fix: 优化相关异常处理

src/renderer/screens/ShowPhoto.vue

@@ -50,10 +50,10 @@ onMounted(() => {
 })
 
 const errorSnackbar = ref<boolean>(false)
-const errorMsg = ref<string>('');
+const errorMsg = ref<string>('')
 const showError = (msg: string) => {
-  errorSnackbar.value = true;
-  errorMsg.value = msg;
+  errorSnackbar.value = true
+  errorMsg.value = msg
 }
 
 function loadImg(): Promise<HTMLImageElement> {
@@ -77,17 +77,7 @@ async function init() {
   canvasEle.height = img.naturalHeight
   const item = photo.list.find((i) => i.url === url)
   photoRole = new PhotoRole(settings.liveHost, `${item?.liveUrl}`, canvasEle)
-
-  photoRole.on('asyncAnswer', async (ans) => {
-    if (ans.playState === 'playing') {
-      microphoneState.value = 'reply'
-      return
-    }
-
-    if (microphoneState.value === 'reply' && ans.playState === 'pause' && (await checkSteps())) {
-      microphoneState.value = 'input'
-    }
-  })
+  photoRole.on('asyncAnswer', onAsyncAnswer)
 
   // initPlayer(videoEle);
 
@@ -112,6 +102,17 @@ async function init() {
   requestAnimationFrame(updateFrame)
 }
 
+async function onAsyncAnswer(ans: PhotoAnswer) {
+  if (ans.playState === 'playing') {
+    microphoneState.value = 'reply'
+    return
+  }
+
+  if (microphoneState.value === 'reply' && ans.playState === 'pause' && (await checkSteps())) {
+    microphoneState.value = 'input'
+  }
+}
+
 function draw(
   ctx: CanvasRenderingContext2D,
   img: HTMLImageElement,
@@ -273,7 +274,7 @@ async function startVoskWsAudioInput() {
   if (microphoneState.value === 'loading') return
 
   if (microphoneState.value === 'input') {
-    endAudioInput()
+    await endAudioInput()
     return
   }
 
@@ -378,6 +379,9 @@ async function endAudioInput() {
   }
   inputContext.asrPartial = ''
   inputContext.answerArray.length = 0
+  inputContext.steps.length = 0
+  // @ts-ignore
+  photoRole?.off('asyncAnswer', onAsyncAnswer)
   await photoRole?.destroy()
 }
 
@@ -417,73 +421,80 @@ async function onQ(question: string) {
 
   microphoneState.value = 'loading'
 
-  const ws = await initLLMSocket()
-  const { pose, stepResolve, stepReject } = createStep()
-  const messageTimeout = setTimeout(async () => {
-    showError('llm：timeout!')
-    await endAudioInput()
-    microphoneState.value = 'waitInput'
-  }, 5000);
-
-  let sliceAnswer = ''
-  let answer = ''
-  let isTime = true
-  let sliceAnswerLength = 10
-
-  inputContext.ws = ws
-  inputContext.answerArray.length = 0
-  inputContext.steps.length = 0
-  inputContext.steps.push(pose)
-  photoRole!.answerArgs = new PhotoAnswer()
-
-  ws.onmessage = (message) => {
-    try {
-      const { text, event } = JSON.parse(message.data) as {
-        event: string
-        message_num: number
-        text: string
-      }
-
-      if (event === 'stream_end') {
-        inputContext.answerArray.push({ text: sliceAnswer, isLast: true })
-        sliceAnswer = ''
-        runTTSTask()
-        inputContext.ws?.close()
-        console.log('----------------> answer: ', answer)
-        stepResolve('chat')
-        return
-      }
+  try {
+    const ws = await initLLMSocket()
+    const { pose, stepResolve, stepReject } = createStep()
+    const messageTimeout = setTimeout(async () => {
+      showError('llm：timeout!')
+      await endAudioInput()
+      microphoneState.value = 'waitInput'
+    }, 10000)
+
+    let sliceAnswer = ''
+    let answer = ''
+    let isTime = true
+    let sliceAnswerLength = 10
+
+    inputContext.ws = ws
+    inputContext.answerArray.length = 0
+    inputContext.steps.length = 0
+    inputContext.steps.push(pose)
+    photoRole!.answerArgs = new PhotoAnswer()
+    photoRole!.on('asyncAnswer', onAsyncAnswer)
+
+    ws.onmessage = (message) => {
+      try {
+        const { text, event } = JSON.parse(message.data) as {
+          event: string
+          message_num: number
+          text: string
+        }
 
