@yh-ui/hooks 0.1.12 → 0.1.16

package/dist/use-ai/use-ai-stream.mjs

@@ -44,6 +44,37 @@ export const qwenParser = (raw) => {
   }
   return text || null;
 };
+export const claudeParser = (raw) => {
+  const lines = raw.split("\n");
+  let text = "";
+  for (const line of lines) {
+    if (!line.startsWith("data: ")) continue;
+    const data = line.slice(6).trim();
+    try {
+      const json = JSON.parse(data);
+      if (json?.type === "content_block_delta" && json?.delta?.text) {
+        text += json.delta.text;
+      }
+    } catch {
+    }
+  }
+  return text || null;
+};
+export const geminiParser = (raw) => {
+  const lines = raw.split("\n");
+  let text = "";
+  for (const line of lines) {
+    const content = line.startsWith("data: ") ? line.slice(6).trim() : line.trim();
+    if (!content) continue;
+    try {
+      const json = JSON.parse(content);
+      const part = json?.candidates?.[0]?.content?.parts?.[0]?.text;
+      if (part) text += part;
+    } catch {
+    }
+  }
+  return text || null;
+};
 export const plainTextParser = (raw) => raw || null;
 class TypewriterThrottle {
   queue = [];
@@ -167,6 +198,13 @@ export function useAiStream(options) {
     fetchStream,
     stop,
     // 暴露解析器供测试/自定义使用
-    parsers: { openaiParser, ernieParser, qwenParser, plainTextParser }
+    parsers: {
+      openaiParser,
+      ernieParser,
+      qwenParser,
+      claudeParser,
+      geminiParser,
+      plainTextParser
+    }
   };
 }

package/dist/use-ai/use-ai-voice.cjs

@@ -0,0 +1,218 @@
+"use strict";
+
+Object.defineProperty(exports, "__esModule", {
+  value: true
+});
+exports.useAiVoice = useAiVoice;
+var _vue = require("vue");
+function useAiVoice(options = {}) {
+  const {
+    language = "zh-CN",
+    interimResults = true,
+    continuous = false,
+    vad = true,
+    vadThreshold = 2e3,
+    volumeThreshold = 0.05,
+    waveCount = 20,
+    useSTT = true
+  } = options;
+  const isRecording = (0, _vue.ref)(false);
+  const transcript = (0, _vue.ref)("");
+  const interimTranscript = (0, _vue.ref)("");
+  const volume = (0, _vue.ref)(0);
+  const amplitudes = (0, _vue.ref)(new Array(waveCount).fill(5));
+  const audioBlob = (0, _vue.ref)(null);
+  const recognition = (0, _vue.shallowRef)(null);
+  const audioContext = (0, _vue.shallowRef)(null);
+  const analyser = (0, _vue.shallowRef)(null);
+  const stream = (0, _vue.shallowRef)(null);
+  const mediaRecorder = (0, _vue.shallowRef)(null);
+  let chunks = [];
+  let animationId = null;
+  let silenceStart = null;
+  const _window = typeof window !== "undefined" ? window : null;
+  const SpeechRecognition = _window?.SpeechRecognition || _window?.webkitSpeechRecognition;
+  const sttSupported = !!SpeechRecognition;
+  const initMediaRecorder = mediaStream => {
+    chunks = [];
+    const recorder = new MediaRecorder(mediaStream);
+    recorder.ondataavailable = e => {
+      if (e.data.size > 0) chunks.push(e.data);
+    };
+    recorder.onstop = () => {
+      audioBlob.value = new Blob(chunks, {
+        type: "audio/webm"
+      });
+      if (isRecording.value === false && chunks.length > 0) {
+        options.onStop?.(transcript.value, audioBlob.value);
+      }
+    };
+    mediaRecorder.value = recorder;
+  };
+  const initRecognition = () => {
+    if (!sttSupported || !useSTT) return;
+    const recognitionInstance = new SpeechRecognition();
+    recognitionInstance.lang = language;
+    recognitionInstance.interimResults = interimResults;
+    recognitionInstance.continuous = continuous;
+    recognitionInstance.onresult = event => {
+      let currentInterim = "";
