@pie-players/pie-section-player 0.3.17 → 0.3.19

package/dist/index-BRKsQ6Im-_GY8ip1I.js

@@ -0,0 +1,442 @@
+var k = Object.defineProperty, A = (i, t, e) => t in i ? k(i, t, { enumerable: !0, configurable: !0, writable: !0, value: e }) : i[t] = e, h = (i, t, e) => A(i, typeof t != "symbol" ? t + "" : t, e);
+const b = (i) => {
+  const t = (i.providerOptions && typeof i.providerOptions == "object" ? i.providerOptions : {}).__pieTelemetry;
+  return typeof t == "function" ? t : void 0;
+}, R = {
+  pie: 3e3,
+  custom: 3e3
+}, f = (i) => i.replace(/\/+$/, ""), I = (i) => {
+  const t = f(i.apiEndpoint), e = (i.provider || "").toLowerCase();
+  return e === "polly" || e === "google" ? `${t}/${e}/voices` : `${t}/voices`;
+}, T = (i) => i.transportMode === "custom" ? "custom" : i.transportMode === "pie" ? "pie" : i.provider === "custom" ? "custom" : "pie", S = (i, t) => i.endpointMode ? i.endpointMode : t === "custom" ? "rootPost" : "synthesizePath", P = (i, t) => i.endpointValidationMode ? i.endpointValidationMode : t === "custom" ? "none" : "voices", M = (i) => {
+  const t = i.providerOptions || {};
+  if (typeof t.speedRate == "string")
+    return t.speedRate;
+  const e = Number(i.rate ?? 1);
+  return !Number.isFinite(e) || e <= 0.95 ? "slow" : e >= 1.5 ? "fast" : "medium";
+}, E = (i) => {
+  const t = [];
+  let e = 0;
+  const o = i.split(`
+`).map((r) => r.trim()).filter(Boolean);
+  for (const r of o)
+    try {
+      const s = JSON.parse(r), a = typeof s.type == "string" ? s.type : "word", n = typeof s.time == "number" && Number.isFinite(s.time) ? s.time : 0, l = typeof s.value == "string" ? s.value : "", d = typeof s.start == "number" && Number.isFinite(s.start) ? s.start : null, u = typeof s.end == "number" && Number.isFinite(s.end) ? s.end : null, y = d ?? e, m = u ?? y + Math.max(1, l.length || String(s.value || "").length);
+      e = Math.max(m + 1, e), t.push({
+        time: n,
+        type: a,
+        start: y,
+        end: m,
+        value: l
+      });
+    } catch {
+    }
+  return t;
+}, C = {
+  id: "pie",
+  resolveSynthesisUrl: (i) => {
+    const t = S(i, "pie"), e = f(i.apiEndpoint);
+    return t === "rootPost" ? e : `${e}/synthesize`;
+  },
+  buildRequestBody: (i, t) => {
+    const e = t.providerOptions || {}, o = typeof t.engine == "string" ? t.engine : typeof e.engine == "string" ? e.engine : void 0, r = typeof e.sampleRate == "number" && Number.isFinite(e.sampleRate) ? e.sampleRate : void 0, s = e.format === "mp3" || e.format === "ogg" || e.format === "pcm" ? e.format : void 0, a = Array.isArray(e.speechMarkTypes) ? e.speechMarkTypes.filter((n) => n === "word" || n === "sentence" || n === "ssml") : void 0;
+    return {
+      text: i,
+      provider: t.provider || "polly",
+      voice: t.voice,
+      language: t.language,
+      rate: t.rate,
+      engine: o,
+      sampleRate: r,
+      format: s,
+      speechMarkTypes: a,
+      includeSpeechMarks: !0
+    };
+  },
+  parseResponse: async (i) => {
+    const t = await i.json();
+    return {
+      audio: {
+        kind: "base64",
+        data: t.audio,
+        contentType: t.contentType
+      },
+      speechMarks: Array.isArray(t.speechMarks) ? t.speechMarks : []
+    };
+  }
+}, z = {
+  id: "custom",
+  resolveSynthesisUrl: (i) => {
+    const t = S(i, "custom"), e = f(i.apiEndpoint);
+    return t === "synthesizePath" ? `${e}/synthesize` : e;
+  },
