refactor: 删除实时分析/按需识别/纯聊天三种模式切换

三种模式仅为前端 UI 区分，后端无感知，实际使用中价值不大：
- 纯聊天模式名不副实（VAD 仍附带摄像头帧）
- 实时分析增加 API 成本且体验不佳
- 按需识别已是默认且最自然的交互方式

删除后保留按需识别行为：语音带图 + 手动识别按钮，UI 更简洁。

frontend/src/App.css

@@ -644,12 +644,6 @@ body {
   transition: all var(--transition-fast);
 }
 
-.ai-vision-indicator--active {
-  color: var(--color-success);
-  background: rgba(52, 211, 153, 0.15);
-  animation: pulse 2s cubic-bezier(0.25, 0.46, 0.45, 0.94) infinite;
-}
-
 /* ---- 视频控制栏 ---- */
 
 .video-controls {
@@ -750,16 +744,6 @@ body {
   gap: 10px;
 }
 
-.chat-panel-header__mode {
-  font-size: 0.68rem;
-  font-weight: 500;
-  color: var(--color-success);
-  background: rgba(52, 211, 153, 0.08);
-  padding: 3px 10px;
-  border-radius: 12px;
-  letter-spacing: 0.02em;
-}
-
 /* ---- Scenario Quick Selector (chat header) ---- */
 .scenario-selector {
   appearance: none;
@@ -1473,38 +1457,6 @@ body {
   outline-offset: 1px;
 }
 
-/* Mode switcher */
-.video-controls__mode {
-  display: flex;
-  gap: 2px;
-  background: var(--color-surface-2);
-  border-radius: var(--radius-sm);
-  padding: 2px;
-  justify-content: center;
-}
-
-.mode-btn {
-  background: none;
-  border: none;
-  color: var(--color-text-muted);
-  font-size: 0.7rem;
-  font-weight: 500;
-  padding: 5px 12px;
-  border-radius: 6px;
-  cursor: pointer;
-  transition: all var(--transition-fast);
-  white-space: nowrap;
-}
-
-.mode-btn:hover {
-  color: var(--color-text);
-}
-
-.mode-btn--active {
-  background: var(--color-primary);
-  color: white;
-}
-
 /* ---- Text-only mode (video ended, chat preserved) ---- */
 
 .video-controls__text-only {

frontend/src/App.tsx

@@ -24,9 +24,6 @@ import type { Theme } from "./types";
 import type { UserScenario } from "./lib/api/scenarios";
 import "./App.css";
 
-/** AI 视觉模式 */
-type VisionMode = "realtime" | "ondemand" | "chat";
-
 /** 内部组件，确保在 I18nContext.Provider 内部使用 hooks */
 function AppContent() {
   const { isAuthenticated, isLoading, user, logout, accessToken } = useAuth();
@@ -36,7 +33,6 @@ function AppContent() {
   const [theme, setTheme] = useState<Theme>(loadTheme);
   const [elapsed, setElapsed] = useState(0);
   const [sidebarOpen, setSidebarOpen] = useState(false);
-  const [visionMode, setVisionMode] = useState<VisionMode>("ondemand");
   const timerRef = useRef<ReturnType<typeof setInterval> | null>(null);
 
