQueueKey, value: 1)
 }
 
+/**
+ * 同步应用 TTS 音频会话配置（用于需要立即生效的播放/引擎初始化路径）
+ * - Parameters: 无
+ * - Returns: 是否成功激活音频会话
+ * - Throws: 无（内部捕获 AVAudioSession 相关异常并记录日志）
+ */
 /**
  * 同步应用 TTS 音频会话配置（用于需要立即生效的播放/引擎初始化路径）
  * - Parameters: 无
@@ -1397,7 +1416,6 @@ private func ensureAudioSessionQueueSpecificKeySet() {
 private func applyAudioSessionForTTS() -> Bool {
     var success = true
     ensureAudioSessionQueueSpecificKeySet()
-    
     let applyBody = {
         if self.isApplyingAudioSessionConfig { return }
         self.isApplyingAudioSessionConfig = true
@@ -1405,69 +1423,62 @@ private func applyAudioSessionForTTS() -> Bool {
 
         let audioSession = AVAudioSession.sharedInstance()
         let outputs: [AVAudioSessionPortDescription] = audioSession.currentRoute.outputs
-        let hasBluetoothHFP = outputs.contains(where: { $0.portType == .bluetoothHFP })
         let hasWiredHeadphones = outputs.contains(where: { $0.portType == .headphones || $0.portType == .headsetMic })
-        let hasBluetoothA2DP = outputs.contains(where: { $0.portType == .bluetoothA2DP })
-        let hasHeadphones = hasWiredHeadphones || hasBluetoothA2DP
+        let hasBluetoothA2DP = outputs.contains(where: { $0.portType == .bluetoothA2DP || $0.portType == .bluetoothLE })
+        let hasBluetoothHFP = outputs.contains(where: { $0.portType == .bluetoothHFP })
+        let hasHeadphones = hasWiredHeadphones || hasBluetoothA2DP || hasBluetoothHFP
         let otherAudioPlaying = audioSession.isOtherAudioPlaying
         let isRecording = (self.micCapture?.isCapturing == true)
+        self.isHeadphones = hasHeadphones
 
         let desiredCategory: AVAudioSession.Category
         let desiredMode: AVAudioSession.Mode
         var desiredOptions: AVAudioSession.CategoryOptions = []
 
