@lokutor/sdk 1.1.15 → 1.1.19

package/dist/index.js

@@ -24,6 +24,7 @@ __export(index_exports, {
   BrowserAudioManager: () => BrowserAudioManager,
   DEFAULT_URLS: () => DEFAULT_URLS,
   Language: () => Language,
+  LokutorError: () => LokutorError,
   StreamResampler: () => StreamResampler,
   TTSClient: () => TTSClient,
   VoiceAgentClient: () => VoiceAgentClient,
@@ -32,6 +33,7 @@ __export(index_exports, {
   bytesToPcm16: () => bytesToPcm16,
   calculateRMS: () => calculateRMS,
   float32ToPcm16: () => float32ToPcm16,
+  isRetryable: () => isRetryable,
   normalizeAudio: () => normalizeAudio,
   pcm16ToBytes: () => pcm16ToBytes,
   pcm16ToFloat32: () => pcm16ToFloat32,
@@ -66,7 +68,9 @@ var Language = /* @__PURE__ */ ((Language2) => {
 })(Language || {});
 var AUDIO_CONFIG = {
   SAMPLE_RATE: 16e3,
+  SAMPLE_RATE_INPUT: 16e3,
   SPEAKER_SAMPLE_RATE: 44100,
+  SAMPLE_RATE_OUTPUT: 44100,
   CHANNELS: 1,
   CHUNK_DURATION_MS: 20,
   get CHUNK_SIZE() {
@@ -77,6 +81,41 @@ var DEFAULT_URLS = {
   VOICE_AGENT: "wss://api.lokutor.com/ws/agent",
   TTS: "wss://api.lokutor.com/ws/tts"
 };
+var LokutorError = class extends Error {
+  code;
+  detail;
+  retryable;
+  original;
+  constructor(code, message, opts) {
+    super(message);
+    this.name = "LokutorError";
+    this.code = code;
+    this.detail = opts?.detail;
+    this.retryable = opts?.retryable ?? isRetryableCode(code);
+    this.original = opts?.original;
+  }
+};
+function isRetryableCode(code) {
+  const fatal = [
+    "auth.missing_key",
+    "auth.invalid_key",
+    "auth.time_limited",
+    "validation.invalid_voice",
+    "validation.invalid_language",
+    "validation.text_too_long",
+    "validation.speed_out_of_range",
+    "validation.steps_out_of_range",
+    "validation.invalid_request_format",
+    "internal.cancelled"
+  ];
+  return !fatal.includes(code);
+}
+function isRetryable(error) {
+  if (error instanceof LokutorError) {
+    return error.retryable;
+  }
+  return false;
+}
 
 // src/audio-utils.ts
 function pcm16ToFloat32(int16Data) {
@@ -137,6 +176,9 @@ function pcm16ToBytes(data) {
   return new Uint8Array(data.buffer, data.byteOffset, data.byteLength);
 }
 function bytesToPcm16(bytes) {
+  if (bytes.length % 2 !== 0) {
+    bytes = bytes.slice(0, bytes.length - 1);
+  }
   return new Int16Array(bytes.buffer, bytes.byteOffset, bytes.length / 2);
 }
 function normalizeAudio(data, targetPeak = 0.95) {
@@ -186,11 +228,14 @@ var StreamResampler = class {
     combined.set(this.inputBuffer);
     combined.set(inputChunk, this.inputBuffer.length);
     const ratio = this.inputRate / this.outputRate;
-    const outputLength = Math.floor(combined.length / ratio);
+    let outputLength = Math.floor(combined.length / ratio);
     if (outputLength === 0 && !flush) {
       this.inputBuffer = combined;
       return new Float32Array(0);
     }
+    if (flush && outputLength === 0 && combined.length > 0) {
+      outputLength = 1;
+    }
     const output = new Float32Array(outputLength);
     for (let i = 0; i < outputLength; i++) {
       const pos = i * ratio;
@@ -199,10 +244,8 @@ var StreamResampler = class {
       const weight = pos - left;
       output[i] = combined[left] * (1 - weight) + combined[right] * weight;
     }