+  buildRequestBody: (i, t) => {
+    const e = t.providerOptions || {}, o = typeof e.lang_id == "string" ? e.lang_id : t.language || "en-US", r = typeof e.cache == "boolean" ? e.cache : !0;
+    return {
+      text: i,
+      speedRate: M(t),
+      lang_id: o,
+      cache: r
+    };
+  },
+  parseResponse: async (i, t, e, o) => {
+    const r = await i.json(), s = {};
+    if (t.includeAuthOnAssetFetch)
+      for (const [n, l] of Object.entries(e))
+        n.toLowerCase() === "authorization" && (s[n] = l);
+    let a = [];
+    if (typeof r.word == "string" && r.word.length > 0) {
+      const n = await fetch(r.word, {
+        headers: s,
+        signal: o
+      });
+      if (n.ok) {
+        const l = await n.text();
+        a = E(l);
+      }
+    }
+    return {
+      audio: {
+        kind: "url",
+        url: r.audioContent
+      },
+      speechMarks: a
+    };
+  }
+}, B = {
+  pie: C,
+  custom: z
+};
+class O {
+  constructor(t, e) {
+    h(this, "config"), h(this, "adapter"), h(this, "currentAudio", null), h(this, "pausedState", !1), h(this, "wordTimings", []), h(this, "highlightInterval", null), h(this, "intentionallyStopped", !1), h(this, "activeSynthesisController", null), h(this, "synthesisRunId", 0), h(this, "telemetryReporter"), h(this, "onWordBoundary"), this.config = t, this.adapter = e, this.telemetryReporter = b(t);
+  }
+  async emitTelemetry(t, e) {
+    try {
+      await this.telemetryReporter?.(t, e);
+    } catch (o) {
+      console.warn("[ServerTTSProvider] telemetry callback failed:", o);
+    }
+  }
+  async speak(t) {
+    this.stop(), this.intentionallyStopped = !1;
+    const e = ++this.synthesisRunId, o = new AbortController();
+    this.activeSynthesisController = o;
+    const { audioUrl: r, wordTimings: s } = await this.synthesizeSpeech(t, o.signal, e);
+    if (e !== this.synthesisRunId) {
+      URL.revokeObjectURL(r);
+      return;
+    }
+    const a = this.config.rate || 1;
+    return this.wordTimings = s.map((n) => ({
+      ...n,
+      time: n.time / a
+    })), new Promise((n, l) => {
+      const d = new Audio(r);
+      this.currentAudio = d, this.config.rate && (d.playbackRate = Math.max(0.25, Math.min(4, this.config.rate))), this.config.volume !== void 0 && (d.volume = Math.max(0, Math.min(1, this.config.volume))), d.onplay = () => {
+        this.pausedState = !1, this.onWordBoundary && this.wordTimings.length > 0 && this.startWordHighlighting();
+      }, d.onended = () => {
+        this.stopWordHighlighting(), URL.revokeObjectURL(r), this.currentAudio = null, this.wordTimings = [], n();
+      }, d.onerror = (u) => {
+        this.stopWordHighlighting(), URL.revokeObjectURL(r), this.currentAudio = null, this.wordTimings = [], this.intentionallyStopped ? n() : l(new Error("Failed to play audio from server"));
+      }, d.onpause = () => {
+        this.stopWordHighlighting(), this.pausedState = !0;
+      }, d.play().catch(l);
+    });
+  }
+  /**
+   * Call server API to synthesize speech
+   */
+  async synthesizeSpeech(t, e, o) {
+    const r = Date.now();
+    await this.emitTelemetry("pie-tool-backend-call-start", {
+      toolId: "tts",
+      backend: this.config.provider || "server",
+      operation: "synthesize-speech"
+    });
+    const s = {
+      "Content-Type": "application/json",
+      ...this.config.headers
+    };
+    this.config.authToken && (s.Authorization = `Bearer ${this.config.authToken}`);