+      for (let i = event.resultIndex; i < event.results.length; ++i) {
+        if (event.results[i].isFinal) {
+          transcript.value += event.results[i][0].transcript;
+          options.onResult?.(transcript.value);
+        } else {
+          currentInterim += event.results[i][0].transcript;
+        }
+      }
+      interimTranscript.value = currentInterim;
+      options.onPartialResult?.(currentInterim);
+    };
+    recognitionInstance.onerror = event => {
+      if (event.error !== "no-speech" && event.error !== "aborted") {
+        options.onError?.(event);
+      }
+    };
+    recognition.value = recognitionInstance;
+  };
+  const initAudioAnalyzer = async mediaStream => {
+    try {
+      const AudioCtx = window.AudioContext || window.webkitAudioContext;
+      audioContext.value = new AudioCtx();
+      if (audioContext.value.state === "suspended") {
+        await audioContext.value.resume();
+      }
+      analyser.value = audioContext.value.createAnalyser();
+      analyser.value.fftSize = 256;
+      const source = audioContext.value.createMediaStreamSource(mediaStream);
+      source.connect(analyser.value);
+      const bufferLength = analyser.value.frequencyBinCount;
+      const dataArray = new Uint8Array(bufferLength);
+      const process = () => {
+        if (!isRecording.value) {
+          amplitudes.value = new Array(waveCount).fill(5);
+          volume.value = 0;
+          return;
+        }
+        animationId = requestAnimationFrame(process);
+        analyser.value.getByteFrequencyData(dataArray);
+        let total = 0;
+        for (let i = 0; i < bufferLength; i++) total += dataArray[i];
+        const avg = total / bufferLength;
+        volume.value = Math.min(100, avg / 128 * 100);
+        const step = Math.floor(bufferLength / waveCount);
+        const newAmps = [];
+        for (let i = 0; i < waveCount; i++) {
+          const val = dataArray[i * step];
+          newAmps.push(6 + val / 255 * 34);
+        }
+        amplitudes.value = newAmps;
+        if (vad) {
+          const normalizedVol = avg / 255;
+          if (normalizedVol < volumeThreshold) {
+            if (silenceStart === null) silenceStart = Date.now();else if (Date.now() - silenceStart > vadThreshold) {
+              stop();
+            }
+          } else {
+            silenceStart = null;
+          }
+        }
+      };
+      process();
+    } catch (err) {
+      options.onError?.(err);
+    }
+  };
+  const start = async () => {
+    if (isRecording.value) return;
+    try {
+      transcript.value = "";
+      interimTranscript.value = "";
+      audioBlob.value = null;
+      silenceStart = null;
+      stream.value = await navigator.mediaDevices.getUserMedia({
+        audio: true
+      });
+      isRecording.value = true;
+      initMediaRecorder(stream.value);
+      initRecognition();
+      await initAudioAnalyzer(stream.value);
+      mediaRecorder.value?.start(1e3);
+      recognition.value?.start();
+      options.onStart?.();
+    } catch (err) {
+      isRecording.value = false;
+      if (stream.value) {
+        stream.value.getTracks().forEach(t => t.stop());
+        stream.value = null;
+      }
+      console.error("[yh-ui/hooks] useAiVoice start failed:", err);
+      options.onError?.(err);
+    }
+  };
+  const stop = () => {
+    if (!isRecording.value) return;
+    isRecording.value = false;
+    if (stream.value) {
+      stream.value.getTracks().forEach(track => track.stop());
+      stream.value = null;
+    }
+    if (recognition.value) {
+      try {
+        recognition.value.stop();
+      } catch {}
+    }