-    const remainingSamples = Math.ceil(combined.length - outputLength * ratio);
-    this.inputBuffer = combined.slice(
-      combined.length - remainingSamples
-    );
+    const consumed = Math.floor(outputLength * ratio);
+    this.inputBuffer = combined.slice(consumed);
     return output;
   }
   reset() {
@@ -364,6 +407,10 @@ var BrowserAudioManager = class {
       console.warn("AudioContext not initialized");
       return;
     }
+    if (pcm16Data.length % 2 !== 0) {
+      console.warn(`Discarding odd-length PCM buffer (${pcm16Data.length} bytes)`);
+      return;
+    }
     const int16Array = new Int16Array(
       pcm16Data.buffer,
       pcm16Data.byteOffset,
@@ -492,6 +539,54 @@ var BrowserAudioManager = class {
 };
 
 // src/client.ts
+function sdkTraceEnabled() {
+  try {
+    if (typeof window === "undefined") return false;
+    const w = window;
+    return Boolean(w.LOKUTOR_TRACE) || window.localStorage?.getItem("lokutorTrace") === "1";
+  } catch {
+    return false;
+  }
+}
+function sdkTrace(...args) {
+  if (sdkTraceEnabled()) {
+    console.log("[SDK TRACE]", ...args);
+  }
+}
+function nowMs() {
+  if (typeof performance !== "undefined" && performance.now) {
+    return performance.now();
+  }
+  return Date.now();
+}
+function wsToHttp(url) {
+  return url.replace(/^wss:/, "https:").replace(/^ws:/, "http:");
+}
+async function fetchJson(url, timeoutMs = 1e4) {
+  const controller = new AbortController();
+  const timer = setTimeout(() => controller.abort(), timeoutMs);
+  try {
+    const res = await fetch(url, {
+      signal: controller.signal,
+      headers: { Accept: "application/json" }
+    });
+    clearTimeout(timer);
+    if (!res.ok) {
+      throw new LokutorError("internal.error", `HTTP ${res.status} from ${url}`, {
+        detail: await res.text().catch(() => ""),
+        retryable: res.status >= 500
+      });
+    }
+    return await res.json();
+  } catch (err) {
+    clearTimeout(timer);
+    if (err instanceof LokutorError) throw err;
+    throw new LokutorError("internal.error", `Failed to fetch ${url}`, {
+      original: err,
+      retryable: true
+    });
+  }
+}
 function base64ToUint8Array(base64) {
   const binaryString = atob(base64);
   const bytes = new Uint8Array(binaryString.length);
@@ -500,6 +595,39 @@ function base64ToUint8Array(base64) {
   }
   return bytes;
 }
+function normalizeVisemes(payload) {
+  if (!Array.isArray(payload)) return [];
+  const normalized = [];
+  for (const item of payload) {
+    if (!item || typeof item !== "object") continue;
+    const c = String(item.c ?? item.char ?? "sil").toLowerCase();
+    const t = Number(item.t ?? item.timestamp ?? 0);
+    const v = Number(item.v ?? item.id ?? 0);
+    normalized.push({
+      v: Number.isFinite(v) ? v : 0,
+      c,
+      t: Number.isFinite(t) ? t : 0
+    });
+  }
+  return normalized;
+}
+function extractVisemePayload(msg) {
+  if (Array.isArray(msg?.data)) {
+    return normalizeVisemes(msg.data);
+  }
+  if (Array.isArray(msg?.data?.visemes)) {
+    return normalizeVisemes(msg.data.visemes);
+  }
+  if (msg?.data && !Array.isArray(msg.data) && typeof msg.data === "object") {
+    const singularInData = normalizeVisemes([msg.data]);
+    if (singularInData.length > 0) return singularInData;
+  }
