@wq-hook/volcano-react 1.0.0 → 1.0.2

package/dist/index.js

@@ -32,8 +32,13 @@ var index_exports = {};
 __export(index_exports, {
   AudioProgressBar: () => AudioProgressBar_default,
   AudioWaveVisualizer: () => AudioWaveVisualizer_default,
+  StreamingTextSplitter: () => StreamingTextSplitter,
+  clearSessionAudioCache: () => clearSessionAudioCache,
+  findSessionCacheByText: () => findSessionCacheByText,
+  getSessionAudioCache: () => getSessionAudioCache,
   splitTextByDelimiters: () => splitTextByDelimiters,
   useMessageTTS: () => useMessageTTS,
+  useStreamTTS: () => useStreamTTS,
   useVolcanoASR: () => useVolcanoASR,
   useVolcanoTTS: () => useVolcanoTTS
 });
@@ -550,10 +555,12 @@ function useMessageTTS({
   const [isSynthesizing, setIsSynthesizing] = (0, import_react3.useState)(false);
   const [error, setErrorState] = (0, import_react3.useState)(null);
   const [progress, setProgress] = (0, import_react3.useState)(0);
-  const [visualizationData, setVisualizationData] = (0, import_react3.useState)({
-    frequencyData: new Uint8Array(0),
-    timeDomainData: new Uint8Array(0)
-  });
+  const [visualizationData, setVisualizationData] = (0, import_react3.useState)(
+    {
+      frequencyData: new Uint8Array(0),
+      timeDomainData: new Uint8Array(0)
+    }
+  );
   const instanceId = (0, import_react3.useRef)(
     `tts-${Date.now()}-${Math.random().toString(36).slice(2)}`
   ).current;
@@ -761,7 +768,11 @@ function useMessageTTS({
           console.error("Audio playback error:", e, audio.error);
           metricsCollector.record({
             name: "tts_error",
-            labels: { error_code: "playback_error", voice, detail: audio.error?.message || String(audio.error?.code) },
+            labels: {
+              error_code: "playback_error",
+              voice,
+              detail: audio.error?.message || String(audio.error?.code)
+            },
             value: 1,
             timestamp: Date.now()
           });
@@ -779,7 +790,10 @@ function useMessageTTS({
           }
         };
         if (cachedData) {
-          const totalSize = cachedData.reduce((acc, buf) => acc + buf.byteLength, 0);
+          const totalSize = cachedData.reduce(
+            (acc, buf) => acc + buf.byteLength,
+            0
+          );
           metricsCollector.record({
             name: "tts_cache_hit",
             labels: { voice, speed },
@@ -797,7 +811,9 @@ function useMessageTTS({
             })
           );
           if (totalSize === 0) {
-            console.warn("[useMessageTTS] Cached data is empty, falling back to stream");
+            console.warn(
+              "[useMessageTTS] Cached data is empty, falling back to stream"
+            );
           } else {
             const blob = new Blob(cachedData, { type: "audio/mpeg" });
             const url2 = URL.createObjectURL(blob);
@@ -816,7 +832,10 @@ function useMessageTTS({
         }
         console.log("[useMessageTTS] Cache miss, starting stream");
         clientRef.current = (0, import_tts2.WebsocketMSE)({ autoStartSession: true });
-        const formattedText = import_volcano_sdk2.MarkdownFormatter.format(text).replace((0, import_emoji_regex2.default)(), "");
+        const formattedText = import_volcano_sdk2.MarkdownFormatter.format(text).replace(
+          (0, import_emoji_regex2.default)(),
+          ""
+        );
         const segments = splitTextByDelimiters(formattedText);
         const url = clientRef.current.start({
           url: buildFullUrl2(WS_URL, {
@@ -934,6 +953,14 @@ function useMessageTTS({
         console.warn(
           `[useMessageTTS] Voice ${failedVoice} failed, switching to fallback voice ${fallbackVoice}`
         );
+        if (clientRef.current) {
+          clientRef.current.close();
+          clientRef.current = null;
+        }
+        if (audioRef.current) {
+          audioRef.current.pause();
+          audioRef.current = null;
+        }
         executeTTS(text, fallbackVoice);
       } else {
         playFallback(text);
@@ -943,7 +970,7 @@ function useMessageTTS({
   );
