feat：语音对话优化

frontend/src/hooks/useVisionSession.ts

@@ -15,7 +15,7 @@ import { loadConfig, saveConfig } from "../lib/storage";
 import { useI18n } from "../lib/i18n";
 import { useCamera } from "../components/CameraManager";
 import { useMicrophone } from "../components/MicManager";
-import { useVAD, sampleFrame, compareFrames } from "../components/EdgeProcessor";
+import { useVAD, sampleFrame } from "../components/EdgeProcessor";
 import { useWebSocketManager } from "../components/WebSocketManager";
 import { useObservationMode } from "./useObservationMode";
 import type { ChatMessage, SessionConfig, ServerMessage, LLMDoneMessage } from "../types";
@@ -194,24 +194,13 @@ export function useVisionSession(accessToken?: string | null) {
         ttsPlayerRef.current?.stop();
         setIsAudioPlaying(false);
 
+        // 尝试捕获图像帧（允许为 null，纯语音场景无需画面）
         const frame = captureFrame();
-        if (!frame) {
-          console.warn("[Session] 无法捕获图像帧");
-          return;
-        }
 
-        // 关键帧检测：与上一帧对比，相似度过高则跳过
+        // 更新帧参考（供观察模式使用，但不以此阻断语音请求）
         const video = videoRef.current;
         if (video) {
           const currentSample = sampleFrame(video);
-          if (currentSample && prevFrameRef.current) {
-            const { similarity } = compareFrames(prevFrameRef.current, currentSample);
-            if (similarity > 0.9) {
-              console.log(`[Session] 画面无变化 (similarity=${similarity.toFixed(2)})，跳过`);
-              prevFrameRef.current = currentSample;
-              return;
-            }
-          }
           if (currentSample) {
             prevFrameRef.current = currentSample;
           }
@@ -221,7 +210,7 @@ export function useVisionSession(accessToken?: string | null) {
         send({
           type: "query",
           request_id: requestId,
-          image: dataUrlToBase64(frame),
+          image: frame ? dataUrlToBase64(frame) : "",
           audio: encodeAudioToBase64(audio),
         });
 
@@ -304,6 +293,20 @@ export function useVisionSession(accessToken?: string | null) {
         case "error":
           console.error("[Session] 服务端错误:", msg.code, msg.message);
           showToast(getErrorMessage(msg.code, t), "error");
+          // STT 失败时：将"语音识别中..."占位消息替换为失败提示
+          if (msg.code === "STT_ERROR") {
+            setMessages((prev) => {
+              const updated = [...prev];
+              const lastUserIdx = updated.findLastIndex((m) => m.role === "user");
+              if (lastUserIdx >= 0 && updated[lastUserIdx].content === t("session.recognizing")) {
+                updated[lastUserIdx] = {
+                  ...updated[lastUserIdx],
+                  content: t("session.sttFailed"),
+                };
+              }
+              return updated;
+            });
+          }
           setIsProcessing(false);
           break;
       }