feat: 改造语音的实现逻辑

src/constant/speech.js

@@ -1,6 +1,7 @@
 // App 端 yao-asdRealSpeech 使用的阿里云 DashScope 实时语音识别配置。
 // 将 apikey 填成实际的 DashScope API Key 后，App 端语音识别即可发起连接。
 export const appSpeechRecognitionOptions = {
-  apikey: "SnoHqdtJ832riRg4",
+  apikey: "sk-2cab1c221b4b47749119d33ab991360a",
   language_hints: ["zh"],
+  saveAudioFile: false,
 };

src/pages/ChatMain/ChatInputArea/index.vue

@@ -1,6 +1,10 @@
 <template>
-  <view class="input-area-wrapper">
+  <view
-    <view v-if="!visibleWaveBtn" class="area-input">
+    class="input-area-wrapper"
+    @touchend="handleVoiceTouchEndFromContainer"
+    @touchcancel="handleVoiceTouchEndFromContainer"
+  >
+    <view class="area-input">
       <!-- 语音/键盘切换 -->
       <view
         v-if="isSpeechRecognitionSupported"
@@ -51,11 +55,16 @@
         <view
           v-if="isVoiceMode"
           class="hold-to-talk-button"
-          @longpress="handleVoiceTouchStart"
+          @touchstart="handleVoiceTouchStart"
           @touchend="handleVoiceTouchEnd"
           @touchcancel="handleVoiceTouchEnd"
         >
-          按住 说话
+          <RecordingWaveBtn
+            v-if="visibleWaveBtn"
+            class="recording-wave-inline"
+            ref="recordingWaveBtnRef"
+          />
+          <text v-else>按住 说话</text>
         </view>
       </view>
 
@@ -71,12 +80,10 @@
       </view>
     </view>
 
-    <!-- 录音按钮 -->
-    <RecordingWaveBtn v-if="visibleWaveBtn" ref="recordingWaveBtnRef" />
-
     <!-- #ifdef APP-PLUS -->
     <yao-asdRealSpeech
-      v-if="isSpeechRecognitionSupported"
+      v-if="isSpeechRecognitionSupported && appSpeechVisible"
+      :key="appSpeechKey"
       ref="appSpeechRef"
       :options="appSpeechOptions"
       @result="handleAppSpeechResult"
@@ -98,7 +105,7 @@ import { appSpeechRecognitionOptions } from "@/constant/speech";
 
 let manager = null;
 let speechProvider = "";
-const isSpeechRecognitionEnabled = ref(false);
+const isSpeechRecognitionEnabled = ref(true);
 const isSpeechRecognitionSupported = ref(false);
 const appSpeechOptions = appSpeechRecognitionOptions;
 
@@ -134,6 +141,8 @@ const emit = defineEmits([
 const textareaRef = ref(null);
 const recordingWaveBtnRef = ref(null);
 const appSpeechRef = ref(null);
+const appSpeechKey = ref(0);
+const appSpeechVisible = ref(true);
 const placeholder = computed(() => {
   const config = getCurrentConfig();
   return `快告诉${config.name}您在想什么~`;
@@ -144,7 +153,10 @@ const keyboardHeight = ref(0);
 const isVoiceMode = ref(false);
 const visibleWaveBtn = ref(false);
 const isRecording = ref(false);
+const isVoicePressing = ref(false);
 const appRecognizedText = ref("");
+const isAppSpeechStarting = ref(false);
+const isAppWaitingToSend = ref(false);
 let watchDogTimer = null;
 let appStopFallbackTimer = null;
 let hasSentAppRecognition = false;
@@ -168,7 +180,12 @@ const resetUI = () => {
 const startWatchDog = (timeout = 10000) => {
   if (watchDogTimer) clearTimeout(watchDogTimer);
   watchDogTimer = setTimeout(() => {
+    watchDogTimer = null;
     console.warn("recording watchdog triggered, forcing UI reset");
+    if (hasActiveVoiceRecognition()) {
+      stopActiveVoiceRecognition({ shouldSend: true });
+      return;
+    }
     resetUI();
   }, timeout);
 };
@@ -180,6 +197,48 @@ const clearAppStopFallback = () => {
   }
 };
 