+    const a = this.adapter.resolveSynthesisUrl(this.config), n = this.adapter.buildRequestBody(t, this.config), l = await (async () => {
+      try {
+        return await fetch(a, {
+          method: "POST",
+          headers: s,
+          body: JSON.stringify(n),
+          signal: e
+        });
+      } catch (c) {
+        throw await this.emitTelemetry("pie-tool-backend-call-error", {
+          toolId: "tts",
+          backend: this.config.provider || "server",
+          operation: "synthesize-speech",
+          duration: Date.now() - r,
+          errorType: "TTSBackendNetworkError",
+          message: c instanceof Error ? c.message : String(c)
+        }), c;
+      }
+    })();
+    if (!l.ok) {
+      const c = await l.json().catch(() => ({})), p = c.message || c.error?.message || `Server returned ${l.status}`;
+      throw await this.emitTelemetry("pie-tool-backend-call-error", {
+        toolId: "tts",
+        backend: this.config.provider || "server",
+        operation: "synthesize-speech",
+        duration: Date.now() - r,
+        statusCode: l.status,
+        errorType: "TTSBackendRequestError",
+        message: p
+      }), new Error(p);
+    }
+    const d = await this.adapter.parseResponse(l, this.config, s, e);
+    if (o !== this.synthesisRunId || e.aborted)
+      throw new Error("Synthesis superseded by a newer request");
+    let u;
+    if (d.audio.kind === "base64")
+      u = this.base64ToBlob(d.audio.data, d.audio.contentType);
+    else {
+      const c = d.audio.url, p = Date.now();
+      await this.emitTelemetry("pie-tool-backend-call-start", {
+        toolId: "tts",
+        backend: this.config.provider || "server",
+        operation: "fetch-synthesized-audio-asset"
+      });
+      const w = {};
+      this.config.includeAuthOnAssetFetch && this.config.authToken && (w.Authorization = `Bearer ${this.config.authToken}`);
+      const g = await (async () => {
+        try {
+          return await fetch(c, {
+            headers: w,
+            signal: e
+          });
+        } catch (v) {
+          throw await this.emitTelemetry("pie-tool-backend-call-error", {
+            toolId: "tts",
+            backend: this.config.provider || "server",
+            operation: "fetch-synthesized-audio-asset",
+            duration: Date.now() - p,
+            errorType: "TTSAssetNetworkError",
+            message: v instanceof Error ? v.message : String(v)
+          }), v;
+        }
+      })();
+      if (!g.ok)
+        throw await this.emitTelemetry("pie-tool-backend-call-error", {
+          toolId: "tts",
+          backend: this.config.provider || "server",
+          operation: "fetch-synthesized-audio-asset",
+          duration: Date.now() - p,
+          statusCode: g.status,
+          errorType: "TTSAssetFetchError",
+          message: `Failed to download synthesized audio (${g.status})`
+        }), new Error(`Failed to download synthesized audio (${g.status})`);
+      u = await g.blob(), await this.emitTelemetry("pie-tool-backend-call-success", {
+        toolId: "tts",
+        backend: this.config.provider || "server",
+        operation: "fetch-synthesized-audio-asset",
+        duration: Date.now() - p
+      });
+    }
+    const y = URL.createObjectURL(u), m = this.parseSpeechMarks(d.speechMarks);
+    return await this.emitTelemetry("pie-tool-backend-call-success", {
+      toolId: "tts",