-        let isHeadphonesConnected = outputs.contains(where: { $0.portType == .headphones || $0.portType == .headsetMic || $0.portType == .bluetoothA2DP })
-        os_log("[liwei] AzureTtsHelper: applyAudioSessionForTTS: %{public}@", log: self.log, type: .debug, self.currentRecognitionMode)
-        if self.currentRecognitionMode == "phone_call" || self.currentRecognitionMode == "push_to_talk" {
+
+        if self.currentRecognitionMode == "phone_call" {
             desiredCategory = .playAndRecord
-            if isHeadphonesConnected {
-                desiredMode = .voiceChat
-                if #available(iOS 18.0, *) {
-                    desiredOptions = [.allowBluetooth, .duckOthers]
-                } else {
-                    desiredOptions = [.allowBluetooth, .allowBluetoothA2DP]
+            if hasHeadphones {
+                desiredMode = .default
+                desiredOptions = [.allowBluetoothA2DP]
+                if let inputs = audioSession.availableInputs, let builtInMic = inputs.first(where: { $0.portType == .builtInMic }) {
+                    try? audioSession.setPreferredInput(builtInMic)
                 }
             } else {
-                desiredMode = .videoChat
+                desiredMode = .voiceChat
                 desiredOptions = [.allowBluetooth, .defaultToSpeaker]
             }
-        } else if self.isTelephonyActive() {
+            if otherAudioPlaying {
+                desiredOptions.insert(.duckOthers)
+            }
+        } else if self.currentRecognitionMode == "push_to_talk" || self.currentRecognitionMode == "normal" || self.currentRecognitionMode == "ble_wakeup" {
+            if isRecording {
+                self.micCapture?.stopCapture()
+                Thread.sleep(forTimeInterval: 0.05)
+            }
             desiredCategory = .playback
-            desiredMode = .default
+            desiredMode = .spokenAudio
+            if otherAudioPlaying {
                 desiredOptions.insert(.duckOthers)
+            }
         } else if isRecording {
             desiredCategory = .playAndRecord
-            if self.isHeadphones {
             desiredMode = .default
-                desiredOptions.insert(.allowBluetooth)
+            if hasHeadphones {
                 desiredOptions.insert(.allowBluetoothA2DP)
-                if otherAudioPlaying {
-                    desiredOptions.insert(.duckOthers)
+                if let inputs = audioSession.availableInputs, let builtInMic = inputs.first(where: { $0.portType == .builtInMic }) {
+                    try? audioSession.setPreferredInput(builtInMic)
                 }
             } else {
-                desiredMode = .videoChat
                 desiredOptions.insert(.allowBluetooth)
-                desiredOptions.insert(.mixWithOthers)
                 desiredOptions.insert(.defaultToSpeaker)
             }
-        } else if hasBluetoothHFP {
-            desiredCategory = .playback
-            desiredMode = .voicePrompt
-            desiredOptions = [.interruptSpokenAudioAndMixWithOthers, .allowBluetooth]
-        } else if hasHeadphones {
-            desiredCategory = .playback
-            desiredMode = .default
             if otherAudioPlaying {
                 desiredOptions.insert(.duckOthers)
             }
         } else {
             desiredCategory = .playback
-            desiredMode = .default
-            if #available(iOS 18.0, *) {
-                desiredOptions = []
-            } else {
-                desiredOptions = [.defaultToSpeaker]
-            }
+            desiredMode = .spokenAudio
             if otherAudioPlaying {
                 desiredOptions.insert(.duckOthers)
             }
@@ -1483,7 +1494,7 @@ private func applyAudioSessionForTTS() -> Bool {
                         _ = try? audioSession.setPreferredIOBufferDuration(0.02)
                     }
                     try audioSession.setActive(true)
-                    if self.currentRecognitionMode == "phone_call", !self.isHeadphones {
+                    if self.currentRecognitionMode == "phone_call", !hasHeadphones {
                         _ = try? audioSession.overrideOutputAudioPort(.speaker)
                     }
                     break
@@ -1551,16 +1562,6 @@ private func isTelephonyActive() -> Bool {
     return false
 }
 
-/**
- * 安全配置音频会话用于 TTS
- * - 播放优先走媒体声道（A2DP），避免切换到通话声道（HFP）
- * - 外部音乐播放时：不启用 mix，默认会打断外部音频；同时可 duck 确保可听见
- * - 录音进行中：保持 playAndRecord + A2DP 输出 + 内置麦输入，避免切到 HFP
- * - 通话中：保持保守策略，避免触发系统断言或破坏通话链路
- * - Parameters: 无
- * - Returns: 无
- * - Throws: 无（内部捕获 AVAudioSession 相关异常并记录日志）
- */
 
