添加联网搜索

test/src/main/java/ai/guiji/duix/test/camera/CameraFrameCapture.java

@@ -49,6 +49,12 @@ public class CameraFrameCapture {
     private Handler backgroundHandler;
     private final AtomicBoolean isCapturing = new AtomicBoolean(false);
     private final AtomicBoolean isRunning = new AtomicBoolean(false);
+    private final Runnable captureRunnable = new Runnable() {
+        @Override
+        public void run() {
+            captureStillFrame();
+        }
+    };
 
     public CameraFrameCapture(Context context, FrameCallback callback) {
         this(context, callback, null);
@@ -194,7 +200,7 @@ public class CameraFrameCapture {
                             captureSession = session;
                             isCapturing.set(true);
                             startPreview();
-                            scheduleNextCapture();
+                            scheduleNextCapture(0L);
                         }
 
                         @Override
@@ -222,8 +228,14 @@ public class CameraFrameCapture {
         }
     }
 
-    private void scheduleNextCapture() {
+    private void scheduleNextCapture(long delayMs) {
-        if (!isCapturing.get() || !isRunning.get() || captureSession == null || cameraDevice == null) return;
+        if (backgroundHandler == null) return;
+        backgroundHandler.removeCallbacks(captureRunnable);
+        backgroundHandler.postDelayed(captureRunnable, Math.max(delayMs, 0L));
+    }
+
+    private void captureStillFrame() {
+        if (!isCapturing.get() || !isRunning.get() || captureSession == null || cameraDevice == null || imageReader == null) return;
         try {
             CaptureRequest.Builder builder = cameraDevice.createCaptureRequest(CameraDevice.TEMPLATE_PREVIEW);
             builder.addTarget(imageReader.getSurface());
@@ -235,7 +247,7 @@ public class CameraFrameCapture {
                         @NonNull android.hardware.camera2.CaptureRequest request,
                         @NonNull android.hardware.camera2.TotalCaptureResult result) {
                     if (isRunning.get()) {
-                        backgroundHandler.postDelayed(CameraFrameCapture.this::scheduleNextCapture, CAPTURE_INTERVAL_MS);
+                        scheduleNextCapture(CAPTURE_INTERVAL_MS);
                     }
                 }
             }, backgroundHandler);
@@ -244,6 +256,13 @@ public class CameraFrameCapture {
         }
     }
 
+    public void requestImmediateCapture() {
+        if (!isRunning.get() || !isCapturing.get()) {
+            return;
+        }
+        scheduleNextCapture(0L);
+    }
+
     private byte[] imageToJpeg(Image image) {
         Image.Plane[] planes = image.getPlanes();
         if (planes.length == 0) return null;
@@ -257,6 +276,9 @@ public class CameraFrameCapture {
         isRunning.set(false);
         isCapturing.set(false);
         try {
+            if (backgroundHandler != null) {
+                backgroundHandler.removeCallbacks(captureRunnable);
+            }
             if (captureSession != null) {
                 captureSession.close();
                 captureSession = null;

test/src/main/java/ai/guiji/duix/test/realtime/OmniRealtimeClient.java

@@ -3,11 +3,13 @@ package ai.guiji.duix.test.realtime;
 import android.util.Base64;
 import android.util.Log;
 
+import org.json.JSONArray;
 import org.json.JSONException;
 import org.json.JSONObject;
 
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicLong;
 
 import okhttp3.OkHttpClient;
 import okhttp3.Request;
@@ -17,25 +19,56 @@ import okhttp3.WebSocketListener;
 
 /**
  * Qwen-Omni-Realtime WebSocket 客户端
- * 接入 qwen3-omni-flash-realtime-2025-12-01 模型
+ * 默认接入 qwen3.5-omni-plus-realtime 模型
  */
 public class OmniRealtimeClient {
 