+    if (mediaRecorder.value && mediaRecorder.value.state !== "inactive") {
+      try {
+        mediaRecorder.value.stop();
+      } catch {}
+    }
+    cleanup();
+  };
+  const cancel = () => {
+    if (!isRecording.value) return;
+    isRecording.value = false;
+    if (stream.value) {
+      stream.value.getTracks().forEach(track => track.stop());
+      stream.value = null;
+    }
+    if (recognition.value) {
+      try {
+        recognition.value.abort();
+      } catch {}
+    }
+    if (mediaRecorder.value && mediaRecorder.value.state !== "inactive") {
+      try {
+        mediaRecorder.value.stop();
+      } catch {}
+    }
+    cleanup();
+  };
+  const cleanup = () => {
+    if (animationId) {
+      cancelAnimationFrame(animationId);
+      animationId = null;
+    }
+    if (audioContext.value && audioContext.value.state !== "closed") {
+      audioContext.value.close().catch(_err => {});
+      audioContext.value = null;
+    }
+    amplitudes.value = new Array(waveCount).fill(5);
+    volume.value = 0;
+  };
+  (0, _vue.onUnmounted)(() => {
+    if (isRecording.value) stop();else cleanup();
+  });
+  return {
+    isRecording,
+    transcript,
+    interimTranscript,
+    amplitudes,
+    volume,
+    audioBlob,
+    start,
+    stop,
+    cancel,
+    sttSupported
+  };
+}

package/dist/use-ai/use-ai-voice.d.ts

@@ -0,0 +1,83 @@
+export interface UseAiVoiceOptions {
+    /**
+     * 语言代码 (用于 SpeechRecognition)
+     * @default 'zh-CN'
+     */
+    language?: string;
+    /**
+     * 是否需要临时结果（在说话过程中实时返回）
+     * @default true
+     */
+    interimResults?: boolean;
+    /**
+     * 是否连续识别
+     * @default false
+     */
+    continuous?: boolean;
+    /**
+     * 智能静音检测（VAD）
+     * 开启后，当检测到长时间无声会自动停止录音
+     * @default true
+     */
+    vad?: boolean;
+    /**
+     * 静音检测阈值 (ms)
+     * @default 2000
+     */
+    vadThreshold?: number;
+    /**
+     * 音量变化敏感度 (0-1)
+     * @default 0.05
+     */
+    volumeThreshold?: number;
+    /**
+     * 返回波形柱的数量（对应 AiVoiceTrigger 的 amplitudes）
+     * @default 20
+     */
+    waveCount?: number;
+    /**
+     * 是否在开始时自动执行浏览器语音识别 (SpeechRecognition)
+     * 如果关闭，则只进行物理音频录制
+     * @default true
+     */
+    useSTT?: boolean;
+    /** 回调事件 */
+    onStart?: () => void;
+    /** 停止回调，包含最终转写文本和录音文件 Blob */
+    onStop?: (transcript: string, blob: Blob | null) => void;
+    onResult?: (transcript: string) => void;
+    onPartialResult?: (transcript: string) => void;
+    onError?: (error: unknown) => void;
+}
+export interface UseAiVoiceReturn {
+    /** 是否正在录音 */
+    isRecording: import('vue').Ref<boolean>;
+    /** 最终转写文本 */
+    transcript: import('vue').Ref<string>;
+    /** 过程中的临时文本 */
+    interimTranscript: import('vue').Ref<string>;
+    /** 实时波形数据 */
+    amplitudes: import('vue').Ref<number[]>;
+    /** 实时音量 (0-100) */
+    volume: import('vue').Ref<number>;
+    /** 录音文件的 Blob */
+    audioBlob: import('vue').Ref<Blob | null>;
+    /** 开始录音 */
+    start: () => Promise<void>;
+    /** 停止录音 */
+    stop: () => void;
+    /** 取消并放弃当前结果 */
+    cancel: () => void;
+    /** 浏览器是否支持 SpeechRecognition (用于显示警告) */
+    sttSupported: boolean;
+}
+/**
+ * useAiVoice - 专业级 AI 语音交互 Hook
+ *
+ * 核心能力：
+ * 1. 【音频录制】：通过 MediaRecorder 真实录制音频并导出 Blob 文件。
+ * 2. 【视觉分析】：通过 Web Audio API 实时输出驱动 AiVoiceTrigger 的波形数组。
+ * 3. 【智能 VAD】：多维检测静音状态，支持自动停顿结束。
+ * 4. 【语音转写】：内置 Web Speech API 实时转写及临时结果反馈。
+ */
+export declare function useAiVoice(options?: UseAiVoiceOptions): UseAiVoiceReturn;

package/dist/use-ai/use-ai-voice.mjs

@@ -0,0 +1,215 @@
+import { ref, onUnmounted, shallowRef } from "vue";