+const showRecordingUI = () => {
+  isRecording.value = true;
+  visibleWaveBtn.value = true;
+
+  nextTick(() => {
+    if (recordingWaveBtnRef.value) {
+      recordingWaveBtnRef.value.startAnimation();
+    }
+  });
+  startWatchDog(10000);
+};
+
+const resetAppSpeechComponent = () => {
+  if (speechProvider !== "app") return;
+
+  isAppSpeechStarting.value = false;
+  appSpeechVisible.value = false;
+  nextTick(() => {
+    appSpeechKey.value += 1;
+    appSpeechVisible.value = true;
+  });
+};
+
+const getPendingAppSpeechText = () => {
+  return (appRecognizedText.value || inputMessage.value || "").trim();
+};
+
+const scheduleAppRecognizedTextSend = (timeout = 1200) => {
+  isAppWaitingToSend.value = true;
+  clearAppStopFallback();
+
+  if (getPendingAppSpeechText()) {
+    sendAppRecognizedText();
+    return;
+  }
+
+  appStopFallbackTimer = setTimeout(() => {
+    appStopFallbackTimer = null;
+    sendAppRecognizedText();
+  }, timeout);
+};
+
 // 保持和父组件同步
 watch(
   () => props.modelValue,
@@ -199,17 +258,24 @@ watch(inputMessage, (val) => {
 // 切换语音/文本模式
 const toggleVoiceMode = () => {
   if (!isSpeechRecognitionSupported.value) return;
+  if (isVoiceMode.value) {
+    stopActiveVoiceRecognition({ shouldSend: false });
+  }
   isVoiceMode.value = !isVoiceMode.value;
 };
 
-// 处理语音按钮长按开始
+// 处理语音按钮按下开始
 const handleVoiceTouchStart = () => {
   if (!isSpeechRecognitionSupported.value) return;
+  if (isRecording.value || isAppSpeechStarting.value || visibleWaveBtn.value) return;
+
+  isVoicePressing.value = true;
 
   try {
     if (speechProvider === "wechat") {
       if (!manager) return;
       manager.start({ lang: "zh_CN" });
+      showRecordingUI();
     } else if (speechProvider === "app") {
       if (!appSpeechOptions.apikey) {
         uni.showToast({
@@ -220,10 +286,15 @@ const handleVoiceTouchStart = () => {
       }
 
       appRecognizedText.value = "";
+      inputMessage.value = "";
+      isAppWaitingToSend.value = false;
       hasSentAppRecognition = false;
+      clearAppStopFallback();
+      isAppSpeechStarting.value = true;
 
       const appSpeech = appSpeechRef.value;
       if (!appSpeech || typeof appSpeech.start !== "function") {
+        isAppSpeechStarting.value = false;
         uni.showToast({
           title: "语音组件未初始化",
           icon: "none",
@@ -235,31 +306,65 @@ const handleVoiceTouchStart = () => {
     } else {
       return;
     }
-
-    isRecording.value = true;
-    visibleWaveBtn.value = true;
-
-    // 启动音频条动画
-    nextTick(() => {
-      if (recordingWaveBtnRef.value) {
-        recordingWaveBtnRef.value.startAnimation();
-      }
-    });
-    startWatchDog(10000);
   } catch (err) {
     console.error("record start error:", err);
+    isAppSpeechStarting.value = false;
     // 保底清理
     resetUI();
   }
 };
 
-// 处理语音按钮长按结束
+const hasActiveVoiceRecognition = () => {
-const handleVoiceTouchEnd = () => {
+  return (
+    isVoicePressing.value ||
+    isRecording.value ||
+    isAppSpeechStarting.value ||
+    visibleWaveBtn.value
+  );
+};
+
+const stopActiveVoiceRecognition = ({ shouldSend = true } = {}) => {
+  isVoicePressing.value = false;
+
   if (!isSpeechRecognitionSupported.value) {
     resetUI();
     return;
   }
 
+  if (speechProvider === "app") {
+    const wasStarting = isAppSpeechStarting.value;
+    const wasRecording = isRecording.value;
+
+    if (!wasStarting && !wasRecording) {
+      if (!shouldSend) {
+        isAppWaitingToSend.value = false;
+        clearAppStopFallback();
+        resetAppSpeechComponent();
+      }
+      resetUI();
+      return false;
+    }
+
+    isAppSpeechStarting.value = false;
+
+    try {
+      appSpeechRef.value?.stop?.();
+      if (wasRecording && shouldSend) {
+        scheduleAppRecognizedTextSend();
+      } else {
+        isAppWaitingToSend.value = false;
+        clearAppStopFallback();
+        resetAppSpeechComponent();
+      }
+    } catch (err) {
+      console.error("record stop error:", err);
+      resetAppSpeechComponent();
+    } finally {
+      resetUI();
+    }
+    return true;
+  }
+
   // 如果本地状态不是录音中，也确保 UI 恢复
   if (!isRecording.value) {
     if (recordingWaveBtnRef.value) {
@@ -270,26 +375,28 @@ const handleVoiceTouchEnd = () => {
       clearTimeout(watchDogTimer);
       watchDogTimer = null;
     }
-    return;
+    return false;
   }
 
   try {
-    if (speechProvider === "wechat") {
+    manager?.stop?.();
-      manager?.stop?.();
-    } else if (speechProvider === "app") {
-      appSpeechRef.value?.stop?.();
-      clearAppStopFallback();
-      appStopFallbackTimer = setTimeout(() => {
-        appStopFallbackTimer = null;
-        sendAppRecognizedText();
-      }, 600);
-    }
   } catch (err) {
     console.error("record stop error:", err);
   } finally {
     // 无论 stop 是否抛错，都保证 UI 恢复
     resetUI();
   }
+  return true;
+};
+
+// 处理语音按钮长按结束
+const handleVoiceTouchEnd = () => {
+  stopActiveVoiceRecognition({ shouldSend: true });
+};
+
+const handleVoiceTouchEndFromContainer = () => {
+  if (!hasActiveVoiceRecognition()) return;
+  stopActiveVoiceRecognition({ shouldSend: true });
 };
 