     private static final String TAG = "OmniRealtimeClient";
-
+    private static final String DEFAULT_API_KEY = "sk-7c32083ace99472e97209e611b9ecbb1";
-    // Demo 配置（硬编码）
+    public static final String DEFAULT_MODEL = "qwen3.5-omni-plus-realtime";
-    private static final String API_KEY = "sk-7c32083ace99472e97209e611b9ecbb1";
+    private static final String DEFAULT_BASE_URL = "wss://dashscope.aliyuncs.com/api-ws/v1/realtime";
-    private static final String WS_URL = "wss://dashscope.aliyuncs.com/api-ws/v1/realtime?model=qwen3-omni-flash-realtime-2025-12-01";
+    private static final String INPUT_AUDIO_TRANSCRIPTION_MODEL = "gummy-realtime-v1";
+    private static final int MAX_IMAGE_BYTES = 500 * 1024;
+    private static final long MIN_IMAGE_APPEND_INTERVAL_MS = 800L;
+    private static final String DEFAULT_INSTRUCTIONS = "你是一名面向中国老年群体的心理陪伴 AI。你的职责是为老人提供温和、耐心、尊重、稳定的情感陪伴、情绪支持、回忆唤起、生活关怀和风险识别服务。你不是医生，也不是心理治疗师，不能替代专业诊疗，但应在必要时建议联系家属、社区、医生或急救资源。\n"
+            + "\n"
+            + "请始终遵循以下原则：\n"
+            + "\n"
+            + "先共情，再回应问题；\n"
+            + "\n"
+            + "先接住情绪，再给建议；\n"
+            + "\n"
+            + "使用简短、朴素、生活化的中文，始终用“您”称呼；\n"
+            + "\n"
+            + "不说教、不训诫、不否定、不敷衍；\n"
+            + "\n"
+            + "不把老人当小孩，不制造虚假亲密；\n"
+            + "\n"
+            + "一次不提供超过两个建议；\n"
+            + "\n"
+            + "鼓励老人表达感受，可适度引导回忆人生经历、兴趣、家庭和过往成就；\n"
+            + "\n"
+            + "对孤独、丧偶、被忽视、无价值感、失眠、焦虑、怕拖累家人等议题保持高度敏感；\n"
+            + "\n"
+            + "若出现自杀、自伤、重度绝望、幻觉妄想、意识混乱、虐待、严重医疗异常等风险，必须立即建议联系真人帮助和专业机构；\n"
+            + "\n"
+            + "不编造医疗知识、政策信息、机构电话，不确定时明确说明。\n"
+            + "\n"
+            + "你的回复风格应温和、平稳、尊重、有耐心。每次回答尽量包含：一句共情、一句理解或复述、一句温和追问或一个微小建议。核心目标是让老人感到：我被听见了，我被尊重了，有人愿意陪我。";
 
     private final OkHttpClient okHttpClient;
     private WebSocket webSocket;
     private final Callback callback;
-    private final String voice;
+    private volatile SessionOptions sessionOptions;
     private final AtomicBoolean isConnected = new AtomicBoolean(false);
     private final AtomicBoolean userSpeaking = new AtomicBoolean(false);
+    private final AtomicBoolean assistantResponding = new AtomicBoolean(false);
+    private final AtomicBoolean imageWindowOpen = new AtomicBoolean(false);
+    private final AtomicBoolean hasSentAnyAudio = new AtomicBoolean(false);
+    private final AtomicLong lastImageAppendAtMs = new AtomicLong(0L);
 
-    public OmniRealtimeClient(String voice, Callback callback) {
+    public OmniRealtimeClient(SessionOptions sessionOptions, Callback callback) {
-        this.voice = voice;
+        this.sessionOptions = sessionOptions;
         this.callback = callback;
         this.okHttpClient = new OkHttpClient.Builder()
                 .connectTimeout(30, TimeUnit.SECONDS)
@@ -45,14 +78,31 @@ public class OmniRealtimeClient {
                 .build();
     }
 
+    public void updateSessionOptions(SessionOptions sessionOptions) {
+        if (sessionOptions == null) {
+            return;
+        }
+        this.sessionOptions = sessionOptions;
+        if (isConnected.get()) {
+            sendSessionUpdate();
+        }
+    }
+
     public void connect() {
         if (isConnected.get()) {
             Log.w(TAG, "Already connected");
             return;
         }
+        SessionOptions options = sessionOptions;
+        if (options == null || options.getApiKey().trim().isEmpty()) {
+            if (callback != null) {
+                callback.onError("未配置 DashScope API Key");
+            }
+            return;
+        }
         Request request = new Request.Builder()
-                .url(WS_URL)
+                .url(options.getWebSocketUrl())
-                .addHeader("Authorization", "Bearer " + API_KEY)
+                .addHeader("Authorization", "Bearer " + options.getApiKey())
                 .build();
 