   const play = (0, import_react3.useCallback)(
     (text) => {
-      const voice = audioParams?.speaker || "default";
+      const voice = audioParams?.speaker || "zh_female_vv_uranus_bigtts";
       return executeTTS(text, voice);
     },
     [audioParams, executeTTS]
@@ -996,7 +1023,9 @@ function useMessageTTS({
     if (audioRef.current) {
       let duration = audioRef.current.duration;
       if (!isFinite(duration) && audioRef.current.buffered.length > 0) {
-        duration = audioRef.current.buffered.end(audioRef.current.buffered.length - 1);
+        duration = audioRef.current.buffered.end(
+          audioRef.current.buffered.length - 1
+        );
       }
       if (isFinite(duration) && duration > 0) {
         const time = percentage / 100 * duration;
@@ -1025,8 +1054,725 @@ function useMessageTTS({
   };
 }
 
-// src/components/AudioWaveVisualizer.tsx
+// src/tts/useStreamTTS.ts
+var import_tts3 = require("@wq-hook/volcano-sdk/tts");
 var import_react4 = require("react");
+
+// src/tts/StreamingTextSplitter.ts
+var import_volcano_sdk3 = require("@wq-hook/volcano-sdk");
+var import_emoji_regex3 = __toESM(require("emoji-regex"));
+var StreamingTextSplitter = class {
+  constructor(options = {}) {
+    /** 当前缓冲区 */
+    this.buffer = "";
+    /** 分段索引计数器 */
+    this.segmentIndex = 0;
+    /** 已完成的分段列表 */
+    this.segments = [];
+    /** 是否已完成 */
+    this.isCompleted = false;
+    this.maxLength = options.maxLength || 150;
+    this.minLength = options.minLength || 10;
+    this.onSegmentComplete = options.onSegmentComplete;
+    this.onAllComplete = options.onAllComplete;
+  }
+  /**
+   * 接收流式文本块
+   * @param chunk - 文本块
+   */
+  onChunk(chunk) {
+    if (!chunk || this.isCompleted) return;
+    this.buffer += chunk;
+    if (this.detectBoundary(chunk)) {
+      const newlineIndex = this.buffer.indexOf("\n");
+      if (newlineIndex !== -1) {
+        if (newlineIndex === 0) {
+          this.buffer = this.buffer.substring(1);
+          return;
+        }
+        const segmentBuffer = this.buffer.substring(0, newlineIndex);
+        this.buffer = this.buffer.substring(newlineIndex + 1);
+        this.flushSegmentWithBuffer(segmentBuffer);
+        while (this.buffer.includes("\n")) {
+          const nextNewlineIndex = this.buffer.indexOf("\n");
+          if (nextNewlineIndex === 0) {
+            this.buffer = this.buffer.substring(1);
+            continue;
+          }
+          const nextSegmentBuffer = this.buffer.substring(0, nextNewlineIndex);
+          this.buffer = this.buffer.substring(nextNewlineIndex + 1);
+          this.flushSegmentWithBuffer(nextSegmentBuffer);
+        }
+      }
+    }
+  }
+  /**
+   * 检测分段边界
+   * @param chunk - 最新接收的文本块
+   * @returns 是否应该分段
+   */
+  detectBoundary(chunk) {
+    if (chunk.includes("\n")) {
+      if (this.buffer.length >= this.maxLength) {
+        this.forceSplitAtSentenceBoundary();
+      }
+      return true;
+    }
+    if (this.buffer.length >= this.maxLength) {
+      this.forceSplitAtSentenceBoundary();
+      return true;
+    }
+    return false;
+  }
+  /**
+   * 在句子边界强制拆分超长段落
+   */
+  forceSplitAtSentenceBoundary() {
+    const content = this.buffer;
+    const sentenceEnders = /[。？！]/g;
+    let lastMatch = null;
+    let match = null;
+    while ((match = sentenceEnders.exec(content)) !== null) {
+      lastMatch = match;
+    }
+    if (lastMatch && lastMatch.index > this.minLength) {
+      const splitPoint = lastMatch.index + 1;
+      const firstPart = content.substring(0, splitPoint);
+      const secondPart = content.substring(splitPoint);
+      this.buffer = firstPart;
+      this.flushSegment();
+      this.buffer = secondPart;
+    } else {
+      const midPoint = Math.floor(content.length / 2);
+      const firstPart = content.substring(0, midPoint);
+      const secondPart = content.substring(midPoint);
+      this.buffer = firstPart;
+      this.flushSegment();
+      this.buffer = secondPart;
+    }
+  }
+  /**
+   * 使用指定缓冲区内容刷新为分段
+   * @param bufferToFlush - 要分段的缓冲区内容
+   */
+  flushSegmentWithBuffer(bufferToFlush) {
+    const content = bufferToFlush;