+export function useAiVoice(options = {}) {
+  const {
+    language = "zh-CN",
+    interimResults = true,
+    continuous = false,
+    vad = true,
+    vadThreshold = 2e3,
+    volumeThreshold = 0.05,
+    waveCount = 20,
+    useSTT = true
+  } = options;
+  const isRecording = ref(false);
+  const transcript = ref("");
+  const interimTranscript = ref("");
+  const volume = ref(0);
+  const amplitudes = ref(new Array(waveCount).fill(5));
+  const audioBlob = ref(null);
+  const recognition = shallowRef(null);
+  const audioContext = shallowRef(null);
+  const analyser = shallowRef(null);
+  const stream = shallowRef(null);
+  const mediaRecorder = shallowRef(null);
+  let chunks = [];
+  let animationId = null;
+  let silenceStart = null;
+  const _window = typeof window !== "undefined" ? window : null;
+  const SpeechRecognition = _window?.SpeechRecognition || _window?.webkitSpeechRecognition;
+  const sttSupported = !!SpeechRecognition;
+  const initMediaRecorder = (mediaStream) => {
+    chunks = [];
+    const recorder = new MediaRecorder(mediaStream);
+    recorder.ondataavailable = (e) => {
+      if (e.data.size > 0) chunks.push(e.data);
+    };
+    recorder.onstop = () => {
+      audioBlob.value = new Blob(chunks, { type: "audio/webm" });
+      if (isRecording.value === false && chunks.length > 0) {
+        options.onStop?.(transcript.value, audioBlob.value);
+      }
+    };
+    mediaRecorder.value = recorder;
+  };
+  const initRecognition = () => {
+    if (!sttSupported || !useSTT) return;
+    const recognitionInstance = new SpeechRecognition();
+    recognitionInstance.lang = language;
+    recognitionInstance.interimResults = interimResults;
+    recognitionInstance.continuous = continuous;
+    recognitionInstance.onresult = (event) => {
+      let currentInterim = "";
+      for (let i = event.resultIndex; i < event.results.length; ++i) {
+        if (event.results[i].isFinal) {
+          transcript.value += event.results[i][0].transcript;
+          options.onResult?.(transcript.value);
+        } else {
+          currentInterim += event.results[i][0].transcript;
+        }
+      }
+      interimTranscript.value = currentInterim;
+      options.onPartialResult?.(currentInterim);
+    };
+    recognitionInstance.onerror = (event) => {
+      if (event.error !== "no-speech" && event.error !== "aborted") {
+        options.onError?.(event);
+      }
+    };
+    recognition.value = recognitionInstance;
+  };
+  const initAudioAnalyzer = async (mediaStream) => {
+    try {
+      const AudioCtx = window.AudioContext || window.webkitAudioContext;
+      audioContext.value = new AudioCtx();
+      if (audioContext.value.state === "suspended") {
+        await audioContext.value.resume();
+      }
+      analyser.value = audioContext.value.createAnalyser();
+      analyser.value.fftSize = 256;
+      const source = audioContext.value.createMediaStreamSource(mediaStream);
+      source.connect(analyser.value);
+      const bufferLength = analyser.value.frequencyBinCount;
+      const dataArray = new Uint8Array(bufferLength);
+      const process = () => {
+        if (!isRecording.value) {
+          amplitudes.value = new Array(waveCount).fill(5);
+          volume.value = 0;
+          return;
+        }
+        animationId = requestAnimationFrame(process);
+        analyser.value.getByteFrequencyData(dataArray);
+        let total = 0;
+        for (let i = 0; i < bufferLength; i++) total += dataArray[i];
+        const avg = total / bufferLength;
+        volume.value = Math.min(100, avg / 128 * 100);
+        const step = Math.floor(bufferLength / waveCount);