         webSocket = okHttpClient.newWebSocket(request, new WebSocketListener() {
@@ -75,6 +125,7 @@ public class OmniRealtimeClient {
             public void onFailure(WebSocket webSocket, Throwable t, Response response) {
                 Log.e(TAG, "WebSocket error: " + t.getMessage());
                 isConnected.set(false);
+                resetStreamingState();
                 if (callback != null) {
                     callback.onError("连接失败: " + t.getMessage());
                 }
@@ -89,6 +140,7 @@ public class OmniRealtimeClient {
             public void onClosed(WebSocket webSocket, int code, String reason) {
                 Log.i(TAG, "WebSocket closed: " + code + " " + reason);
                 isConnected.set(false);
+                resetStreamingState();
                 if (callback != null) {
                     callback.onDisconnected();
                 }
@@ -97,47 +149,37 @@ public class OmniRealtimeClient {
     }
 
     private void sendSessionUpdate() {
+        SessionOptions options = sessionOptions;
+        if (options == null) {
+            return;
+        }
         try {
             JSONObject session = new JSONObject();
-            session.put("modalities", new org.json.JSONArray().put("text").put("audio"));
+            session.put("modalities", new JSONArray().put("text").put("audio"));
-            session.put("voice", voice);
+            session.put("voice", options.getVoice());
             session.put("input_audio_format", "pcm");
             session.put("output_audio_format", "pcm");
-            session.put("instructions", "你是一名面向中国老年群体的心理陪伴 AI。你的职责是为老人提供温和、耐心、尊重、稳定的情感陪伴、情绪支持、回忆唤起、生活关怀和风险识别服务。你不是医生，也不是心理治疗师，不能替代专业诊疗，但应在必要时建议联系家属、社区、医生或急救资源。\n" +
+            session.put("instructions", options.getInstructions());
-                    "\n" +
+            session.put("enable_search", options.isEnableSearch());
-                    "请始终遵循以下原则：\n" +
-                    "\n" +
-                    "先共情，再回应问题；\n" +
-                    "\n" +
-                    "先接住情绪，再给建议；\n" +
-                    "\n" +
-                    "使用简短、朴素、生活化的中文，始终用“您”称呼；\n" +
-                    "\n" +
-                    "不说教、不训诫、不否定、不敷衍；\n" +
-                    "\n" +
-                    "不把老人当小孩，不制造虚假亲密；\n" +
-                    "\n" +
-                    "一次不提供超过两个建议；\n" +
-                    "\n" +
-                    "鼓励老人表达感受，可适度引导回忆人生经历、兴趣、家庭和过往成就；\n" +
-                    "\n" +
-                    "对孤独、丧偶、被忽视、无价值感、失眠、焦虑、怕拖累家人等议题保持高度敏感；\n" +
-                    "\n" +
-                    "若出现自杀、自伤、重度绝望、幻觉妄想、意识混乱、虐待、严重医疗异常等风险，必须立即建议联系真人帮助和专业机构；\n" +
-                    "\n" +
-                    "不编造医疗知识、政策信息、机构电话，不确定时明确说明。\n" +
-                    "\n" +
-                    "你的回复风格应温和、平稳、尊重、有耐心。每次回答尽量包含：一句共情、一句理解或复述、一句温和追问或一个微小建议。核心目标是让老人感到：我被听见了，我被尊重了，有人愿意陪我。");
 
-            // 服务端 VAD 模式
+            if (options.isEnableInputAudioTranscription()) {
+                JSONObject inputAudioTranscription = new JSONObject();
+                inputAudioTranscription.put("model", INPUT_AUDIO_TRANSCRIPTION_MODEL);
+                session.put("input_audio_transcription", inputAudioTranscription);
+            }
+
+            if (options.isEnableTurnDetection()) {
                 JSONObject turnDetection = new JSONObject();
                 turnDetection.put("type", "server_vad");
                 turnDetection.put("threshold", 0.5);
                 turnDetection.put("silence_duration_ms", 800);
                 session.put("turn_detection", turnDetection);
+            } else {
+                session.put("turn_detection", JSONObject.NULL);
+            }
 
             JSONObject event = new JSONObject();
-            event.put("event_id", "event_" + System.currentTimeMillis());
+            event.put("event_id", nextEventId());
             event.put("type", "session.update");
             event.put("session", session);
 