+    if (!content) return;
+    const isPureSymbols = /^[^\p{L}\p{N}]*$/u.test(content);
+    const isTooShort = content.length < 3;
+    if (isPureSymbols && isTooShort) {
+      return;
+    }
+    const formattedContent = import_volcano_sdk3.MarkdownFormatter.format(content).replace((0, import_emoji_regex3.default)(), "");
+    if (!formattedContent) return;
+    let subSegments = [formattedContent];
+    if (formattedContent.length > this.maxLength) {
+      subSegments = this.splitLongSegment(formattedContent);
+    }
+    for (const subSegment of subSegments) {
+      if (!subSegment) continue;
+      const segment = {
+        index: this.segmentIndex++,
+        content: subSegment,
+        length: subSegment.length,
+        sent: false
+      };
+      this.segments.push(segment);
+      this.onSegmentComplete?.(segment);
+    }
+  }
+  /**
+   * 刷新当前缓冲区为分段
+   */
+  flushSegment() {
+    const content = this.buffer.trim();
+    if (!content) {
+      this.buffer = "";
+      return;
+    }
+    const isPureSymbols = /^[^\p{L}\p{N}]*$/u.test(content);
+    const isTooShort = content.length < 3;
+    if (isPureSymbols && isTooShort) {
+      this.buffer = "";
+      return;
+    }
+    const formattedContent = import_volcano_sdk3.MarkdownFormatter.format(content).replace((0, import_emoji_regex3.default)(), "");
+    if (!formattedContent) {
+      this.buffer = "";
+      return;
+    }
+    let subSegments = [formattedContent];
+    if (formattedContent.length > this.maxLength) {
+      subSegments = this.splitLongSegment(formattedContent);
+    }
+    for (const subSegment of subSegments) {
+      if (!subSegment) continue;
+      const segment = {
+        index: this.segmentIndex++,
+        content: subSegment,
+        length: subSegment.length,
+        sent: false
+      };
+      this.segments.push(segment);
+      this.onSegmentComplete?.(segment);
+    }
+    this.buffer = "";
+  }
+  /**
+   * 拆分超长分段
+   * @param segment - 超长的分段
+   * @returns 拆分后的分段数组
+   */
+  splitLongSegment(segment) {
+    const result = [];
+    let current = "";
+    for (const char of segment) {
+      current += char;
+      const shouldSplit = /[。？！，,]/.test(char);
+      if (shouldSplit && current.length <= this.maxLength) {
+        result.push(current);
+        current = "";
+      } else if (current.length >= this.maxLength) {
+        result.push(current);
+        current = "";
+      }
+    }
+    if (current) {
+      result.push(current);
+    }
+    return result.filter((s) => s.length > 0);
+  }
+  /**
+   * 完成流式输入
+   * 处理剩余的缓冲区内容
+   */
+  complete() {
+    if (this.isCompleted) return;
+    this.isCompleted = true;
+    while (this.buffer.includes("\n")) {
+      const newlineIndex = this.buffer.indexOf("\n");
+      if (newlineIndex === 0) {
+        this.buffer = this.buffer.substring(1);
+        continue;
+      }
+      const segmentBuffer = this.buffer.substring(0, newlineIndex);
+      this.buffer = this.buffer.substring(newlineIndex + 1);
+      this.flushSegmentWithBuffer(segmentBuffer);
+    }
+    if (this.buffer.trim()) {
+      this.flushSegment();
+    }
+    this.onAllComplete?.(this.segments);
+  }
+  /**
+   * 重置分段器状态
+   */
+  reset() {
+    this.buffer = "";
+    this.segmentIndex = 0;
+    this.segments = [];
+    this.isCompleted = false;
+  }
+  /**
+   * 获取当前缓冲区内容
+   */
+  getBuffer() {
+    return this.buffer;
+  }
+  /**
+   * 获取已分段的列表
+   */
+  getSegments() {
+    return this.segments;
+  }
+  /**
+   * 获取统计信息
+   */
+  getStats() {
+    return {
+      bufferLength: this.buffer.length,
+      segmentCount: this.segments.length,
+      totalChars: this.segments.reduce((sum, seg) => sum + seg.length, 0)
+    };
+  }
+};
+
+// src/tts/useStreamTTS.ts
+var WS_URL2 = "wss://openspeech.bytedance.com/api/v3/tts/bidirection";
+var activeInstances2 = /* @__PURE__ */ new Map();
+var sessionAudioCache = /* @__PURE__ */ new Map();
+function buildFullUrl3(url, params) {
+  const arr = [];
+  for (const key in params) {
+    if (Object.prototype.hasOwnProperty.call(params, key)) {