-      answer += text
-      photoRole!.answerArgs!.answer += answer
-      photoRole!.answerArgs!._typingAnswer.push(answer)
-      isTime && console.time('sliceAnswer')
-      isTime = false
-      clearTimeout(messageTimeout);
-
-      const textArr = text.split('')
-      for (let i = 0; i < textArr.length; i++) {
-        const t = textArr[i]
-        sliceAnswer += t
-        if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= sliceAnswerLength) {
-          console.timeEnd('sliceAnswer')
-          sliceAnswerLength = settings.llmToTTSSliceLength
+        if (event === 'stream_end') {
           inputContext.answerArray.push({ text: sliceAnswer, isLast: true })
-          runTTSTask()
           sliceAnswer = ''
-          isTime = true
+          runTTSTask()
+          inputContext.ws?.close()
+          console.log('----------------> answer: ', answer)
+          stepResolve('chat')
+          return
         }
+
+        answer += text
+        photoRole!.answerArgs!.answer += answer
+        photoRole!.answerArgs!._typingAnswer.push(answer)
+        isTime && console.time('sliceAnswer')
+        isTime = false
+        clearTimeout(messageTimeout)
+
+        const textArr = text.split('')
+        for (let i = 0; i < textArr.length; i++) {
+          const t = textArr[i]
+          sliceAnswer += t
+          if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= sliceAnswerLength) {
+            console.timeEnd('sliceAnswer')
+            sliceAnswerLength = settings.llmToTTSSliceLength
+            inputContext.answerArray.push({ text: sliceAnswer, isLast: true })
+            runTTSTask()
+            sliceAnswer = ''
+            isTime = true
+          }
+        }
+      } catch (error) {
+        showError('llm：' + error)
+        endAudioInput().then(() => {
+          microphoneState.value = 'waitInput'
+        })
+        stepReject(JSON.stringify(error))
       }
-    } catch (error) {
-      showError('llm：'+ error)
-      endAudioInput().then(() => {
-        microphoneState.value = 'waitInput'
-      })
-      stepReject(JSON.stringify(error))
     }
-  }
 
-  ws.send(JSON.stringify({ prompt: question, historys_list: [] }))
+    ws.send(JSON.stringify({ prompt: question, historys_list: [] }))
+  } catch (error) {
+    console.error(error)
+    microphoneState.value = 'input'
+    showError(`llm：${JSON.stringify(error)}`)
+  }
 }
 
 let isTTSRunning = false
@@ -525,7 +536,7 @@ async function runTTSTask() {
       })
     }
   } catch (error) {
-    showError('tts：'+ error)
+    showError('tts：' + error)
     endAudioInput().then(() => {
       microphoneState.value = 'waitInput'
     })
@@ -577,7 +588,7 @@ async function down() {
   try {
     await startVoskWsAudioInput()
   } catch (error) {
-    showError('start：'+ error)
+    showError('start：' + error)
     endAudioInput().then(() => {
       microphoneState.value = 'waitInput'
     })
@@ -631,22 +642,12 @@ async function down() {
     </v-btn>
   </div>
 
-  <v-snackbar
-      v-model="errorSnackbar"
-      multi-line
-      :timeout="3000"
-    >
-      {{ errorMsg }}
-
-      <template #actions>
-        <v-btn
-          color="red"
-          variant="text"
-          @click="errorSnackbar = false"
-        >
-          Close
-        </v-btn>
-      </template>
+  <v-snackbar v-model="errorSnackbar" multi-line :timeout="3000">
+    {{ errorMsg }}
+
+    <template #actions>
+      <v-btn color="red" variant="text" @click="errorSnackbar = false"> Close </v-btn>
+    </template>
   </v-snackbar>
 </template>
 <style scoped>

src/renderer/screens/ShowVideo.vue

@@ -35,8 +35,10 @@ const inputContext: {
   voskWs?: WebSocket
   asrPartial: string
   playingAudio?: HTMLAudioElement
+  answerArray: string[]
 } = {
-  asrPartial: ''
+  asrPartial: '',
+  answerArray: []
 }
 
 onMounted(() => {
@@ -48,10 +50,10 @@ router.beforeEach((g) => {
 })
 
 const errorSnackbar = ref<boolean>(false)
-const errorMsg = ref<string>('');
+const errorMsg = ref<string>('')
 const showError = (msg: string) => {
-  errorSnackbar.value = true;
-  errorMsg.value = msg;
+  errorSnackbar.value = true
+  errorMsg.value = msg
 }
 
 async function initVosk({
@@ -275,6 +277,7 @@ function endAudioInput() {
   videos[1].value?.pause()
   videos[0].value?.pause()
   isPlayRunning = false
+  inputContext.answerArray.length = 0
 }
 
 const canplay = () => {
@@ -344,81 +347,82 @@ async function onQ(question: string) {
   }
 