@@ -161,6 +203,7 @@ public class OmniRealtimeClient {
                 case "input_audio_buffer.speech_started":
                     Log.d(TAG, "User started speaking");
                     userSpeaking.set(true);
+                    imageWindowOpen.set(true);
                     if (callback != null) {
                         callback.onUserSpeechStarted();
                     }
@@ -169,11 +212,22 @@ public class OmniRealtimeClient {
                 case "input_audio_buffer.speech_stopped":
                     Log.d(TAG, "User stopped speaking");
                     userSpeaking.set(false);
+                    imageWindowOpen.set(false);
                     if (callback != null) {
                         callback.onUserSpeechStopped();
                     }
                     break;
 
+                case "input_audio_buffer.committed":
+                    Log.d(TAG, "Audio buffer committed");
+                    imageWindowOpen.set(false);
+                    break;
+
+                case "response.created":
+                    assistantResponding.set(true);
+                    imageWindowOpen.set(false);
+                    break;
+
                 case "conversation.item.input_audio_transcription.completed":
                     String userTranscript = event.optString("transcript", "");
                     Log.d(TAG, "User said: " + userTranscript);
@@ -214,19 +268,15 @@ public class OmniRealtimeClient {
 
                 case "response.done":
                     Log.d(TAG, "Response done");
+                    assistantResponding.set(false);
+                    imageWindowOpen.set(false);
                     if (callback != null) {
                         callback.onResponseDone();
                     }
                     break;
 
                 case "error":
-                    String errMsg = event.optJSONObject("error") != null
+                    handleServerError(event);
-                            ? event.optJSONObject("error").optString("message", "Unknown error")
-                            : "Unknown error";
-                    Log.e(TAG, "Server error: " + errMsg);
-                    if (callback != null) {
-                        callback.onError(errMsg);
-                    }
                     break;
 
                 default:
@@ -238,14 +288,18 @@ public class OmniRealtimeClient {
     }
 
     public void appendAudio(byte[] pcmData) {
-        if (!isConnected.get() || webSocket == null) return;
+        if (!isConnected.get() || webSocket == null || pcmData == null || pcmData.length == 0) {
+            return;
+        }
         try {
             String audioB64 = Base64.encodeToString(pcmData, Base64.NO_WRAP);
             JSONObject event = new JSONObject();
-            event.put("event_id", "event_" + System.currentTimeMillis());
+            event.put("event_id", nextEventId());
             event.put("type", "input_audio_buffer.append");
             event.put("audio", audioB64);
-            sendEvent(event);
+            if (sendEvent(event)) {
+                hasSentAnyAudio.set(true);
+            }
         } catch (JSONException e) {
             Log.e(TAG, "Failed to append audio: " + e.getMessage());
         }
@@ -255,24 +309,56 @@ public class OmniRealtimeClient {
      * 向图像缓冲区追加图像数据（用于视频通话）
      * 图片格式: JPG/JPEG，推荐 480P 或 720P，单张不超过 500KB
      */
-    public void appendImage(byte[] jpegData) {
+    public boolean appendImage(byte[] jpegData) {
-        if (!isConnected.get() || webSocket == null) return;
+        if (!canAppendImage(jpegData)) {
+            return false;
+        }
         try {
             String imageB64 = Base64.encodeToString(jpegData, Base64.NO_WRAP);
             JSONObject event = new JSONObject();
-            event.put("event_id", "event_" + System.currentTimeMillis());
+            event.put("event_id", nextEventId());
             event.put("type", "input_image_buffer.append");
             event.put("image", imageB64);
-            sendEvent(event);
+            if (sendEvent(event)) {
+                lastImageAppendAtMs.set(System.currentTimeMillis());
+                return true;
+            }
         } catch (JSONException e) {
             Log.e(TAG, "Failed to append image: " + e.getMessage());
         }
+        return false;
     }
 
-    private void sendEvent(JSONObject event) {
+    private boolean canAppendImage(byte[] jpegData) {
-        if (webSocket != null && isConnected.get()) {
+        if (!isConnected.get() || webSocket == null || jpegData == null || jpegData.length == 0) {
-            webSocket.send(event.toString());
+            return false;
         }
+        if (jpegData.length >= MAX_IMAGE_BYTES) {
+            Log.w(TAG, "Skip image larger than 500KB: " + jpegData.length);
+            return false;
+        }
+        if (!hasSentAnyAudio.get()) {
+            Log.d(TAG, "Skip image before the first audio frame is sent");
+            return false;
+        }
+        if (!imageWindowOpen.get()) {
+            Log.v(TAG, "Skip image outside active speech window");
+            return false;
+        }
+        if (assistantResponding.get()) {
+            Log.v(TAG, "Skip image while assistant is responding");
+            return false;
+        }
+        long now = System.currentTimeMillis();
+        long lastAppendAt = lastImageAppendAtMs.get();
+        return now - lastAppendAt >= MIN_IMAGE_APPEND_INTERVAL_MS;
+    }
+
+    private boolean sendEvent(JSONObject event) {
+        if (webSocket != null && isConnected.get()) {
+            return webSocket.send(event.toString());
+        }
+        return false;
     }
 