+      arr.push(`${key}=${encodeURIComponent(params[key])}`);
+    }
+  }
+  return `${url}?${arr.join("&")}`;
+}
+function useStreamTTS({
+  ttsConfig,
+  audioParams,
+  autoPlay = true,
+  metricsCollector = new NoopMetricsCollector(),
+  onPlayStart,
+  onPlayPause,
+  onPlayResume,
+  onPlayEnd,
+  onError,
+  visualization,
+  maxSegmentLength = 150
+}) {
+  const [isConnected, setIsConnected] = (0, import_react4.useState)(false);
+  const [isSessionStarted, setIsSessionStarted] = (0, import_react4.useState)(false);
+  const [isSynthesizing, setIsSynthesizing] = (0, import_react4.useState)(false);
+  const [isPlaying, setIsPlaying] = (0, import_react4.useState)(false);
+  const [isPaused, setIsPaused] = (0, import_react4.useState)(false);
+  const [error, setErrorState] = (0, import_react4.useState)(null);
+  const [streamText, setStreamText] = (0, import_react4.useState)("");
+  const [progress, setProgress] = (0, import_react4.useState)(0);
+  const [visualizationData, setVisualizationData] = (0, import_react4.useState)({
+    frequencyData: new Uint8Array(0),
+    timeDomainData: new Uint8Array(0)
+  });
+  const instanceId = (0, import_react4.useRef)(`tts-stream-${Date.now()}-${Math.random().toString(36).slice(2)}`).current;
+  const clientRef = (0, import_react4.useRef)(null);
+  const audioRef = (0, import_react4.useRef)(null);
+  const audioContextRef = (0, import_react4.useRef)(null);
+  const analyserRef = (0, import_react4.useRef)(null);
+  const sourceRef = (0, import_react4.useRef)(null);
+  const audioUrlRef = (0, import_react4.useRef)(null);
+  const streamTextRef = (0, import_react4.useRef)("");
+  const isConnectedRef = (0, import_react4.useRef)(false);
+  const isSessionStartedRef = (0, import_react4.useRef)(false);
+  const calledSessionStartedRef = (0, import_react4.useRef)(false);
+  const splitterRef = (0, import_react4.useRef)(null);
+  const segmentQueueRef = (0, import_react4.useRef)([]);
+  const isSendingRef = (0, import_react4.useRef)(false);
+  const sessionAudioBuffersRef = (0, import_react4.useRef)([]);
+  const isStreamFinishedRef = (0, import_react4.useRef)(false);
+  const isSessionFinishedRef = (0, import_react4.useRef)(false);
+  const resolveAllSegmentsSentRef = (0, import_react4.useRef)(null);
+  const currentVoiceRef = (0, import_react4.useRef)("");
+  const initAudioContext = (0, import_react4.useCallback)(() => {
+    if (!audioRef.current) return;
+    if (!audioContextRef.current) {
+      const AudioContextClass = window.AudioContext || window.webkitAudioContext;
+      audioContextRef.current = new AudioContextClass();
+    }
+    if (audioContextRef.current.state === "suspended") {
+      audioContextRef.current.resume();
+    }
+    if (!analyserRef.current) {
+      analyserRef.current = audioContextRef.current.createAnalyser();
+      analyserRef.current.fftSize = visualization?.fftSize || 256;
+    }
+    if (!sourceRef.current) {
+      try {
+        sourceRef.current = audioContextRef.current.createMediaElementSource(audioRef.current);
+        sourceRef.current.connect(analyserRef.current);
+        analyserRef.current.connect(audioContextRef.current.destination);
+      } catch (e) {
+      }
+    }
+  }, [visualization?.fftSize]);
+  const cleanupAudio = (0, import_react4.useCallback)(() => {
+    if (audioUrlRef.current) {
+      URL.revokeObjectURL(audioUrlRef.current);
+      audioUrlRef.current = null;
+    }
+    if (audioRef.current) {
+      audioRef.current.onerror = null;
+      audioRef.current.onended = null;
+      audioRef.current.onpause = null;
+      audioRef.current.onplay = null;
+      audioRef.current.ontimeupdate = null;
+      audioRef.current.pause();
+      audioRef.current.src = "";
+      audioRef.current = null;
+    }
+    if (sourceRef.current) {
+      try {
+        sourceRef.current.disconnect();
+      } catch (e) {
+      }
+      sourceRef.current = null;
+    }
+  }, []);
+  const stopOthers = (0, import_react4.useCallback)(() => {
+    activeInstances2.forEach((instance, id) => {
+      if (id !== instanceId) {
+        instance.pause();
+      }
+    });