     /// 为 TTS 安全配置音频会话：
     /// - 使用异步队列，避免在同一串行队列上 sync 导致死锁
@@ -1568,23 +1569,10 @@ private func isTelephonyActive() -> Bool {
     /// - 外部音频播放时启用 spoken audio 策略，确保能在 A2DP 上播放
     /// - 仅当类别或模式发生变化时才 setCategory，降低 AVAudioSessionRouteChangeReason.categoryChange 的触发概率
     private func safeConfigureAudioSessionForTTS() {
-    os_log("[liwei] AzureTtsHelper: safeConfigureAudioSessionForTTS - recognitionMode: %{public}@", log: log, type: .debug, currentRecognitionMode)
-            // 使用异步，避免串行队列重入自锁
+        if !isTtsBusy() { return }
         audioSessionQueue.async { [weak self] in
             guard let self = self else { return }
-
-                let audioSession = AVAudioSession.sharedInstance()
-
-                // 显式指定 outputs 的类型，确保闭包参数类型可推断
-                let outputs: [AVAudioSessionPortDescription] = audioSession.currentRoute.outputs
-
-                // 使用 contains(where:) 明确谓词闭包。
-                // 保持使用 audioSessionQueue.async，防止在串行队列上自锁。
-                // 仅在类别或模式发生变化时调用 setCategory，减少 categoryChange 的回调风暴。
-                let hasBluetoothHFP = outputs.contains(where: { $0.portType == .bluetoothHFP })
-                _ = hasBluetoothHFP
-                _ = audioSession
-
+            if !self.isTtsBusy() { return }
             _ = self.applyAudioSessionForTTS()
         }
     }
@@ -1617,8 +1605,6 @@ private func deactivateAudioSessionAfterTTSIfIdle() {
             self.playerNode = nil
             self.pcmFormat = nil
         }
-        let session = AVAudioSession.sharedInstance()
-        _ = try? session.setActive(false, options: .notifyOthersOnDeactivation)
     }
     if DispatchQueue.getSpecific(key: audioSessionQueueKey) != nil {
         work()
@@ -1645,8 +1631,10 @@ private func deactivateAudioSessionAfterTTSIfIdle() {
         os_log("音频中断开始（可能为通话/Siri），停止TTS", log: log, type: .info)
     case .ended:
         audioInterrupted = false
+        if isTtsBusy() {
             os_log("音频中断结束，重新配置会话", log: log, type: .info)
             safeConfigureAudioSessionForTTS()
+        }
     @unknown default:
         break
     }
@@ -1660,42 +1648,55 @@ private func deactivateAudioSessionAfterTTSIfIdle() {
  * - 配置过程中（isApplyingAudioSessionConfig = true）直接忽略
  */
 @objc private func handleRouteChange(_ notification: Notification) {
-    guard let info = notification.userInfo,
-          let reasonValue = info[AVAudioSessionRouteChangeReasonKey] as? UInt,
-          let reason = AVAudioSession.RouteChangeReason(rawValue: reasonValue) else { return }
-
-    os_log("音频路由变化: %{public}@", log: log, type: .info, String(describing: reason))
-
-    // 配置中触发的路由变化，直接忽略，避免重入导致死循环
-    if isApplyingAudioSessionConfig {
-        os_log("配置进行中，忽略本次路由变化", log: log, type: .debug)
-        return
-    }
-
-    // 忽略会造成循环的原因：categoryChange（你的日志 reason=3），以及 routeConfigurationChange 等
-    switch reason {
-    case .categoryChange, .routeConfigurationChange, .wakeFromSleep, .noSuitableRouteForCategory:
-        os_log("忽略路由变化原因: %{public}@", log: log, type: .debug, String(describing: reason))
-        return
-    default:
-        break
-    }
-
-    // 通话期间避免重配，避免与电话路由冲突
-    if isTelephonyActive() {
-        os_log("通话期间检测到路由变化，跳过重配", log: log, type: .info)
-        return
-    }
+    // guard let info = notification.userInfo,
+    //       let reasonValue = info[AVAudioSessionRouteChangeReasonKey] as? UInt,
+    //       let reason = AVAudioSession.RouteChangeReason(rawValue: reasonValue) else { return }
+
+    // if !isTtsBusy() { return }
+    // os_log("音频路由变化: %{public}@", log: log, type: .info, String(describing: reason))
+
+    // // 配置中触发的路由变化，直接忽略，避免重入导致死循环
+    // if isApplyingAudioSessionConfig {
+    //     os_log("配置进行中，忽略本次路由变化", log: log, type: .debug)
+    //     return
+    // }
+
+    // // 忽略会造成循环的原因：categoryChange（你的日志 reason=3），以及 routeConfigurationChange 等
+    // switch reason {
+    // case .categoryChange, .routeConfigurationChange, .wakeFromSleep, .noSuitableRouteForCategory:
+    //     os_log("忽略路由变化原因: %{public}@", log: log, type: .debug, String(describing: reason))
+    //     return
+    // default:
+    //     break
+    // }
+
+    // // 通话期间避免重配，避免与电话路由冲突
+    // if isTelephonyActive() {
+    //     os_log("通话期间检测到路由变化，跳过重配", log: log, type: .info)
+    //     return
+    // }
+
+    // let isActivelyPlayingNow =
+    //     speaking ||
+    //     isPlaying ||
+    //     (audioPlayer?.isPlaying == true) ||
+    //     (playerNode?.isPlaying == true)
+    // if isActivelyPlayingNow {
+    //     routeChangeDebounceWorkItem?.cancel()
+    //     _ = stop()
+    //     os_log("路由变化时正在播报，已停止以避免音源冲突", log: log, type: .info)
+    //     return
+    // }
 