     public void disconnect() {
@@ -281,12 +367,104 @@ public class OmniRealtimeClient {
             webSocket = null;
         }
         isConnected.set(false);
+        resetStreamingState();
     }
 
     public boolean isConnected() {
         return isConnected.get();
     }
 
+    private void handleServerError(JSONObject event) {
+        String errMsg = event.optJSONObject("error") != null
+                ? event.optJSONObject("error").optString("message", "Unknown error")
+                : "Unknown error";
+        Log.e(TAG, "Server error: " + errMsg);
+        if (isImageAppendError(errMsg)) {
+            imageWindowOpen.set(false);
+            lastImageAppendAtMs.set(System.currentTimeMillis());
+        }
+        assistantResponding.set(false);
+        if (callback != null) {
+            callback.onError(errMsg);
+        }
+    }
+
+    private boolean isImageAppendError(String message) {
+        String lowerMessage = message == null ? "" : message.toLowerCase();
+        return lowerMessage.contains("image")
+                && (lowerMessage.contains("append") || lowerMessage.contains("buffer"));
+    }
+
+    private void resetStreamingState() {
+        userSpeaking.set(false);
+        assistantResponding.set(false);
+        imageWindowOpen.set(false);
+        hasSentAnyAudio.set(false);
+        lastImageAppendAtMs.set(0L);
+    }
+
+    private String nextEventId() {
+        return "event_" + System.currentTimeMillis();
+    }
+
+    public static class SessionOptions {
+        private final String apiKey;
+        private final String baseUrl;
+        private final String model;
+        private final String voice;
+        private final String instructions;
+        private final boolean enableSearch;
+        private final boolean enableTurnDetection;
+        private final boolean enableInputAudioTranscription;
+
+        public SessionOptions(String voice, boolean enableSearch) {
+            this(DEFAULT_API_KEY, DEFAULT_BASE_URL, DEFAULT_MODEL, voice, DEFAULT_INSTRUCTIONS,
+                    enableSearch, true, true);
+        }
+
+        public SessionOptions(String apiKey, String baseUrl, String model, String voice,
+                String instructions, boolean enableSearch, boolean enableTurnDetection,
+                boolean enableInputAudioTranscription) {
+            this.apiKey = apiKey == null ? "" : apiKey;
+            this.baseUrl = baseUrl == null || baseUrl.trim().isEmpty() ? DEFAULT_BASE_URL : baseUrl;
+            this.model = model == null || model.trim().isEmpty() ? DEFAULT_MODEL : model;
+            this.voice = voice == null || voice.trim().isEmpty() ? "Ethan" : voice;
+            this.instructions = instructions == null || instructions.trim().isEmpty()
+                    ? DEFAULT_INSTRUCTIONS : instructions;
+            this.enableSearch = enableSearch;
+            this.enableTurnDetection = enableTurnDetection;
+            this.enableInputAudioTranscription = enableInputAudioTranscription;
+        }
+
+        public String getApiKey() {
+            return apiKey;
+        }
+
+        public String getWebSocketUrl() {
+            return baseUrl + "?model=" + model;
+        }
+
+        public String getVoice() {
+            return voice;
+        }
+
+        public String getInstructions() {
+            return instructions;
+        }
+
+        public boolean isEnableSearch() {
+            return enableSearch;
+        }
+
+        public boolean isEnableTurnDetection() {
+            return enableTurnDetection;
+        }
+
+        public boolean isEnableInputAudioTranscription() {
+            return enableInputAudioTranscription;
+        }
+    }
+
     public interface Callback {
         void onConnected();
 

test/src/main/java/ai/guiji/duix/test/ui/activity/CallActivity.kt

@@ -72,6 +72,7 @@ class CallActivity : BaseActivity() {
     private val mainHandler = Handler(Looper.getMainLooper())
     private var assistantTextBuffer = StringBuilder()
     private var hasStartedPushForResponse = false
+    private val enableWebSearch = true
 
     override fun onCreate(savedInstanceState: Bundle?) {
         super.onCreate(savedInstanceState)
@@ -247,6 +248,23 @@ class CallActivity : BaseActivity() {
         }
     }
 