   // ---- 自建情景管理 ----
@@ -91,8 +87,6 @@ function AppContent() {
     config,
     updateConfig,
     stats,
-    mode,
-    isObserving,
     startSession,
     stopSession,
     stopVideo,
@@ -365,13 +359,13 @@ function AppContent() {
               <div className="detail-badge">HD</div>
             )}
             {/* AI 视觉状态指示 */}
-            {isConnected && visionMode !== "chat" && (
-              <div className={`ai-vision-indicator ${isObserving ? "ai-vision-indicator--active" : ""}`}>
+            {isConnected && (
+              <div className="ai-vision-indicator">
                 <svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" strokeWidth="2" strokeLinecap="round" strokeLinejoin="round">
                   <path d="M1 12s4-8 11-8 11 8 11 8-4 8-11 8-11-8-11-8z" />
                   <circle cx="12" cy="12" r="3" />
                 </svg>
-                <span>{isObserving ? tr("video.observing") : "AI"}</span>
+                <span>AI</span>
               </div>
             )}
             {isSpeaking && (
@@ -427,27 +421,6 @@ function AppContent() {
                     </select>
                   </div>
                 </div>
-                {/* 模式切换器 */}
-                <div className="video-controls__mode">
-                  <button
-                    className={`mode-btn ${visionMode === "realtime" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("realtime")}
-                  >
-                    {tr("controls.mode.realtime")}
-                  </button>
-                  <button
-                    className={`mode-btn ${visionMode === "ondemand" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("ondemand")}
-                  >
-                    {tr("controls.mode.ondemand")}
-                  </button>
-                  <button
-                    className={`mode-btn ${visionMode === "chat" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("chat")}
-                  >
-                    {tr("controls.mode.chat")}
-                  </button>
-                </div>
               </>
             ) : isCameraOn ? (
               <>
@@ -467,8 +440,7 @@ function AppContent() {
                   >
                     🎤
                   </button>
-                  {/* 识别画面按钮（按需模式下显示） */}
-                  {visionMode === "ondemand" && (
+                  {/* 识别画面按钮 */}
                   <button
                     className="btn--recognize"
                     onClick={handleRecognize}
@@ -476,7 +448,6 @@ function AppContent() {
                   >
                     🔍 {tr("controls.recognize")}
                   </button>
-                  )}
                   {isProcessing && (
                     <button className="btn btn--warning" onClick={interrupt}>
                       {tr("controls.interrupt")}
@@ -486,27 +457,6 @@ function AppContent() {
                     {tr("controls.stopVideo")}
                   </button>
                 </div>
-                {/* 通话态模式切换 */}
-                <div className="video-controls__mode">
-                  <button
-                    className={`mode-btn ${visionMode === "realtime" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("realtime")}
-                  >
-                    {tr("controls.mode.realtime")}
-                  </button>
-                  <button
-                    className={`mode-btn ${visionMode === "ondemand" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("ondemand")}
-                  >
-                    {tr("controls.mode.ondemand")}
-                  </button>
-                  <button
-                    className={`mode-btn ${visionMode === "chat" ? "mode-btn--active" : ""}`}
-                    onClick={() => setVisionMode("chat")}
-                  >
-                    {tr("controls.mode.chat")}
-                  </button>
-                </div>
               </>
             ) : (
               <>
@@ -547,9 +497,6 @@ function AppContent() {
                   </option>
                 ))}
               </select>
-              {isConnected && mode === "observation" && (
-                <span className="chat-panel-header__mode">{tr("chat.mode.observation")}</span>
-              )}
               {isConnected && stats.queryCount > 0 && (
                 <span className="chat-panel-header__stats">
                   {stats.queryCount} {tr("statusbar.recognitions")}