     // 去抖处理：300ms 内只执行最后一次重配
     // 忽略 categoryChange 等
     // 去抖：在 audioSessionQueue 上延迟调用
-    routeChangeDebounceWorkItem?.cancel()
-    let work = DispatchWorkItem { [weak self] in
-        self?.safeConfigureAudioSessionForTTS()
-    }
-    routeChangeDebounceWorkItem = work
-    audioSessionQueue.asyncAfter(deadline: .now() + 0.3, execute: work)
+    // routeChangeDebounceWorkItem?.cancel()
+    // let work = DispatchWorkItem { [weak self] in
+    //     self?.safeConfigureAudioSessionForTTS()
+    // }
+    // routeChangeDebounceWorkItem = work
+    // audioSessionQueue.asyncAfter(deadline: .now() + 0.3, execute: work)
 }
 
 /**

local_plugins/azure_speech/ios/azure_speech/Sources/tools/MicrophoneCapture.swift

@@ -92,8 +92,7 @@ public class MicrophoneCapture: NSObject {
         }
         
         do {
-            try audioSession.setPreferredSampleRate(sampleRate)
-            try audioSession.setPreferredIOBufferDuration(0.005) // 5ms缓冲
+            try audioSession.setPreferredIOBufferDuration(0.02)
             print("音频会话配置成功")
         } catch {
             print("音频会话配置失败: \(error.localizedDescription)")
@@ -306,15 +305,18 @@ public class MicrophoneCapture: NSObject {
             sampleRate = audioSession.sampleRate
         }
 
-        // iOS 13+ 启用语音处理（可能在部分路由/模式下失败，失败不应阻断录音启动）
         if #available(iOS 13.0, *) {
+            let session = self.audioSession ?? AVAudioSession.sharedInstance()
+            let outputs = session.currentRoute.outputs
+            let hasBluetoothA2DPOutput = outputs.contains(where: { $0.portType == .bluetoothA2DP || $0.portType == .bluetoothLE })
+            let shouldEnableVoiceProcessing = (session.category == .playAndRecord) && (session.mode == .voiceChat || session.mode == .videoChat) && !hasBluetoothA2DPOutput
             do {
-                try audioInputNode.setVoiceProcessingEnabled(true)
+                try audioInputNode.setVoiceProcessingEnabled(shouldEnableVoiceProcessing)
             } catch {
                 if let nsError = error as NSError? {
-                    print("启用语音处理失败: domain=\(nsError.domain) code=\(nsError.code) desc=\(nsError.localizedDescription)")
+                    print("语音处理设置失败(enable=\(shouldEnableVoiceProcessing)): domain=\(nsError.domain) code=\(nsError.code) desc=\(nsError.localizedDescription)")
                 } else {
-                    print("启用语音处理失败: \(error.localizedDescription)")
+                    print("语音处理设置失败(enable=\(shouldEnableVoiceProcessing)): \(error.localizedDescription)")
                 }
             }
         }
@@ -514,14 +516,7 @@ public class MicrophoneCapture: NSObject {
         // 修复：使用安全的清理方法
         cleanupAudioEngine()
         