+  }, [instanceId]);
+  const pause = (0, import_react4.useCallback)(() => {
+    if (audioRef.current) {
+      audioRef.current.pause();
+    }
+    setIsPaused(true);
+    setIsPlaying(false);
+    onPlayPause?.();
+  }, [onPlayPause]);
+  const resume = (0, import_react4.useCallback)(() => {
+    stopOthers();
+    if (audioRef.current) {
+      audioRef.current.play();
+    }
+    setIsPaused(false);
+    setIsPlaying(true);
+    onPlayResume?.();
+    activeInstances2.set(instanceId, { pause });
+  }, [stopOthers, instanceId, pause, onPlayResume]);
+  const sendNextSegment = (0, import_react4.useCallback)(() => {
+    if (!clientRef.current || !isSessionStartedRef.current || isSendingRef.current || isSessionFinishedRef.current) {
+      return;
+    }
+    if (segmentQueueRef.current.length === 0) {
+      if (isStreamFinishedRef.current && !isSessionFinishedRef.current) {
+        console.log("[useStreamTTS] All segments sent, finishing session");
+        isSessionFinishedRef.current = true;
+        clientRef.current.finishSession();
+        resolveAllSegmentsSentRef.current?.();
+      }
+      return;
+    }
+    isSendingRef.current = true;
+    const segment = segmentQueueRef.current.shift();
+    console.log(`[useStreamTTS] Sending segment ${segment.index}: ${segment.content.substring(0, 30)}...`);
+    clientRef.current.sendText(segment.content);
+    segment.sent = true;
+    isSendingRef.current = false;
+    setTimeout(() => sendNextSegment(), 0);
+  }, []);
+  const stop = (0, import_react4.useCallback)(() => {
+    if (clientRef.current) {
+      clientRef.current.close();
+      clientRef.current = null;
+    }
+    cleanupAudio();
+    setIsConnected(false);
+    isConnectedRef.current = false;
+    setIsSessionStarted(false);
+    isSessionStartedRef.current = false;
+    calledSessionStartedRef.current = false;
+    setIsPlaying(false);
+    setIsPaused(false);
+    setIsSynthesizing(false);
+    setProgress(0);
+    activeInstances2.delete(instanceId);
+    streamTextRef.current = "";
+    setStreamText("");
+    segmentQueueRef.current = [];
+    isSendingRef.current = false;
+    sessionAudioBuffersRef.current = [];
+    isStreamFinishedRef.current = false;
+    isSessionFinishedRef.current = false;
+    splitterRef.current?.reset();
+  }, [cleanupAudio, instanceId]);
+  const connect = (0, import_react4.useCallback)(async () => {
+    stop();
+    setErrorState(null);
+    setProgress(0);
+    sessionAudioBuffersRef.current = [];
+    isStreamFinishedRef.current = false;
+    streamTextRef.current = "";
+    setStreamText("");
+    segmentQueueRef.current = [];
+    isSendingRef.current = false;
+    isSessionStartedRef.current = false;
+    calledSessionStartedRef.current = false;
+    setIsSessionStarted(false);
+    const voice = audioParams?.speaker || "zh_female_vv_uranus_bigtts";
+    currentVoiceRef.current = voice;
+    const startTime = Date.now();
+    metricsCollector.record({
+      name: "tts_request",
+      labels: { voice, text_length: 0 },
+      value: 1,
+      timestamp: startTime
+    });
+    try {
+      const audio = new Audio();
+      audio.crossOrigin = "anonymous";
+      audioRef.current = audio;
+      audio.onplay = () => {
+        setIsPlaying(true);
+        setIsPaused(false);
+        onPlayStart?.();
+        initAudioContext();
+        activeInstances2.set(instanceId, { pause });
+      };
+      audio.onended = () => {
+        setIsPlaying(false);
+        setIsPaused(false);
+        onPlayEnd?.();
+        activeInstances2.delete(instanceId);
+      };
+      audio.onerror = (e) => {
+        console.error("[useStreamTTS] Audio playback error:", e, audio.error);
+        setErrorState(audio.error?.message || "Audio playback error");
+        onError?.(new Error(audio.error?.message || "Audio playback error"));
+      };
+      audio.ontimeupdate = () => {
+        let duration = audio.duration;
+        if (!isFinite(duration) && audio.buffered.length > 0) {
+          duration = audio.buffered.end(audio.buffered.length - 1);
+        }
+        if (isFinite(duration) && duration > 0) {