+  if (msg && !Array.isArray(msg) && typeof msg === "object") {
+    const singularAtRoot = normalizeVisemes([msg]);
+    if (singularAtRoot.length > 0) return singularAtRoot;
+  }
+  return [];
+}
 var VoiceAgentClient = class {
   ws = null;
   apiKey;
@@ -521,16 +649,19 @@ var VoiceAgentClient = class {
   audioManager = null;
   enableAudio = false;
   currentGeneration = 0;
+  listeners = {};
   // Connection resilience
   isUserDisconnect = false;
   reconnecting = false;
   reconnectAttempts = 0;
   maxReconnectAttempts = 5;
+  serverUrl;
   constructor(config) {
     this.apiKey = config.apiKey;
     this.prompt = config.prompt;
     this.voice = config.voice || "F1" /* F1 */;
     this.language = config.language || "en" /* ENGLISH */;
+    this.serverUrl = config.serverUrl || DEFAULT_URLS.VOICE_AGENT;
     this.onTranscription = config.onTranscription;
     this.onResponse = config.onResponse;
     this.onAudioCallback = config.onAudio;
@@ -543,23 +674,43 @@ var VoiceAgentClient = class {
   }
   /**
    * Connect to the Lokutor Voice Agent server
+   * @param customAudioManager Optional replacement for the default audio hardware handler
    */
-  async connect() {
+  async connect(customAudioManager) {
     this.isUserDisconnect = false;
-    if (this.enableAudio) {
-      if (!this.audioManager) {
+    if (this.enableAudio || customAudioManager) {
+      if (customAudioManager) {
+        this.audioManager = customAudioManager;
+      } else if (!this.audioManager && typeof window !== "undefined") {
         this.audioManager = new BrowserAudioManager();
       }
-      await this.audioManager.init();
+      if (this.audioManager) {
+        await this.audioManager.init();
+      }
     }
     return new Promise((resolve, reject) => {
+      let settled = false;
+      const settle = (fn) => {
+        if (!settled) {
+          settled = true;
+          fn();
+        }
+      };
       try {
-        let url = DEFAULT_URLS.VOICE_AGENT;
+        let url = this.serverUrl;
         if (this.apiKey) {
           const separator = url.includes("?") ? "&" : "?";
           url += `${separator}api_key=${this.apiKey}`;
         }
-        console.log(`\u{1F517} Connecting to ${DEFAULT_URLS.VOICE_AGENT}...`);
+        const redactedUrl = url.replace(/api_key=[^&]+/, "api_key=***");
+        sdkTrace("ws.connect", {
+          endpoint: this.serverUrl,
+          url: redactedUrl,
+          enableAudio: this.enableAudio,
+          wantVisemes: this.wantVisemes,
+          hasAudioManager: Boolean(this.audioManager)
+        });
+        console.log(`\u{1F517} Connecting to ${this.serverUrl}...`);
         this.ws = new WebSocket(url);
         this.ws.binaryType = "arraybuffer";
         this.ws.onopen = async () => {
@@ -567,6 +718,7 @@ var VoiceAgentClient = class {
           this.reconnectAttempts = 0;
           this.reconnecting = false;
           console.log("\u2705 Connected to voice agent!");
+          sdkTrace("ws.open");
           this.sendConfig();
           if (this.audioManager) {
             await this.audioManager.startMicrophone((data) => {
@@ -575,22 +727,54 @@ var VoiceAgentClient = class {
               }
             });
           }
-          resolve(true);
+          settle(() => resolve(true));
         };