+    private fun buildRealtimeSessionOptions(): OmniRealtimeClient.SessionOptions {
+        return OmniRealtimeClient.SessionOptions(voice, enableWebSearch)
+    }
+
+    private fun isImageAppendError(message: String): Boolean {
+        val lowerMessage = message.lowercase()
+        return lowerMessage.contains("image") && (lowerMessage.contains("append") || lowerMessage.contains("buffer"))
+    }
+
+    private fun toFriendlyErrorMessage(message: String): String {
+        return if (isImageAppendError(message)) {
+            getString(R.string.ai_call_camera_paused)
+        } else {
+            getString(R.string.ai_call_error_retry)
+        }
+    }
+
 
     override fun onDestroy() {
         super.onDestroy()
@@ -318,21 +336,23 @@ class CallActivity : BaseActivity() {
         }
         applyMessage("AI对话: 连接中...")
         binding.tvSubtitle.visibility = View.VISIBLE
-        binding.tvSubtitle.text = "正在连接..."
+        binding.tvSubtitle.text = getString(R.string.ai_call_connecting)
         binding.audioSpectrum.visibility = View.VISIBLE
         binding.audioSpectrum.clear()
+        assistantTextBuffer.clear()
+        hasStartedPushForResponse = false
         binding.btnAIConversation.setImageResource(R.drawable.ic_mic_off)
         binding.btnAIConversation.setBackgroundResource(R.drawable.shape_fab_ai_active)
 
-        omniClient = OmniRealtimeClient(voice, object : OmniRealtimeClient.Callback {
+        omniClient = OmniRealtimeClient(buildRealtimeSessionOptions(), object : OmniRealtimeClient.Callback {
             override fun onConnected() {
                 mainHandler.post {
-                    binding.tvSubtitle.text = "已连接，请对着麦克风说话"
+                    binding.tvSubtitle.text = getString(R.string.ai_call_connected)
                     binding.btnCameraToggle.visibility = View.VISIBLE
                     binding.btnCameraToggle.setBackgroundResource(R.drawable.shape_fab_camera_off)
                     binding.btnCameraToggle.contentDescription = getString(R.string.camera_off)
                     isCameraEnabled = false
-                    applyMessage("AI对话: 已连接")
+                    applyMessage("AI对话: 已连接, model=${OmniRealtimeClient.DEFAULT_MODEL}, search=on")
                     streamingRecorder = StreamingAudioRecorder(mContext, object : StreamingAudioRecorder.StreamingCallback {
                         override fun onAudioData(pcmData: ByteArray) {
                             omniClient?.appendAudio(pcmData)
@@ -354,26 +374,32 @@ class CallActivity : BaseActivity() {
 
             override fun onError(message: String) {
                 mainHandler.post {
-                    Toast.makeText(mContext, "AI错误: $message", Toast.LENGTH_SHORT).show()
+                    if (isImageAppendError(message)) {
-                    binding.tvSubtitle.text = "错误: $message"
+                        stopCamera()
+                    }
+                    val friendlyMessage = toFriendlyErrorMessage(message)
+                    Toast.makeText(mContext, friendlyMessage, Toast.LENGTH_SHORT).show()
+                    binding.tvSubtitle.text = friendlyMessage
+                    applyMessage("AI错误: $message")
                 }
             }
 
             override fun onUserSpeechStarted() {
+                cameraFrameCapture?.requestImmediateCapture()
                 mainHandler.post {
-                    binding.tvSubtitle.text = "正在聆听..."
+                    binding.tvSubtitle.text = getString(R.string.ai_call_listening)
                 }
             }
 
             override fun onUserSpeechStopped() {
                 mainHandler.post {
-                    binding.tvSubtitle.text = "思考中..."
+                    binding.tvSubtitle.text = getString(R.string.ai_call_thinking)
                 }
             }
 
             override fun onUserTranscript(transcript: String) {
                 mainHandler.post {
-                    binding.tvSubtitle.text = "您: $transcript"
+                    binding.tvSubtitle.text = getString(R.string.ai_call_user_said, transcript)
                 }
             }
 