+          setProgress(audio.currentTime / duration * 100);
+        }
+      };
+      clientRef.current = (0, import_tts3.WebsocketMSE)({ autoStartSession: false });
+      splitterRef.current = new StreamingTextSplitter({
+        maxLength: maxSegmentLength,
+        onSegmentComplete: (segment) => {
+          segmentQueueRef.current.push(segment);
+          console.log(`[useStreamTTS] Segment ${segment.index} queued (${segment.length} chars)`);
+          if (isSessionStartedRef.current) {
+            sendNextSegment();
+          }
+        },
+        onAllComplete: () => {
+          console.log(`[useStreamTTS] All segments completed, total: ${segmentQueueRef.current.length} in queue`);
+        }
+      });
+      const url = clientRef.current.start({
+        url: buildFullUrl3(WS_URL2, {
+          api_access_key: `Jwt; ${ttsConfig.token}`,
+          api_app_key: ttsConfig.appid,
+          api_resource_id: ttsConfig.resourceId || "seed-tts-2.0"
+        }),
+        config: {
+          user: {
+            uid: `req-${Date.now()}`
+          },
+          namespace: ttsConfig.namespace || "BidirectionalTTS",
+          req_params: {
+            speaker: voice,
+            audio_params: {
+              sample_rate: audioParams?.sample_rate || 24e3,
+              format: audioParams?.format || "mp3",
+              speech_rate: audioParams?.speech_rate,
+              pitch_rate: audioParams?.pitch_rate,
+              loudness_rate: audioParams?.loudness_rate
+            },
+            additions: JSON.stringify({
+              enable_language_detector: true,
+              disable_markdown_filter: true,
+              enable_latex_tn: true
+            })
+          }
+        },
+        // ===== 关键回调 =====
+        onStart: () => {
+          setIsConnected(true);
+          isConnectedRef.current = true;
+          console.log("[useStreamTTS] WebSocket connected, waiting for text...");
+        },
+        onSessionStarted: () => {
+          setIsSessionStarted(true);
+          isSessionStartedRef.current = true;
+          console.log("[useStreamTTS] Session started, can send text now");
+          if (segmentQueueRef.current.length > 0) {
+            sendNextSegment();
+          }
+        },
+        onMessage: (data) => {
+          setIsSynthesizing(true);
+          if (sessionAudioBuffersRef.current.length === 0) {
+            metricsCollector.record({
+              name: "tts_latency",
+              labels: { stage: "first_packet", voice },
+              value: Date.now() - startTime,
+              timestamp: Date.now()
+            });
+          }
+          const buffer = data instanceof ArrayBuffer ? data.slice(0) : new Uint8Array(data).buffer;
+          sessionAudioBuffersRef.current.push(buffer);
+        },
+        onSessionFinished: () => {
+          setIsSynthesizing(false);
+          setIsSessionStarted(false);
+          isSessionStartedRef.current = false;
+          calledSessionStartedRef.current = false;
+          if (sessionAudioBuffersRef.current.length > 0 && streamTextRef.current) {
+            const speed = audioParams?.speech_rate || 0;
+            const cacheKey = TTSCache.generateKey(streamTextRef.current, voice, speed);
+            TTSCache.set(cacheKey, [...sessionAudioBuffersRef.current]);
+            sessionAudioCache.set(instanceId, {
+              streamText: streamTextRef.current,
+              audioBuffers: [...sessionAudioBuffersRef.current],
+              timestamp: Date.now(),
+              voice,
+              speed
+            });
+            console.log(`[useStreamTTS] Session finished, cached ${sessionAudioBuffersRef.current.length} audio buffers`);
+          }
+          metricsCollector.record({
+            name: "tts_synthesis_finished",
+            labels: { voice, text_length: streamTextRef.current.length },
+            value: Date.now() - startTime,
+            timestamp: Date.now()
+          });
+        },
+        onError: (err) => {
+          console.error("[useStreamTTS] TTS error:", err);