+        const newAmps = [];
+        for (let i = 0; i < waveCount; i++) {
+          const val = dataArray[i * step];
+          newAmps.push(6 + val / 255 * 34);
+        }
+        amplitudes.value = newAmps;
+        if (vad) {
+          const normalizedVol = avg / 255;
+          if (normalizedVol < volumeThreshold) {
+            if (silenceStart === null) silenceStart = Date.now();
+            else if (Date.now() - silenceStart > vadThreshold) {
+              stop();
+            }
+          } else {
+            silenceStart = null;
+          }
+        }
+      };
+      process();
+    } catch (err) {
+      options.onError?.(err);
+    }
+  };
+  const start = async () => {
+    if (isRecording.value) return;
+    try {
+      transcript.value = "";
+      interimTranscript.value = "";
+      audioBlob.value = null;
+      silenceStart = null;
+      stream.value = await navigator.mediaDevices.getUserMedia({ audio: true });
+      isRecording.value = true;
+      initMediaRecorder(stream.value);
+      initRecognition();
+      await initAudioAnalyzer(stream.value);
+      mediaRecorder.value?.start(1e3);
+      recognition.value?.start();
+      options.onStart?.();
+    } catch (err) {
+      isRecording.value = false;
+      if (stream.value) {
+        stream.value.getTracks().forEach((t) => t.stop());
+        stream.value = null;
+      }
+      console.error("[yh-ui/hooks] useAiVoice start failed:", err);
+      options.onError?.(err);
+    }
+  };
+  const stop = () => {
+    if (!isRecording.value) return;
+    isRecording.value = false;
+    if (stream.value) {
+      stream.value.getTracks().forEach((track) => track.stop());
+      stream.value = null;
+    }
+    if (recognition.value) {
+      try {
+        recognition.value.stop();
+      } catch {
+      }
+    }
+    if (mediaRecorder.value && mediaRecorder.value.state !== "inactive") {
+      try {
+        mediaRecorder.value.stop();
+      } catch {
+      }
+    }
+    cleanup();
+  };
+  const cancel = () => {
+    if (!isRecording.value) return;
+    isRecording.value = false;
+    if (stream.value) {
+      stream.value.getTracks().forEach((track) => track.stop());
+      stream.value = null;
+    }
+    if (recognition.value) {
+      try {
+        recognition.value.abort();
+      } catch {
+      }
+    }
+    if (mediaRecorder.value && mediaRecorder.value.state !== "inactive") {
+      try {
+        mediaRecorder.value.stop();
+      } catch {
+      }
+    }
+    cleanup();
+  };
+  const cleanup = () => {
+    if (animationId) {
+      cancelAnimationFrame(animationId);
+      animationId = null;
+    }
+    if (audioContext.value && audioContext.value.state !== "closed") {
+      audioContext.value.close().catch((_err) => {
+      });
+      audioContext.value = null;
+    }
+    amplitudes.value = new Array(waveCount).fill(5);
+    volume.value = 0;
+  };
+  onUnmounted(() => {
+    if (isRecording.value) stop();
+    else cleanup();
+  });
+  return {
+    isRecording,
+    transcript,
+    interimTranscript,
+    amplitudes,
+    volume,
+    audioBlob,
+    start,
+    stop,
+    cancel,
+    sttSupported
+  };
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yh-ui/hooks",
-  "version": "0.1.12",
+  "version": "0.1.16",
   "description": "YH-UI composition hooks",
   "type": "module",
   "sideEffects": false,
@@ -24,8 +24,8 @@
   ],
   "dependencies": {
     "dayjs": "^1.11.19",
-    "@yh-ui/locale": "0.1.12",
-    "@yh-ui/utils": "0.1.12"
+    "@yh-ui/locale": "0.1.16",
+    "@yh-ui/utils": "0.1.16"
   },
   "devDependencies": {
     "vue": "^3.5.27",