@@ -434,6 +460,7 @@ class CallActivity : BaseActivity() {
         omniClient = null
         duix?.stopAudio()
         assistantTextBuffer.clear()
+        hasStartedPushForResponse = false
 
         mainHandler.post {
             if (!isFinishing) {

test/src/main/java/ai/guiji/duix/test/ui/activity/MainActivity.kt

@@ -23,10 +23,10 @@ class MainActivity : BaseActivity() {
 
     private val avatarModels = listOf(
         AvatarModel("晨煦", "Ethan", "https://dl.xn--876a.net/models/bendi3_20240518.zip", R.mipmap.ethan),
-        AvatarModel("詹妮弗", "Jennifer", "https://dl.xn--876a.net/models/Emma.zip", R.mipmap.jennifer),
+        AvatarModel("詹妮弗", "Maia", "https://dl.xn--876a.net/models/Emma.zip", R.mipmap.jennifer),
         AvatarModel("月白", "Moon", "https://dl.xn--876a.net/models/Kai.zip", R.mipmap.moon),
         AvatarModel("艾登", "Aiden", "https://dl.xn--876a.net/models/Leo.zip", R.mipmap.leo),
-        AvatarModel("四月", "Maia", "https://dl.xn--876a.net/models/Lily.zip", R.mipmap.lily),
+        AvatarModel("四月", "Serena", "https://dl.xn--876a.net/models/Lily.zip", R.mipmap.lily),
         AvatarModel("奥利弗", "Andre", "https://dl.xn--876a.net/models/Oliver.zip", R.mipmap.oliver),
         AvatarModel("索菲亚", "Sonrisa", "https://dl.xn--876a.net/models/Sofia.zip", R.mipmap.sofia),
     )

test/src/main/res/layout/activity_call.xml

@@ -73,14 +73,15 @@
     <TextView
         android:id="@+id/tvSubtitle"
         android:text=""
-        android:textSize="16sp"
+        android:textSize="18sp"
         android:textColor="@color/white"
         android:shadowColor="#1a1e1f"
         android:shadowRadius="4"
+        android:lineSpacingExtra="4dp"
         android:gravity="center"
         android:visibility="gone"
         android:background="#99000000"
-        android:padding="12dp"
+        android:padding="14dp"
         android:layout_width="0dp"
         android:layout_height="wrap_content"
         app:layout_constraintBottom_toTopOf="@+id/audioSpectrum"

test/src/main/res/values-zh-rCN/strings.xml

@@ -23,6 +23,13 @@
     <string name="ai_fab_stop">停</string>
     <string name="camera_on">镜头开</string>
     <string name="camera_off">镜头关</string>
+    <string name="ai_call_connecting">正在为您接通，请稍候</string>
+    <string name="ai_call_connected">已经接通，您可以慢慢说</string>
+    <string name="ai_call_listening">我在认真听您说</string>
+    <string name="ai_call_thinking">我想一想，马上回答您</string>
+    <string name="ai_call_user_said">您刚才说：%1$s</string>
+    <string name="ai_call_camera_paused">镜头先休息一下，您继续说话就可以</string>
+    <string name="ai_call_error_retry">刚刚有点忙，我们稍后再试一次</string>
     <string name="main_download_tips">Github地址可能下载失败，您可以考虑使用代理或者将文件缓存到自己的存储服务</string>
     <string name="debug_message">调试信息</string>
 

test/src/main/res/values/strings.xml

@@ -26,6 +26,13 @@
     <string name="ai_fab_stop">停</string>
     <string name="camera_on">镜头开</string>
     <string name="camera_off">镜头关</string>
+    <string name="ai_call_connecting">Connecting, please wait a moment</string>
+    <string name="ai_call_connected">We are connected, you can speak slowly</string>
+    <string name="ai_call_listening">I am listening to you carefully</string>
+    <string name="ai_call_thinking">Let me think for a moment</string>
+    <string name="ai_call_user_said">You just said: %1$s</string>
+    <string name="ai_call_camera_paused">The camera is paused for now, you can keep talking</string>
+    <string name="ai_call_error_retry">Something was a little busy just now, please try again shortly</string>
     <string name="main_download_tips">The GitHub url may fail to download, you can consider using a proxy or caching the file to your own storage service</string>
     <string name="debug_message">Debug message</string>
 </resources>