+          setErrorState(err.msg || "TTS error");
+          onError?.(new Error(err.msg || "TTS error"));
+          setIsSynthesizing(false);
+        }
+      });
+      audioUrlRef.current = url;
+      audio.src = url;
+      if (autoPlay) {
+        try {
+          await audio.play();
+        } catch (e) {
+          console.warn("[useStreamTTS] Autoplay blocked:", e);
+        }
+      }
+    } catch (err) {
+      console.error("[useStreamTTS] Connect error:", err);
+      setErrorState(String(err));
+      onError?.(err instanceof Error ? err : new Error(String(err)));
+    }
+  }, [
+    ttsConfig,
+    audioParams,
+    autoPlay,
+    stop,
+    instanceId,
+    onPlayStart,
+    onPlayEnd,
+    initAudioContext,
+    pause,
+    metricsCollector,
+    maxSegmentLength,
+    sendNextSegment,
+    onError
+  ]);
+  const onMessage = (0, import_react4.useCallback)((chunk) => {
+    if (!chunk) return;
+    streamTextRef.current += chunk;
+    setStreamText(streamTextRef.current);
+    if (!calledSessionStartedRef.current && !isSessionStartedRef.current && clientRef.current && isConnectedRef.current) {
+      console.log("[useStreamTTS] First text received, starting session...");
+      calledSessionStartedRef.current = true;
+      clientRef.current.startSession();
+    }
+    splitterRef.current?.onChunk(chunk);
+  }, []);
+  const finishStream = (0, import_react4.useCallback)(async () => {
+    isStreamFinishedRef.current = true;
+    splitterRef.current?.complete();
+    console.log(`[useStreamTTS] Stream finished, ${segmentQueueRef.current.length} segments remaining in queue`);
+    if (segmentQueueRef.current.length > 0 || isSendingRef.current) {
+      await new Promise((resolve) => {
+        resolveAllSegmentsSentRef.current = resolve;
+      });
+    } else if (clientRef.current && isSessionStartedRef.current && !isSessionFinishedRef.current) {
+      isSessionFinishedRef.current = true;
+      clientRef.current.finishSession();
+    }
+  }, []);
+  const seek = (0, import_react4.useCallback)((percentage) => {
+    if (audioRef.current) {
+      let duration = audioRef.current.duration;
+      if (!isFinite(duration) && audioRef.current.buffered.length > 0) {
+        duration = audioRef.current.buffered.end(audioRef.current.buffered.length - 1);
+      }
+      if (isFinite(duration) && duration > 0) {
+        const time = percentage / 100 * duration;
+        if (isFinite(time)) {
+          audioRef.current.currentTime = time;
+          setProgress(percentage);
+        }
+      }
+    }
+  }, []);
+  const getFrequencyData = (0, import_react4.useCallback)(() => {
+    if (!analyserRef.current) return new Uint8Array(0);
+    const dataArray = new Uint8Array(analyserRef.current.frequencyBinCount);
+    analyserRef.current.getByteFrequencyData(dataArray);
+    return dataArray;
+  }, []);
+  const getTimeDomainData = (0, import_react4.useCallback)(() => {
+    if (!analyserRef.current) return new Uint8Array(0);
+    const dataArray = new Uint8Array(analyserRef.current.frequencyBinCount);
+    analyserRef.current.getByteTimeDomainData(dataArray);
+    return dataArray;
+  }, []);
+  (0, import_react4.useEffect)(() => {
+    if (!visualization?.enabled) return;
+    let animId;
+    let lastUpdate = 0;
+    const interval = visualization.refreshInterval || 0;
+    const update = (timestamp) => {
+      if (isPlaying && !isPaused) {
+        if (timestamp - lastUpdate >= interval) {
+          setVisualizationData({
+            frequencyData: getFrequencyData(),
+            timeDomainData: getTimeDomainData()
+          });
+          lastUpdate = timestamp;
+        }
+        animId = requestAnimationFrame(update);
+      }
+    };
+    if (isPlaying && !isPaused) {
+      animId = requestAnimationFrame(update);
+    }
+    return () => {
+      if (animId) cancelAnimationFrame(animId);
+    };
+  }, [isPlaying, isPaused, visualization, getFrequencyData, getTimeDomainData]);
+  (0, import_react4.useEffect)(() => {
+    return () => {
+      stop();
+      if (audioContextRef.current) {
+        audioContextRef.current.close();