frontend/src/hooks/useObservationMode.ts

@@ -1,88 +0,0 @@
-// ============================================================
-// useObservationMode — 观察模式 Hook
-// 职责：定时采帧 → 关键帧检测 → 画面变化时触发回调
-// 来源：docs/05-用户故事.md US-05（持续场景监控）
-// ============================================================
-
-import { useCallback, useEffect, useRef, useState } from "react";
-import { sampleFrame, compareFrames } from "../components/EdgeProcessor";
-
-/** 画面变化显著阈值 */
-const CHANGE_THRESHOLD = 0.85;
-/** 采样间隔（ms） */
-const SAMPLE_INTERVAL = 5000;
-
-export interface ObservationOptions {
-  /** 画面变化回调，携带当前帧的 DataURL */
-  onChange?: (frameDataUrl: string) => void;
-}
-
-export function useObservationMode(options?: ObservationOptions) {
-  const [isObserving, setIsObserving] = useState(false);
-  const intervalRef = useRef<ReturnType<typeof setInterval> | null>(null);
-  const prevFrameRef = useRef<Uint8ClampedArray | null>(null);
-  const optionsRef = useRef(options);
-
-  useEffect(() => {
-    optionsRef.current = options;
-  }, [options]);
-
-  /**
-   * 启动观察模式
-   * @param video 摄像头 video 元素
-   * @param captureFrame 从 video 捕获 DataURL 的函数
-   */
-  const startObserving = useCallback(
-    (video: HTMLVideoElement | null, captureFrame: () => string | null) => {
-      if (!video) return;
-
-      // 立即采一帧作为基准
-      prevFrameRef.current = sampleFrame(video);
-
-      intervalRef.current = setInterval(() => {
-        const current = sampleFrame(video);
-        if (!current) return;
-
-        if (prevFrameRef.current) {
-          const { similarity } = compareFrames(prevFrameRef.current, current);
-
-          if (similarity < CHANGE_THRESHOLD) {
-            console.log(
-              `[Observation] 画面变化 (similarity=${similarity.toFixed(2)})`,
-            );
-            const frameDataUrl = captureFrame();
-            if (frameDataUrl) {
-              optionsRef.current?.onChange?.(frameDataUrl);
-            }
-          }
-        }
-
-        prevFrameRef.current = current;
-      }, SAMPLE_INTERVAL);
-
-      setIsObserving(true);
-    },
-    [],
-  );
-
-  /** 停止观察模式 */
-  const stopObserving = useCallback(() => {
-    if (intervalRef.current) {
-      clearInterval(intervalRef.current);
-      intervalRef.current = null;
-    }
-    prevFrameRef.current = null;
-    setIsObserving(false);
-  }, []);
-
-  // 组件卸载时清理
-  useEffect(() => {
-    return () => {
-      if (intervalRef.current) {
-        clearInterval(intervalRef.current);
-      }
-    };
-  }, []);
-
-  return { isObserving, startObserving, stopObserving };
-}

frontend/src/hooks/useVisionSession.ts

@@ -17,11 +17,8 @@ import { useCamera } from "../components/CameraManager";
 import { useMicrophone } from "../components/MicManager";
 import { useVAD, sampleFrame } from "../components/EdgeProcessor";
 import { useWebSocketManager } from "../components/WebSocketManager";
-import { useObservationMode } from "./useObservationMode";
 import type { ChatMessage, SessionConfig, ServerMessage, LLMDoneMessage } from "../types";
 
-export type SessionMode = "dialogue" | "observation";
-
 export interface SessionStats {
   queryCount: number;
   totalTokens: number;
@@ -35,7 +32,6 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
   const [isAudioPlaying, setIsAudioPlaying] = useState(false);
   const [config, setConfig] = useState<SessionConfig>(loadConfig);
   const [stats, setStats] = useState<SessionStats>({ queryCount: 0, totalTokens: 0 });
-  const [mode, setMode] = useState<SessionMode>("dialogue");
   const [isCameraOn, setIsCameraOn] = useState(false);
   const [isMicOn, setIsMicOn] = useState(false);
 
@@ -78,49 +74,6 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
     conversationIdRef.current = conversationId;
   }, [conversationId]);
 
-  // 观察模式：画面变化时自动发送 query
-  const { isObserving, startObserving, stopObserving } = useObservationMode({
-    onChange: useCallback(
-      (frameDataUrl: string) => {
-        if (isProcessingRef.current) return;
-
-        const requestId = uuidv4();
-        send({
-          type: "query",
-          request_id: requestId,
-          image: dataUrlToBase64(frameDataUrl),
-          audio: "", // 观察模式无音频
-        });
-
-        setMessages((prev) => [
-          ...prev,
-          {
-            id: uuidv4(),
-            role: "user",
-            content: t("session.changeDetected"),
-            timestamp: Date.now(),
-          },
-        ]);
-        setStats((prev) => ({ ...prev, queryCount: prev.queryCount + 1 }));
-        setIsProcessing(true);
-      },
-      [send],
-    ),
-  });
-
-  /** 切换对话/观察模式 */
-  const toggleMode = useCallback(() => {
-    setMode((prev) => {
-      const next = prev === "dialogue" ? "observation" : "dialogue";
-      if (next === "observation") {
-        startObserving(videoRef.current, captureFrame);
-      } else {
-        stopObserving();
-      }
-      return next;
-    });
-  }, [videoRef, captureFrame, startObserving, stopObserving]);
-
   // WebSocket 连接成功后发送 config + flush 待发消息
   useEffect(() => {
     if (status === "connected") {
@@ -343,8 +296,6 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
 