   // 视频链接匹配不上，直接走大模型
-  const ws = await initLLMSocket()
-  const answerArray: string[] = []
-  const messageTimeout = setTimeout(() => {
-    showError('llm：timeout!')
-    endAudioInput()
-    microphoneState.value = 'waitInput'
-  }, 5000);
-  let sliceAnswer = ''
-  let answer = ''
-  let isTime = true
-  let sliceAnswerLength = 10
-  inputContext.ws = ws
-
-  ws.onmessage = (message) => {
-    if (microphoneState.value === 'input') {
-      return
-    }
-
-    try {
-      const { text, event } = JSON.parse(message.data) as {
-        event: string
-        message_num: number
-        text: string
+  try {
+    const ws = await initLLMSocket()
+    const messageTimeout = setTimeout(() => {
+      showError('llm：timeout!')
+      endAudioInput()
+      microphoneState.value = 'waitInput'
+    }, 10000)
+    let sliceAnswer = ''
+    let answer = ''
+    let isTime = true
+    let sliceAnswerLength = 10
+    inputContext.ws = ws
+    ws.onmessage = (message) => {
+      if (microphoneState.value === 'input') {
+        return
       }
 
-      if (event === 'stream_end') {
-        answerArray.push(sliceAnswer)
-        runTTSTask(answerArray)
-        sliceAnswer = ''
+      try {
+        const { text, event } = JSON.parse(message.data) as {
+          event: string
+          message_num: number
+          text: string
+        }
 
-        answerArray.push(sliceAnswer)
-        sliceAnswer = ''
-        inputContext.ws?.close()
-        console.log('----------------> answer: ', answer)
-        return
-      }
+        if (event === 'stream_end') {
+          inputContext.answerArray.push(sliceAnswer)
+          runTTSTask()
+          sliceAnswer = ''
 
-      answer += text
-      isTime && console.time('sliceAnswer')
-      isTime = false
-      clearTimeout(messageTimeout);
-
-      const textArr = text.split('')
-      for (let i = 0; i < textArr.length; i++) {
-        const t = textArr[i]
-        sliceAnswer += t
-        if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= sliceAnswerLength) {
-          console.timeEnd('sliceAnswer')
-          sliceAnswerLength = settings.llmToTTSSliceLength
-          answerArray.push(sliceAnswer)
-          runTTSTask(answerArray)
+          inputContext.answerArray.push(sliceAnswer)
           sliceAnswer = ''
-          isTime = true
+          inputContext.ws?.close()
+          console.log('----------------> answer: ', answer)
+          return
         }
+
+        answer += text
+        isTime && console.time('sliceAnswer')
+        isTime = false
+        clearTimeout(messageTimeout)
+
+        const textArr = text.split('')
+        for (let i = 0; i < textArr.length; i++) {
+          const t = textArr[i]
+          sliceAnswer += t
+          if (/[。，？！；,.?!;]/.test(t) && sliceAnswer.length >= sliceAnswerLength) {
+            console.timeEnd('sliceAnswer')
+            sliceAnswerLength = settings.llmToTTSSliceLength
+            inputContext.answerArray.push(sliceAnswer)
+            runTTSTask()
+            sliceAnswer = ''
+            isTime = true
+          }
+        }
+      } catch (error) {
+        console.error(error)
+        showError(`message：${error}`)
+        microphoneState.value = 'waitInput'
       }
-    } catch (error) {
-      console.error(error);
-      showError(`message：${error}`)
-      microphoneState.value = 'waitInput'
     }
+    ws.send(JSON.stringify({ prompt: question, historys_list: [] }))
+  } catch (error) {
+    console.error(error)
+    microphoneState.value = 'input'
+    showError(`llm：${JSON.stringify(error)}`)
   }
-
-  ws.send(JSON.stringify({ prompt: question, historys_list: [] }))
 }
 
 let isTTSRunning = false
-async function runTTSTask(tasks: string[]) {
+async function runTTSTask() {
   if (isTTSRunning) return
   isTTSRunning = true
 
-  microphoneState.value = 'loading'
-
   try {
-    while (tasks.length) {
-      const task = tasks.shift()
+    while (inputContext.answerArray.length) {
+      const task = inputContext.answerArray.shift()
       if (!task) break
       if (task.trim().length < 1) continue
 
@@ -560,22 +564,12 @@ async function down() {
     </v-chip>
   </div>
 
-  <v-snackbar
-      v-model="errorSnackbar"
-      multi-line
-      :timeout="3000"
-    >
-      {{ errorMsg }}
-
-      <template #actions>
-        <v-btn
-          color="red"
-          variant="text"
-          @click="errorSnackbar = false"
-        >
-          Close
-        </v-btn>
-      </template>
+  <v-snackbar v-model="errorSnackbar" multi-line :timeout="3000">
+    {{ errorMsg }}
+
+    <template #actions>
+      <v-btn color="red" variant="text" @click="errorSnackbar = false"> Close </v-btn>
+    </template>
   </v-snackbar>
 </template>