+      backend: this.config.provider || "server",
+      operation: "synthesize-speech",
+      duration: Date.now() - r
+    }), { audioUrl: y, wordTimings: m };
+  }
+  /**
+   * Convert base64 to Blob
+   */
+  base64ToBlob(t, e) {
+    const o = atob(t), r = new Array(o.length);
+    for (let a = 0; a < o.length; a++)
+      r[a] = o.charCodeAt(a);
+    const s = new Uint8Array(r);
+    return new Blob([s], { type: e });
+  }
+  /**
+   * Parse speech marks into word timings
+   */
+  parseSpeechMarks(t) {
+    return t.filter((e) => e.type === "word").map((e, o) => ({
+      time: e.time,
+      wordIndex: o,
+      charIndex: e.start,
+      length: e.end - e.start
+    }));
+  }
+  /**
+   * Start word highlighting synchronized with audio playback
+   */
+  startWordHighlighting() {
+    if (this.stopWordHighlighting(), !this.currentAudio || !this.onWordBoundary || this.wordTimings.length === 0) {
+      console.log("[ServerTTSProvider] Cannot start highlighting:", {
+        hasAudio: !!this.currentAudio,
+        hasCallback: !!this.onWordBoundary,
+        wordTimingsCount: this.wordTimings.length
+      });
+      return;
+    }
+    console.log("[ServerTTSProvider] Starting word highlighting with", this.wordTimings.length, "word timings"), console.log("[ServerTTSProvider] Playback rate:", this.currentAudio.playbackRate), console.log("[ServerTTSProvider] First 3 timings:", this.wordTimings.slice(0, 3));
+    let t = -1;
+    this.highlightInterval = window.setInterval(() => {
+      if (!this.currentAudio) {
+        this.stopWordHighlighting();
+        return;
+      }
+      const e = this.currentAudio.currentTime * 1e3;
+      for (let o = 0; o < this.wordTimings.length; o++) {
+        const r = this.wordTimings[o];
+        if (e >= r.time && o > t) {
+          this.onWordBoundary && (console.log("[ServerTTSProvider] Highlighting word at charIndex:", r.charIndex, "length:", r.length, "time:", r.time, "currentTime:", e), this.onWordBoundary("", r.charIndex, r.length)), t = o;
+          break;
+        }
+      }
+    }, 50);
+  }
+  /**
+   * Stop word highlighting
+   */
+  stopWordHighlighting() {
+    this.highlightInterval !== null && (clearInterval(this.highlightInterval), this.highlightInterval = null);
+  }
+  pause() {
+    this.currentAudio && !this.pausedState && (this.currentAudio.pause(), this.stopWordHighlighting(), this.pausedState = !0);
+  }
+  resume() {
+    this.currentAudio && this.pausedState && (this.currentAudio.play(), this.pausedState = !1, this.onWordBoundary && this.wordTimings.length > 0 && this.startWordHighlighting());
+  }
+  stop() {
+    this.synthesisRunId += 1, this.activeSynthesisController && (this.activeSynthesisController.abort(), this.activeSynthesisController = null), this.stopWordHighlighting(), this.currentAudio && (this.intentionallyStopped = !0, this.currentAudio.pause(), this.currentAudio.src && URL.revokeObjectURL(this.currentAudio.src), this.currentAudio.src = "", this.currentAudio = null), this.pausedState = !1, this.wordTimings = [];
+  }
+  isPlaying() {
+    return this.currentAudio !== null && !this.pausedState;
+  }
+  isPaused() {
+    return this.pausedState;
+  }
+  /**
+   * Update settings dynamically (rate, pitch, voice)
+   * Note: Voice changes require resynthesis, so voice updates are stored but
+   * take effect on the next speak() call. Rate can be applied to current playback.