   /** 结束会话 */
   const stopSession = useCallback(async () => {
-    stopObserving();
-    setMode("dialogue");
     await stopVAD();
     stopMic();
     stopCamera();
@@ -359,12 +310,10 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
     prevFrameRef.current = null;
     setIsCameraOn(false);
     setIsMicOn(false);
-  }, [stopObserving, stopVAD, stopMic, stopCamera, disconnect]);
+  }, [stopVAD, stopMic, stopCamera, disconnect]);
 
   /** 结束视频，保留聊天和连接 */
   const stopVideo = useCallback(async () => {
-    stopObserving();
-    setMode("dialogue");
     await stopVAD();
     stopMic();
     stopCamera();
@@ -375,7 +324,7 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
     setIsCameraOn(false);
     setIsMicOn(false);
     // 不断开 WebSocket，不清空消息、统计
-  }, [stopObserving, stopVAD, stopMic, stopCamera]);
+  }, [stopVAD, stopMic, stopCamera]);
 
   /** 摄像头开关 */
   const toggleCamera = useCallback(async () => {
@@ -480,9 +429,6 @@ export function useVisionSession(accessToken?: string | null, conversationId?: s
     config,
     updateConfig,
     stats,
-    mode,
-    isObserving,
-    toggleMode,
     startSession,
     stopSession,
     stopVideo,

frontend/src/lib/i18n/en-US.ts

@@ -48,7 +48,6 @@ export const enUS: TranslationMap = {
   "scenario.interpreter.hint": "AI will translate your speech in real-time (Chinese-English), without explanations",
 
   // Video indicators
-  "video.observing": "👁️ Observing",
   "video.listening": "🎤 Listening...",
   "video.playing": "🔊 Playing...",
   "video.initVad": "Initializing voice detection...",
@@ -67,8 +66,6 @@ export const enUS: TranslationMap = {
   "controls.cameraOn": "Turn on camera",
   "controls.micOff": "Turn off microphone",
   "controls.micOn": "Turn on microphone",
-  "controls.observing": "👁️ Observing",
-  "controls.observation": "👁️ Observe",
   "controls.interrupt": "⏹ Interrupt",
   "controls.stop": "End Session",
   "controls.stopVideo": "End Video",
@@ -77,7 +74,6 @@ export const enUS: TranslationMap = {
 
   // Chat panel
   "chat.title": "Chat",
-  "chat.mode.observation": "Observing",
   "chat.reconnecting": "Connection lost, reconnecting...",
   "chat.connecting": "Connecting to server...",
   "chat.vadInit": "Initializing voice detection...",
@@ -91,7 +87,6 @@ export const enUS: TranslationMap = {
   "chat.scenarioSwitched": "Switched to {name} mode",
 
   // Session messages
-  "session.changeDetected": "👁️ Scene change detected",
   "session.recognizing": "(Recognizing speech...)",
   "session.sttFailed": "(Speech recognition failed, please try again)",
   "session.noSpeech": "(No speech detected)",
@@ -141,9 +136,6 @@ export const enUS: TranslationMap = {
   "controls.device.camera": "Camera",
   "controls.device.mic": "Microphone",
   "controls.device.default": "Default",
-  "controls.mode.realtime": "Realtime",
-  "controls.mode.ondemand": "On-demand",
-  "controls.mode.chat": "Chat only",
 
   // Status bar
   "statusbar.ready": "Ready · Select devices to start",

frontend/src/lib/i18n/ja-JP.ts

@@ -48,7 +48,6 @@ export const jaJP: TranslationMap = {
   "scenario.interpreter.hint": "AIがあなたの発言をリアルタイムで翻訳します（中日相互翻訳、説明なし）",
 