         this.ws.onmessage = async (event) => {
           if (event.data instanceof ArrayBuffer) {
+            sdkTrace("ws.message.binary", { bytes: event.data.byteLength });
             this.handleBinaryMessage(new Uint8Array(event.data));
           } else {
+            sdkTrace("ws.message.text", { length: String(event.data).length });
             this.handleTextMessage(event.data.toString());
           }
         };
         this.ws.onerror = (err) => {
-          console.error("\u274C WebSocket error:", err);
-          if (this.onError) this.onError(err);
-          if (!this.isConnected) reject(err);
+          const error = new LokutorError("ws.close", "WebSocket connection error", {
+            detail: `readyState=${this.ws?.readyState}, bufferedAmount=${this.ws?.bufferedAmount}`,
+            original: err,
+            retryable: true
+          });
+          console.error("\u274C WebSocket error:", error.message);
+          sdkTrace("ws.error", { code: error.code, message: error.message });
+          if (this.onError) this.onError(error);
+          if (!this.isConnected) {
+            settle(() => reject(error));
+          }
         };
-        this.ws.onclose = () => {
+        this.ws.onclose = (event) => {
           this.isConnected = false;
+          const diagnostic = {
+            code: event.code,
+            reason: event.reason,
+            wasClean: event.wasClean,
+            url: this.serverUrl,
+            isUserDisconnect: this.isUserDisconnect,
+            reconnectAttempts: this.reconnectAttempts
+          };
+          sdkTrace("ws.close", diagnostic);
+          if (!settled && !this.isUserDisconnect) {
+            const error = new LokutorError("ws.close", `WebSocket closed unexpectedly (code ${event.code})`, {
+              detail: event.reason || "No reason provided",
+              retryable: event.code !== 1008
+            });
+            settle(() => reject(error));
+            return;
+          }
+          if (!event.wasClean && event.code === 1006 && this.reconnectAttempts === 0 && !this.isUserDisconnect) {
+            console.error("\u274C Connection rejected (code 1006). Likely causes: invalid API key, endpoint unavailable, or CORS blocked.");
+            console.error("   URL:", this.serverUrl.replace(/api_key=[^&]+/, "api_key=***"));
+          }
+          console.log(`\u{1F50C} WebSocket closed \u2014 code: ${event.code}, reason: "${event.reason || "none"}", clean: ${event.wasClean}`);
           if (!this.isUserDisconnect && this.reconnectAttempts < this.maxReconnectAttempts) {
             this.reconnecting = true;
             this.reconnectAttempts++;
@@ -606,20 +790,46 @@ var VoiceAgentClient = class {
           }
         };
       } catch (err) {
-        if (this.onError) this.onError(err);
-        reject(err);
+        const error = err instanceof LokutorError ? err : new LokutorError("internal.error", "Failed to create WebSocket connection", { original: err });
+        if (this.onError) this.onError(error);
+        settle(() => reject(error));
       }
     });
   }
+  /**
+   * The "Golden Path" - Starts a managed session with hardware handled automatically.
+   * This is the recommended way to start a conversation in browser environments.