+   */
+  updateSettings(t) {
+    t.rate !== void 0 && (this.config.rate = t.rate, this.currentAudio && (this.currentAudio.playbackRate = Math.max(0.25, Math.min(4, t.rate)))), t.pitch !== void 0 && (this.config.pitch = t.pitch), t.voice !== void 0 && (this.config.voice = t.voice);
+  }
+}
+class W {
+  constructor() {
+    h(this, "providerId", "server-tts"), h(this, "providerName", "Server TTS"), h(this, "version", "1.0.0"), h(this, "config", null), h(this, "adapter", null), h(this, "telemetryReporter");
+  }
+  async emitTelemetry(t, e) {
+    try {
+      await this.telemetryReporter?.(t, e);
+    } catch (o) {
+      console.warn("[ServerTTSProvider] telemetry callback failed:", o);
+    }
+  }
+  /**
+   * Initialize the server TTS provider.
+   *
+   * This is designed to be fast by default (no API calls).
+   * Set validateEndpoint: true in config to test API availability during initialization.
+   *
+   * @performance Default: <10ms, With validation: 100-500ms
+   */
+  async initialize(t) {
+    const e = t;
+    if (!e.apiEndpoint)
+      throw new Error("apiEndpoint is required for ServerTTSProvider");
+    this.config = e, this.telemetryReporter = b(e);
+    const o = T(e);
+    if (this.adapter = B[o], e.validateEndpoint) {
+      const r = Date.now();
+      if (await this.emitTelemetry("pie-tool-backend-call-start", {
+        toolId: "tts",
+        backend: e.provider || "server",
+        operation: "validate-endpoint"
+      }), !await this.testAPIAvailability())
+        throw await this.emitTelemetry("pie-tool-backend-call-error", {
+          toolId: "tts",
+          backend: e.provider || "server",
+          operation: "validate-endpoint",
+          duration: Date.now() - r,
+          errorType: "TTSEndpointValidationError",
+          message: `Server TTS API not available at ${e.apiEndpoint}`
+        }), new Error(`Server TTS API not available at ${e.apiEndpoint}`);
+      await this.emitTelemetry("pie-tool-backend-call-success", {
+        toolId: "tts",
+        backend: e.provider || "server",
+        operation: "validate-endpoint",
+        duration: Date.now() - r
+      });
+    }
+    return new O(e, this.adapter);
+  }
+  /**
+   * Test if API endpoint is available (with timeout).
+   *
+   * @performance 100-500ms depending on network
+   */
+  async testAPIAvailability() {
+    if (!this.config || !this.adapter)
+      return !1;
+    try {
+      const t = { ...this.config.headers };
+      this.config.authToken && (t.Authorization = `Bearer ${this.config.authToken}`);
+      const e = new AbortController(), o = setTimeout(() => e.abort(), 5e3), r = P(this.config, this.adapter.id);
+      if (r === "none")
+        return clearTimeout(o), !0;
+      const s = r === "voices" ? I(this.config) : this.adapter.resolveSynthesisUrl(this.config), a = r === "voices" ? "GET" : "OPTIONS";
+      try {
+        const n = await fetch(s, {
+          method: a,
+          headers: t,
+          signal: e.signal
+        });
+        return clearTimeout(o), n.ok || n.status === 405;
+      } catch {
+        return clearTimeout(o), !1;
+      }
+    } catch {
+      return !1;
+    }
+  }
+  supportsFeature(t) {
+    switch (t) {
+      case "pause":
+      case "resume":
+      case "wordBoundary":
+      case "voiceSelection":
+      case "rateControl":
+        return !0;
+      case "pitchControl":
+        return !1;
+      default:
+        return !1;
+    }
+  }
+  getCapabilities() {
+    const t = this.config ? T(this.config) : "pie";
+    return {
+      supportsPause: !0,
+      supportsResume: !0,
+      supportsWordBoundary: !0,
+      // ✅ Via speech marks from server
+      supportsVoiceSelection: !0,
+      supportsRateControl: !0,
+      supportsPitchControl: !1,
+      // Depends on server provider
+      maxTextLength: R[t]
+    };
+  }
+  destroy() {
+    this.config = null, this.adapter = null, this.telemetryReporter = void 0;
+  }
+}
+export {
+  W as ServerTTSProvider
+};