 // 处理发送原语音
@@ -336,46 +443,87 @@ const getAppSpeechText = (res) => {
 };
 
 const sendAppRecognizedText = () => {
-  if (hasSentAppRecognition) return;
+  if (hasSentAppRecognition) return false;
 
-  const text = appRecognizedText.value.trim();
+  const text = getPendingAppSpeechText();
   if (!text) {
     console.log("没有说话");
-    return;
+    return false;
   }
 
   hasSentAppRecognition = true;
+  isAppWaitingToSend.value = false;
   clearAppStopFallback();
   inputMessage.value = text;
   emit("send", text);
   appRecognizedText.value = "";
+  return true;
 };
 
 const handleAppSpeechResult = (res) => {
+  if (hasSentAppRecognition) return;
+
   const text = getAppSpeechText(res);
   if (!text) return;
 
   appRecognizedText.value = text;
   inputMessage.value = text;
+
+  if (isAppWaitingToSend.value) {
+    sendAppRecognizedText();
+  }
 };
 
 const handleAppSpeechChange = (msg) => {
   if (!msg || !msg.status) return;
 
   if (msg.status === "START") {
-    isRecording.value = true;
+    isAppSpeechStarting.value = false;
+    if (!isVoicePressing.value) {
+      appSpeechRef.value?.stop?.();
+      resetAppSpeechComponent();
+      resetUI();
+      return;
+    }
+
+    showRecordingUI();
     return;
   }
 
   if (msg.status === "STOP") {
+    isAppSpeechStarting.value = false;
+    resetUI();
+    if (!sendAppRecognizedText() && !isAppWaitingToSend.value) {
+      resetAppSpeechComponent();
+    }
+    return;
+  }
+
+  if (msg.status === "FINISH") {
+    isAppSpeechStarting.value = false;
     resetUI();
     sendAppRecognizedText();
+    resetAppSpeechComponent();
+    return;
+  }
+
+  if (msg.status === "CLOSE") {
+    isAppSpeechStarting.value = false;
+    resetUI();
+    if (isAppWaitingToSend.value) {
+      sendAppRecognizedText();
+    }
+    resetAppSpeechComponent();
     return;
   }
 
   if (msg.status === "ERROR") {
     console.error("app speech recognition error", msg.msg);
+    isAppSpeechStarting.value = false;
+    isAppWaitingToSend.value = false;
+    clearAppStopFallback();
     resetUI();
+    resetAppSpeechComponent();
   }
 };
 
@@ -413,6 +561,9 @@ onUnmounted(() => {
     }
   }
   clearAppStopFallback();
+  isAppSpeechStarting.value = false;
+  isAppWaitingToSend.value = false;
+  isVoicePressing.value = false;
   resetUI();
 });
 

src/pages/ChatMain/ChatInputArea/styles/index.scss

@@ -44,6 +44,13 @@
     -webkit-user-select: none;
   }
 
+  .recording-wave-inline {
+    width: 100%;
+    margin: 0;
+    margin-bottom: 0;
+    box-shadow: none;
+  }
+
   .input-container-send {
     display: flex;
     align-items: center;

src/uni_modules/yao-asdRealSpeech/components/yao-asdRealSpeech/yao-asdRealSpeech.vue