+      }
+    };
+  }, [stop]);
+  return {
+    isConnected,
+    isSessionStarted,
+    isSynthesizing,
+    isPlaying,
+    isPaused,
+    error,
+    streamText,
+    progress,
+    connect,
+    onMessage,
+    finishStream,
+    pause,
+    resume,
+    stop,
+    seek,
+    getFrequencyData,
+    getTimeDomainData,
+    visualizationData
+  };
+}
+function getSessionAudioCache(instanceId) {
+  return sessionAudioCache.get(instanceId);
+}
+function clearSessionAudioCache(instanceId) {
+  sessionAudioCache.delete(instanceId);
+}
+function findSessionCacheByText(streamText, voice, speed) {
+  for (const entry of sessionAudioCache.values()) {
+    if (entry.streamText === streamText && entry.voice === voice && entry.speed === speed) {
+      return entry;
+    }
+  }
+  return void 0;
+}
+
+// src/components/AudioWaveVisualizer.tsx
+var import_react5 = require("react");
 var import_jsx_runtime = require("react/jsx-runtime");
 var AudioWaveVisualizer = ({
   isPlaying,
@@ -1041,8 +1787,8 @@ var AudioWaveVisualizer = ({
   className,
   styleObj
 }) => {
-  const canvasRef = (0, import_react4.useRef)(null);
-  const requestRef = (0, import_react4.useRef)(null);
+  const canvasRef = (0, import_react5.useRef)(null);
+  const requestRef = (0, import_react5.useRef)(null);
   const progressBackground = Array.isArray(color) ? `linear-gradient(90deg, ${color[0]}, ${color[1]})` : color;
   const textColor = Array.isArray(color) ? color[0] : color;
   const draw = () => {
@@ -1107,7 +1853,7 @@ var AudioWaveVisualizer = ({
       requestRef.current = requestAnimationFrame(draw);
     }
   };
-  (0, import_react4.useEffect)(() => {
+  (0, import_react5.useEffect)(() => {
     if (isPlaying && !isPaused) {
       requestRef.current = requestAnimationFrame(draw);
     } else {
@@ -1151,7 +1897,7 @@ var AudioWaveVisualizer = ({
 var AudioWaveVisualizer_default = AudioWaveVisualizer;
 
 // src/components/AudioProgressBar.tsx
-var import_react5 = require("react");
+var import_react6 = require("react");
 var import_jsx_runtime2 = require("react/jsx-runtime");
 var AudioProgressBar = ({
   progress,
@@ -1164,16 +1910,16 @@ var AudioProgressBar = ({
   className,
   style
 }) => {
-  const progressBarRef = (0, import_react5.useRef)(null);
-  const containerRef = (0, import_react5.useRef)(null);
-  const progressTextRef = (0, import_react5.useRef)(null);
-  const thumbRef = (0, import_react5.useRef)(null);
-  const displayedProgress = (0, import_react5.useRef)(0);
-  const requestRef = (0, import_react5.useRef)(null);
-  const isDragging = (0, import_react5.useRef)(false);
-  const isHovering = (0, import_react5.useRef)(false);
-  const isTouch = (0, import_react5.useRef)(false);
-  (0, import_react5.useEffect)(() => {
+  const progressBarRef = (0, import_react6.useRef)(null);
+  const containerRef = (0, import_react6.useRef)(null);
+  const progressTextRef = (0, import_react6.useRef)(null);
+  const thumbRef = (0, import_react6.useRef)(null);
+  const displayedProgress = (0, import_react6.useRef)(0);
+  const requestRef = (0, import_react6.useRef)(null);
+  const isDragging = (0, import_react6.useRef)(false);
+  const isHovering = (0, import_react6.useRef)(false);
+  const isTouch = (0, import_react6.useRef)(false);
+  (0, import_react6.useEffect)(() => {
     const match = window.matchMedia("(pointer: coarse)");
     isTouch.current = match.matches;
     if (isTouch.current && thumbRef.current) {
@@ -1212,7 +1958,7 @@ var AudioProgressBar = ({
       requestRef.current = requestAnimationFrame(animate);
     }
   };
-  (0, import_react5.useEffect)(() => {
+  (0, import_react6.useEffect)(() => {
     requestRef.current = requestAnimationFrame(animate);
     return () => {
       if (requestRef.current) {
@@ -1399,8 +2145,13 @@ var AudioProgressBar_default = AudioProgressBar;
 0 && (module.exports = {
   AudioProgressBar,
   AudioWaveVisualizer,
+  StreamingTextSplitter,
+  clearSessionAudioCache,
+  findSessionCacheByText,
+  getSessionAudioCache,
   splitTextByDelimiters,
   useMessageTTS,
+  useStreamTTS,
   useVolcanoASR,
   useVolcanoTTS
 });