+   */
+  async startManaged(config) {
+    this.enableAudio = true;
+    if (config?.audioManager) {
+      this.audioManager = config.audioManager;
+    } else if (!this.audioManager) {
+      if (typeof window === "undefined") {
+        throw new LokutorError("internal.error", "startManaged() requires a browser environment. Pass a custom audioManager for non-browser runtimes.", { retryable: false });
+      }
+      this.audioManager = new BrowserAudioManager();
+    }
+    await this.connect();
+    return this;
+  }
   /**
    * Send initial configuration to the server
    */
   sendConfig() {
     if (!this.ws || !this.isConnected) return;
-    this.ws.send(JSON.stringify({ type: "prompt", data: this.prompt }));
+    this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
     this.ws.send(JSON.stringify({ type: "voice", data: this.voice }));
     this.ws.send(JSON.stringify({ type: "language", data: this.language }));
-    this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
+    this.ws.send(JSON.stringify({ type: "prompt", data: this.prompt }));
+    this.ws.send(JSON.stringify({ type: "rates", playback: 44100, input: 16e3 }));
+    sdkTrace("ws.send.config", {
+      promptLen: this.prompt?.length || 0,
+      voice: this.voice,
+      language: this.language,
+      visemes: this.wantVisemes,
+      tools: this.tools?.length || 0
+    });
     if (this.tools && this.tools.length > 0) {
       this.ws.send(JSON.stringify({ type: "tools", data: this.tools }));
     }
@@ -653,6 +863,15 @@ var VoiceAgentClient = class {
   handleTextMessage(text) {
     try {
       const msg = JSON.parse(text);
+      if (!msg || typeof msg !== "object") {
+        return;
+      }
+      sdkTrace("ws.recv.type", {
+        type: msg.type,
+        hasData: Object.prototype.hasOwnProperty.call(msg, "data"),
+        dataKind: Array.isArray(msg.data) ? "array" : typeof msg.data,
+        generation: msg.generation ?? null
+      });
       switch (msg.type) {
         case "audio":
           if (msg.data) {
@@ -665,7 +884,7 @@ var VoiceAgentClient = class {
           this.messages.push({
             role,
             text: msg.data,
-            timestamp: Date.now()
+            timestamp: nowMs()
           });
           if (msg.role === "user") {
             if (this.onTranscription) this.onTranscription(msg.data);
@@ -697,36 +916,91 @@ var VoiceAgentClient = class {
           console.log(`${icons[msg.data] || ""} Status: ${msg.data}`);
           break;
         case "visemes":
-          if (Array.isArray(msg.data) && msg.data.length > 0) {
-            this.emit("visemes", msg.data);
+        case "viseme": {
+          const msgGen = msg.generation ?? this.currentGeneration;
+          if (msgGen < this.currentGeneration) {
+            sdkTrace("visemes.discard", { msgGen, currentGen: this.currentGeneration });
+            break;
+          }
+          const normalized = extractVisemePayload(msg);
+          const explicitlyEmptyArray = Array.isArray(msg?.data) || Array.isArray(msg?.data?.visemes);
+          sdkTrace("visemes.recv", {
+            rawType: msg.type,
+            normalizedCount: normalized.length,
+            first: normalized[0] ?? null
+          });
+          if (normalized.length > 0 || explicitlyEmptyArray) {
+            this.emit("visemes", normalized);
           }
           break;
-        case "error":
-          if (this.onError) this.onError(msg.data);
-          console.error(`\u274C Server error: ${msg.data}`);
+        }
+        case "error": {
+          const backendCode = msg.data?.code ?? "internal.error";
+          const backendMessage = msg.data?.message ?? msg.data ?? "Unknown server error";
+          const backendDetail = msg.data?.detail;
+          const backendRetryable = msg.data?.retryable ?? true;
+          const error = new LokutorError(backendCode, backendMessage, {
+            detail: backendDetail,
+            retryable: backendRetryable
+          });
+          if (this.onError) this.onError(error);