@@ -16,13 +16,15 @@
 		},
 		methods:{
 			start(){
+				this.task_id='';
 				this.webSocekt();
 			},
 			stop(){
-				this.recorderStatus="stop";
+				this.recorderStatus={
-				if(ws){
+					status:"stop",
-					ws.close();
+					saveAudioFile: !this.options || this.options.saveAudioFile !== false,
-				}
+					time: Date.now()
+				};
 			},
 			toShowToast(){
 				uni.showToast({
@@ -31,9 +33,53 @@
 				});
 				this.recorderStatus="stop";
 			},
-			startRecorder(){
+				startRecorder(){
 				this.$emit('change',{status:"START",msg:"开始录音"});
 			},
+				recordStopped(data){
+					setTimeout(()=>{
+						this.finishTask();
+					},80);
+					if(data && data.saveAudioFile === false){
+						this.$emit('change',{status:"STOP",msg:{}});
+					}
+				},
+				finishTask(){
+					if(!ws || !this.task_id){
+						this.closeSocket();
+						return;
+					}
+
+					var param={
+						header: {
+							action: "finish-task",
+							task_id: this.task_id,
+							streaming: "duplex"
+						},
+						payload:{
+							input:{}
+						}
+					};
+
+					ws.send({
+						data:JSON.stringify(param),
+						fail:(err)=>{
+							this.$emit('change',{status:"ERROR",msg:err});
+							this.closeSocket();
+						}
+					});
+				},
+				closeSocket(){
+					if(ws){
+						var socket=ws;
+						ws=null;
+						try{
+							socket.close();
+						}catch(e){
+
+						}
+					}
+				},
 			webSocekt(){
 				ws = uni.connectSocket({
 					url: 'wss://dashscope.aliyuncs.com/api-ws/v1/inference',
@@ -75,12 +121,25 @@
 					var msg=JSON.parse(data.data)
 					if(msg.header.event=="task-started"){
 						this.task_id=msg.header.task_id;
-						this.recorderStatus="start";
+						this.recorderStatus={
+							status:"start",
+							time: Date.now()
+						};
 					}
 
 					if(msg.header.event=="result-generated"){
 						this.$emit('result',msg.payload.output);
 					}
+
+					if(msg.header.event=="task-finished"){
+						this.$emit('change',{status:"FINISH",msg:msg.payload || {}});
+						this.closeSocket();
+					}
+
+					if(msg.header.event=="task-failed"){
+						this.$emit('change',{status:"ERROR",msg:msg});
+						this.closeSocket();
+					}
 				});
 				
 				ws.onError((err)=>{
@@ -88,6 +147,7 @@
 				});
 				
 				ws.onClose(()=>{
+					ws=null;
 					this.$emit('change',{status:"CLOSE",msg:"连接已关闭"});
 				})
 			},
@@ -103,6 +163,7 @@
 				return result;
 			},
 			frameRecorded({isLastFrame,frameBuffer}){
+				if(!ws || !frameBuffer) return;
 				ws.send({data:uni.base64ToArrayBuffer(frameBuffer)});
 			},
 			recordedChunks(base64){
@@ -210,13 +271,15 @@
 
 		},
 		methods: {
-			async startRecord(val){
+				async startRecord(val){
 
-				if(val==null) return;
+					if(val==null) return;
+					const status = typeof val === 'object' ? val.status : val;
+					const saveAudioFile = typeof val === 'object' ? val.saveAudioFile !== false : true;
 
-				if(val == "start"){
+					if(status == "start"){
-					try{
+						try{
-						recordedChunks = [];
+							recordedChunks = [];
 						// 配置参数
 						var a, i = 16000,
 							s = 1280;
@@ -281,9 +344,9 @@
 						this.$ownerInstance.callMethod('toShowToast');
 						
 					}
-				}else if(val == "stop"){
+					}else if(status == "stop"){
-					this.onStop();
+						this.onStop(saveAudioFile);
-				}
+					}
 								
 			},
 			onFrameRecorded({isLastFrame,frameBuffer}){
@@ -298,7 +361,7 @@
 				}
 				return window.btoa(binary);
 			},
-			async onRecordedChunks(chunks){
+				async onRecordedChunks(chunks){
 				
 				var mergedBuffer=this.mergeAudioBuffers(chunks);
 				// 2. 将合并的二进制流转换为WAV格式（需要添加WAV文件头）
@@ -307,13 +370,11 @@
 				const base64 = await this.blobToBase64(wavBlob);
 				this.$ownerInstance.callMethod('recordedChunks',base64)
 			},
-			onStop(){
+				onStop(saveAudioFile = true){
 
-				this.onFrameRecorded({isLastFrame:true,frameBuffer:''});
+					const chunks = recordedChunks;
-				
+					recordedChunks = [];
-				this.onRecordedChunks(recordedChunks);
+					if (mediaStream) {
-				recordedChunks = [];
-				if (mediaStream) {
 					// 停止所有媒体轨道
 					mediaStream.getTracks().forEach(track => track.stop());
 					mediaStream = null;
@@ -327,11 +388,15 @@
 				  
 				  if (audioContext) {
 					// 关闭音频上下文
-					audioContext.close().then(() => {
+						audioContext.close().then(() => {
-					  audioContext = null;
+						  audioContext = null;
-					});
+						});
-				}
+					}
-			},
+					this.$ownerInstance.callMethod('recordStopped',{saveAudioFile});
+					if(saveAudioFile){
+						this.onRecordedChunks(chunks);
+					}
+				},
 			//合并所有buffer
 			mergeAudioBuffers(buffers) {
 			   let totalLength = buffers.reduce((acc, buf) => acc + buf.byteLength, 0);