Merge branch 'master' of https://gitea.xiaowang.eu.org/wx44wx/AI-VideoAssistant

web/pages/Assistants.tsx

@@ -2207,6 +2207,8 @@ type DebugTextPromptDialogState = {
   open: boolean;
   message: string;
   pendingResult?: DebugPromptPendingResult;
+  promptType?: 'text' | 'voice';
+  voiceText?: string;
 };
 
 type DebugChoicePromptDialogState = {
@@ -2341,7 +2343,11 @@ export const DebugDrawer: React.FC<{
   const [inputText, setInputText] = useState('');
   const [isLoading, setIsLoading] = useState(false);
   const [callStatus, setCallStatus] = useState<'idle' | 'calling' | 'active'>('idle');
-  const [textPromptDialog, setTextPromptDialog] = useState<DebugTextPromptDialogState>({ open: false, message: '' });
+  const [textPromptDialog, setTextPromptDialog] = useState<DebugTextPromptDialogState>({
+    open: false,
+    message: '',
+    promptType: 'text',
+  });
   const [choicePromptDialog, setChoicePromptDialog] = useState<DebugChoicePromptDialogState>({ open: false, question: '', options: [] });
   const textPromptDialogRef = useRef(textPromptDialog);
   const choicePromptDialogRef = useRef(choicePromptDialog);
@@ -2538,7 +2544,7 @@ export const DebugDrawer: React.FC<{
         closeWs();
         stopPromptVoicePlayback();
         promptDialogQueueRef.current = [];
-        setTextPromptDialog({ open: false, message: '' });
+        setTextPromptDialog({ open: false, message: '', promptType: 'text' });
         setChoicePromptDialog({ open: false, question: '', options: [] });
         if (audioCtxRef.current) {
           void audioCtxRef.current.close();
@@ -2861,11 +2867,17 @@ export const DebugDrawer: React.FC<{
 
   const activatePromptDialog = (item: DebugPromptQueueItem) => {
     if (item.kind === 'text') {
+      const nextVoiceText = String(item.payload.voiceText || '').trim();
       setTextPromptDialog({
         open: true,
         message: item.payload.message,
         pendingResult: item.payload.pendingResult,
+        promptType: item.payload.promptType || 'text',
+        voiceText: nextVoiceText || undefined,
       });
+      if (nextVoiceText) {
+        void playPromptVoice(nextVoiceText);
+      }
       return;
     }
     const nextVoiceText = String(item.payload.voiceText || '').trim();
@@ -2902,16 +2914,19 @@ export const DebugDrawer: React.FC<{
     if (!snapshot.open && !opts?.force) return;
     const pending = snapshot?.pendingResult;
     const message = snapshot?.message || '';
-    setTextPromptDialog({ open: false, message: '' });
+    const promptType = snapshot?.promptType === 'voice' ? 'voice' : 'text';
+    stopPromptVoicePlayback();
+    setTextPromptDialog({ open: false, message: '', promptType: 'text' });
     if (pending?.waitForResponse) {
       emitClientToolResult(
         {
           tool_call_id: pending.toolCallId,
           name: pending.toolName,
           output: {
-            message: 'text_prompt_closed',
+            message: promptType === 'voice' ? 'voice_prompt_closed' : 'text_prompt_closed',
             action,
             msg: message,
+            prompt_type: promptType,
           },
           status: { code: 200, message: 'ok' },
         },
@@ -3108,7 +3123,7 @@ export const DebugDrawer: React.FC<{
     setCallStatus('idle');
     clearResponseTracking();
     setMessages([]);
-    setTextPromptDialog({ open: false, message: '' });
+    setTextPromptDialog({ open: false, message: '', promptType: 'text' });
     setChoicePromptDialog({ open: false, question: '', options: [] });
     lastUserFinalRef.current = '';
     setIsLoading(false);
@@ -3467,7 +3482,7 @@ export const DebugDrawer: React.FC<{
     micFrameBufferRef.current = new Uint8Array(0);
     stopPromptVoicePlayback();
     promptDialogQueueRef.current = [];
-    setTextPromptDialog({ open: false, message: '' });
+    setTextPromptDialog({ open: false, message: '', promptType: 'text' });
     setChoicePromptDialog({ open: false, question: '', options: [] });
     setTextSessionStarted(false);
     stopPlaybackImmediately();
@@ -3729,13 +3744,26 @@ export const DebugDrawer: React.FC<{
                   resultPayload.output = { message: "Missing required argument 'msg'" };
                   resultPayload.status = { code: 422, message: 'invalid_arguments' };
                 } else {
-                  void playPromptVoice(msg);
+                  enqueuePromptDialog({
-                  if (waitForResponse) {
+                    kind: 'text',
-                    // Voice prompt playback is fire-and-forget; keep previous wait behavior stable.
+                    payload: {
-                    // Client ack is returned immediately after dispatch.
+                      message: msg,
+                      promptType: 'voice',
+                      voiceText: msg,
+                      pendingResult: {
+                        toolCallId: toolCallId,
+                        toolName,
+                        toolDisplayName,
+                        waitForResponse,
+                      },
+                    },
+                  });
+                  if (!waitForResponse) {
+                    resultPayload.output = { message: 'voice_prompt_shown', msg };
+                    resultPayload.status = { code: 200, message: 'ok' };
+                  } else {
+                    return;
                   }
-                  resultPayload.output = { message: 'voice_prompt_sent', msg };
-                  resultPayload.status = { code: 200, message: 'ok' };
                 }
               } else if (toolName === 'text_msg_prompt') {
                 const msg = String(parsedArgs?.msg || '').trim();
@@ -4524,7 +4552,9 @@ export const DebugDrawer: React.FC<{
                   <X className="h-4 w-4" />
                 </button>
                 <div className="mb-3 pr-6">
-                  <div className="text-[10px] font-black tracking-[0.14em] uppercase text-amber-300">文本消息提示</div>
+                  <div className="text-[10px] font-black tracking-[0.14em] uppercase text-amber-300">
+                    {textPromptDialog.promptType === 'voice' ? '语音消息提示' : '文本消息提示'}
+                  </div>
                   <p className="mt-2 text-sm leading-6 text-foreground whitespace-pre-wrap break-words">{textPromptDialog.message}</p>
                 </div>
                 <div className="flex justify-end">