+          console.error(`\u274C Server error: [${error.code}] ${error.message}`);
           break;
+        }
         case "tool_call":
           console.log(`\u{1F6E0}\uFE0F Tool Call: ${msg.name}(${msg.arguments})`);
           break;
       }
     } catch (e) {
+      sdkTrace("ws.recv.parse_error", { preview: text?.slice(0, 120) });
+      console.debug("Failed to parse message:", e);
     }
   }
-  audioListeners = [];
-  emit(event, data) {
-    if (event === "audio") {
-      if (this.onAudioCallback) this.onAudioCallback(data);
-      this.audioListeners.forEach((l) => l(data));
-    } else if (event === "visemes") {
-      if (this.onVisemesCallback) this.onVisemesCallback(data);
-      this.visemeListeners.forEach((l) => l(data));
+  /**
+   * Register an event listener (for Python parity)
+   */
+  on(event, callback) {
+    if (!this.listeners[event]) {
+      this.listeners[event] = [];
+    }
+    this.listeners[event].push(callback);
+    return this;
+  }
+  /**
+   * Internal emitter for all events
+   */
+  emit(event, ...args) {
+    const legacyMap = {
+      "transcription": "onTranscription",
+      "response": "onResponse",
+      "audio": "onAudioCallback",
+      "visemes": "onVisemesCallback",
+      "status": "onStatus",
+      "error": "onError"
+    };
+    const legacyKey = legacyMap[event];
+    if (legacyKey && this[legacyKey]) {
+      try {
+        this[legacyKey](...args);
+      } catch (e) {
+        console.error(`Error in legacy callback ${legacyKey}:`, e);
+      }
+    }
+    if (this.listeners[event]) {
+      this.listeners[event].forEach((cb) => {
+        try {
+          cb(...args);
+        } catch (e) {
+          console.error(`Error in listener for ${event}:`, e);
+        }
+      });
     }
   }
   onAudio(callback) {
-    this.audioListeners.push(callback);
+    this.on("audio", callback);
   }
   onVisemes(callback) {
-    this.visemeListeners.push(callback);
+    this.on("visemes", callback);
   }
   /**
    * Disconnect from the server
@@ -741,6 +1015,21 @@ var VoiceAgentClient = class {
       this.audioManager.cleanup();
     }
     this.isConnected = false;
+    this.reconnecting = false;
+    this.reconnectAttempts = 0;
+  }
+  /**
+   * Returns true if the client is currently connected.
+   */
+  get connected() {
+    return this.isConnected;
+  }
+  /**
+   * Returns the current generation counter.
+   * Useful for correlating audio/viseme chunks with utterances.
+   */
+  get generation() {
+    return this.currentGeneration;
   }
   /**
    * Toggles the microphone mute state (if managed by client)
@@ -763,17 +1052,70 @@ var VoiceAgentClient = class {
     }
     return 0;
   }
+  /**
+   * Fetch available voice styles from the server.
+   * No authentication required.
+   */
+  static async fetchVoices(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    const data = await fetchJson(`${url}/voices`);
+    return data.voices || [];
+  }
+  /**
+   * Fetch supported languages from the server.
+   * No authentication required.
+   */
+  static async fetchLanguages(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    const data = await fetchJson(`${url}/languages`);
+    return data.languages || [];
+  }
+  /**
+   * Fetch loaded TTS model versions from the server.
+   * No authentication required.
+   */
+  static async fetchModels(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    const data = await fetchJson(`${url}/models`);
+    return data.models || [];
+  }
+  /**
+   * Fetch server configuration (limits and defaults).
+   * No authentication required.
+   */
+  static async fetchConfig(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    return fetchJson(`${url}/config`);
+  }
+  /**
+   * Fetch rich runtime status from the server.
+   * No authentication required.
+   */
+  static async fetchStatus(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    return fetchJson(`${url}/status`);
+  }
+  /**
+   * Fetch health/liveness status from the server.
+   * No authentication required.