   // Video indicators
-  "video.observing": "👁️ 観察中",
   "video.listening": "🎤 聞き取り中...",
   "video.playing": "🔊 再生中...",
   "video.initVad": "音声検出を初期化中...",
@@ -67,8 +66,6 @@ export const jaJP: TranslationMap = {
   "controls.cameraOn": "カメラをオン",
   "controls.micOff": "マイクをオフ",
   "controls.micOn": "マイクをオン",
-  "controls.observing": "👁️ 観察中",
-  "controls.observation": "👁️ 観察モード",
   "controls.interrupt": "⏹ 中断",
   "controls.stop": "対話を終了",
   "controls.stopVideo": "ビデオ終了",
@@ -77,7 +74,6 @@ export const jaJP: TranslationMap = {
 
   // Chat panel
   "chat.title": "チャット",
-  "chat.mode.observation": "観察モード",
   "chat.reconnecting": "接続が切断されました。再接続中...",
   "chat.connecting": "サーバーに接続中...",
   "chat.vadInit": "音声検出を初期化中...",
@@ -91,7 +87,6 @@ export const jaJP: TranslationMap = {
   "chat.scenarioSwitched": "{name} モードに切り替えました",
 
   // Session messages
-  "session.changeDetected": "👁️ シーン変化を検出",
   "session.recognizing": "（音声認識中...）",
   "session.sttFailed": "（音声認識に失敗しました。もう一度お試しください）",
   "session.noSpeech": "（音声が検出されませんでした）",
@@ -141,9 +136,6 @@ export const jaJP: TranslationMap = {
   "controls.device.camera": "カメラ",
   "controls.device.mic": "マイク",
   "controls.device.default": "デフォルト",
-  "controls.mode.realtime": "リアルタイム",
-  "controls.mode.ondemand": "オンデマンド",
-  "controls.mode.chat": "チャットのみ",
 
   // Status bar
   "statusbar.ready": "準備完了 · デバイスを選択して開始",

frontend/src/lib/i18n/zh-CN.ts

@@ -48,7 +48,6 @@ export const zhCN: TranslationMap = {
   "scenario.interpreter.hint": "AI 会实时翻译你的话（中英互译），无解释评论",
 
   // Video indicators
-  "video.observing": "👁️ 观察中",
   "video.listening": "🎤 正在聆听...",
   "video.playing": "🔊 正在播放...",
   "video.initVad": "正在初始化语音检测...",
@@ -67,8 +66,6 @@ export const zhCN: TranslationMap = {
   "controls.cameraOn": "开启摄像头",
   "controls.micOff": "关闭麦克风",
   "controls.micOn": "开启麦克风",
-  "controls.observing": "👁️ 观察中",
-  "controls.observation": "👁️ 观察模式",
   "controls.interrupt": "⏹ 打断",
   "controls.stop": "结束对话",
   "controls.stopVideo": "结束视频",
@@ -77,7 +74,6 @@ export const zhCN: TranslationMap = {
 
   // Chat panel
   "chat.title": "对话",
-  "chat.mode.observation": "观察模式",
   "chat.reconnecting": "连接已断开，正在重连...",
   "chat.connecting": "正在连接服务...",
   "chat.vadInit": "正在初始化语音检测...",
@@ -91,7 +87,6 @@ export const zhCN: TranslationMap = {
   "chat.scenarioSwitched": "已切换到 {name} 模式",
 
   // Session messages
-  "session.changeDetected": "👁️ 画面变化检测",
   "session.recognizing": "（语音识别中...）",
   "session.sttFailed": "（语音识别失败，请重试）",
   "session.noSpeech": "（未识别到语音）",
@@ -141,9 +136,6 @@ export const zhCN: TranslationMap = {
   "controls.device.camera": "摄像头",
   "controls.device.mic": "麦克风",
   "controls.device.default": "默认",
-  "controls.mode.realtime": "实时分析",
-  "controls.mode.ondemand": "按需识别",
-  "controls.mode.chat": "纯聊天",
 
   // Status bar
   "statusbar.ready": "就绪 · 选择设备后开始通话",