-        // 安全地处理音频会话
-        guard let audioSession = audioSession else { return }
-        do {
-            try audioSession.setActive(false, options: .notifyOthersOnDeactivation)
-        } catch {
-            print("音频会话配置失败: \(error.localizedDescription)")
-        }
-        
+        // 安全地处理音频会话：这里不主动 deactivate，避免与后续播报会话切换打架导致卡顿/失败
     }
     
     /**

local_plugins/azure_speech/ios/azure_speech/Sources/tools/SimpleAudioReceiver.swift

@@ -166,25 +166,20 @@ public class SimpleAudioReceiver: NSObject {
 
         if audioSourceType == .microphone || isHybridExternalMode {
             // --- 麦克风输入逻辑 (包括混合模式) ---
-            desiredCategory = .playAndRecord
-            
             if currentRecognitionMode == "phone_call" {
-                desiredMode = .voiceChat
                 if isHeadphones {
-                    // 修正：增加 A2DP 选项以支持高质量蓝牙音频播放
-                    desiredOptions = [.allowBluetooth, .allowBluetoothA2DP]
+                    desiredCategory = .playAndRecord
+                    desiredMode = .default
+                    desiredOptions = [.allowBluetoothA2DP]
                 } else {
-                    desiredOptions = [.allowBluetooth, .defaultToSpeaker]
-                }
-            } else { // normal, push_to_talk, ble_wakeup 等
-                if isHeadphones {
-                    // 修正：使用 .voiceChat 模式以避免 .videoChat 强制外放
+                    desiredCategory = .playAndRecord
                     desiredMode = .voiceChat
-                    desiredOptions = [.allowBluetooth, .allowBluetoothA2DP, .mixWithOthers]
-                } else {
-                    desiredMode = .videoChat
-                    desiredOptions = [.mixWithOthers, .defaultToSpeaker]
+                    desiredOptions = [.allowBluetooth, .defaultToSpeaker]
                 }
+            } else { // normal, push_to_talk, ble_wakeup 等：仅录音，不需要同时播报
+                desiredCategory = .record
+                desiredMode = .default
+                desiredOptions = []
             }
             
             // 如果在真实通话中，强制使用更适合的录音设置
@@ -197,8 +192,8 @@ public class SimpleAudioReceiver: NSObject {
         } else if audioSourceType == .external {
             // --- 纯外部源 (仅播放) 逻辑 ---
             desiredCategory = .playback
-            desiredMode = .videoChat
-            desiredOptions = isHeadphones ? [.allowBluetoothA2DP, .mixWithOthers] : [.mixWithOthers, .defaultToSpeaker]
+            desiredMode = .spokenAudio
+            desiredOptions = isHeadphones ? [.allowBluetoothA2DP, .mixWithOthers] : [.mixWithOthers]
         } else {
             // 兜底，理论上不会执行
             desiredCategory = .playAndRecord
@@ -217,11 +212,13 @@ public class SimpleAudioReceiver: NSObject {
         // --- 路由设置 ---
         if audioSourceType == .microphone || isHybridExternalMode {
             // 为需要麦克风的场景设置输入输出
+            if desiredCategory == .playAndRecord {
                 if isHeadphones {
                     try audioSession.overrideOutputAudioPort(.none) // 允许耳机/蓝牙输出
                 } else {
                     try audioSession.overrideOutputAudioPort(.speaker) // 强制外放
                 }
+            }
 
             // 设置首选输入设备
             if let inputs = audioSession.availableInputs {
@@ -248,7 +245,7 @@ public class SimpleAudioReceiver: NSObject {
                 }
             }
         } else { // 纯外部源
-            try audioSession.overrideOutputAudioPort(.none)
+            // .overrideOutputAudioPort 仅对 .playAndRecord 有效，.playback 会触发 -50
         }
     }
     
@@ -280,7 +277,8 @@ public class SimpleAudioReceiver: NSObject {
             out.portType == .bluetoothA2DP ||
             out.portType == .bluetoothHFP ||
             out.portType == .bluetoothLE ||
-            out.portType == .headphones
+            out.portType == .headphones ||
+            out.portType == .headsetMic
         }
     }