+   */
+  static async fetchHealth(baseUrl) {
+    const url = wsToHttp(baseUrl || DEFAULT_URLS.VOICE_AGENT);
+    return fetchJson(`${url}/health`);
+  }
   /**
    * Update the system prompt mid-conversation
    */
   updatePrompt(newPrompt) {
     this.prompt = newPrompt;
-    if (this.ws && this.isConnected) {
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
       try {
         this.ws.send(JSON.stringify({ type: "prompt", data: newPrompt }));
         console.log(`\u2699\uFE0F Updated prompt: ${newPrompt.substring(0, 50)}...`);
       } catch (error) {
-        console.error("Error updating prompt:", error);
+        const err = new LokutorError("internal.error", "Failed to update prompt", { original: error });
+        if (this.onError) this.onError(err);
+        console.error("Error updating prompt:", err.message);
       }
     } else {
       console.warn("Not connected - prompt will be updated on next connection");
@@ -805,15 +1147,28 @@ var TTSClient = class {
    */
   synthesize(options) {
     return new Promise((resolve, reject) => {
+      let activityTimeout;
+      let ws;
+      let startTime;
+      let firstByteReceived = false;
+      const refreshTimeout = () => {
+        if (activityTimeout) clearTimeout(activityTimeout);
+        activityTimeout = setTimeout(() => {
+          console.log("\u23F1\uFE0F TTS synthesis reached inactivity timeout (2s) - resolving");
+          if (ws) ws.close();
+          resolve();
+        }, 2e3);
+      };
       try {
         let url = DEFAULT_URLS.TTS;
         if (this.apiKey) {
           const separator = url.includes("?") ? "&" : "?";
           url += `${separator}api_key=${this.apiKey}`;
         }
-        const ws = new WebSocket(url);
+        ws = new WebSocket(url);
         ws.binaryType = "arraybuffer";
         ws.onopen = () => {
+          refreshTimeout();
           const req = {
             text: options.text,
             voice: options.voice || "F1" /* F1 */,
@@ -823,28 +1178,65 @@ var TTSClient = class {
             visemes: options.visemes || false
           };
           ws.send(JSON.stringify(req));
+          startTime = nowMs();
         };
         ws.onmessage = async (event) => {
+          refreshTimeout();
           if (event.data instanceof ArrayBuffer) {
+            if (!firstByteReceived) {
+              const ttfb = nowMs() - startTime;
+              if (options.onTTFB) options.onTTFB(ttfb);
+              firstByteReceived = true;
+            }
             if (options.onAudio) options.onAudio(new Uint8Array(event.data));
-          } else {
-            try {
-              const msg = JSON.parse(event.data.toString());
-              if (Array.isArray(msg) && options.onVisemes) {
-                options.onVisemes(msg);
+            return;
+          }
+          const text = event.data.toString();
+          if (text === "EOS") {
+            if (activityTimeout) clearTimeout(activityTimeout);
+            ws.close();
+            resolve();
+            return;
+          }
+          try {
+            const msg = JSON.parse(text);
+            if (msg.type === "audio" && msg.data) {
+              const audioBuffer = base64ToUint8Array(msg.data);
+              if (!firstByteReceived) {
+                const ttfb = nowMs() - startTime;
+                if (options.onTTFB) options.onTTFB(ttfb);
+                firstByteReceived = true;
               }
-            } catch (e) {
+              if (options.onAudio) options.onAudio(audioBuffer);
+              return;
+            }
+            if (msg.type === "visemes" && Array.isArray(msg.data) && options.onVisemes) {
+              options.onVisemes(normalizeVisemes(msg.data));
+              return;
+            }
+            if (Array.isArray(msg) && options.onVisemes) {
+              options.onVisemes(normalizeVisemes(msg));
+              return;
+            }
+            if (msg.type === "eos") {
+              if (activityTimeout) clearTimeout(activityTimeout);
+              ws.close();
+              resolve();
             }
+          } catch (e) {
           }
         };
         ws.onerror = (err) => {
+          if (activityTimeout) clearTimeout(activityTimeout);
           if (options.onError) options.onError(err);
           reject(err);
         };
         ws.onclose = () => {
+          if (activityTimeout) clearTimeout(activityTimeout);
           resolve();
         };
       } catch (err) {
+        if (activityTimeout) clearTimeout(activityTimeout);
         if (options.onError) options.onError(err);
         reject(err);
       }
@@ -866,6 +1258,7 @@ async function simpleTTS(options) {
   BrowserAudioManager,
   DEFAULT_URLS,
   Language,
+  LokutorError,
   StreamResampler,
   TTSClient,
   VoiceAgentClient,
@@ -874,6 +1267,7 @@ async function simpleTTS(options) {
   bytesToPcm16,
   calculateRMS,
   float32ToPcm16,
+  isRetryable,
   normalizeAudio,
   pcm16ToBytes,
   pcm16ToFloat32,