@drawdream/livespeech 0.1.0 → 0.1.2

package/dist/index.mjs

@@ -7,7 +7,7 @@ var Region = {
 };
 var REGION_ENDPOINTS = {
   "ap-northeast-2": "wss://talk.drawdream.co.kr",
-  "us-west-2": "wss://talk..drawdream.ca"
+  "us-west-2": "wss://talk.drawdream.ca"
   // Coming soon
 };
 function getEndpointForRegion(region) {
@@ -293,22 +293,13 @@ var WebSocketConnection = class {
   /**
    * Handle incoming message
    */
-  handleMessage(data, onFirstConnect) {
+  handleMessage(data) {
     const message = parseServerMessage(data);
     if (!message) {
       this.logger.warn("Invalid message received:", data);
       return;
     }
     this.logger.debug("Received message:", message.type);
-    if (message.type === "connected") {
-      this.connectionId = message.connectionId;
-      this.state = "connected";
-      this.retryController.reset();
-      this.startPingInterval();
-      this.events.onOpen?.(message.connectionId);
-      onFirstConnect?.();
-      return;
-    }
     if (message.type === "pong") {
       this.logger.debug("Pong received");
       return;
@@ -574,24 +565,17 @@ var CONFIG_DEFAULTS = {
   reconnectDelay: 1e3,
   debug: false
 };
-var SESSION_DEFAULTS = {
-  voiceId: "en-US-Standard-A",
-  languageCode: "en-US",
-  inputFormat: "pcm16",
-  outputFormat: "pcm16",
-  sampleRate: 16e3
-};
 var LiveSpeechClient = class {
   config;
   connection;
   audioEncoder;
   logger;
   sessionId = null;
-  sessionConfig = null;
+  isStreaming = false;
   // Event listeners using a simple map
   eventListeners = /* @__PURE__ */ new Map();
   // Simplified handlers
-  transcriptHandler = null;
+  userTranscriptHandler = null;
   responseHandler = null;
   audioHandler = null;
   errorHandler = null;
@@ -653,6 +637,12 @@ var LiveSpeechClient = class {
   get hasActiveSession() {
     return this.sessionId !== null;
   }
+  /**
+   * Check if audio streaming is active
+   */
+  get isAudioStreaming() {
+    return this.isStreaming;
+  }
   /**
    * Connect to the server
    */
@@ -666,7 +656,7 @@ var LiveSpeechClient = class {
   disconnect() {
     this.logger.info("Disconnecting...");
     this.sessionId = null;
-    this.sessionConfig = null;
+    this.isStreaming = false;
     this.connection.disconnect();
   }
   /**
@@ -679,16 +669,6 @@ var LiveSpeechClient = class {
     if (this.sessionId) {
       throw new Error("Session already active. Call endSession() first.");
     }
-    const resolvedConfig = {
-      prePrompt: config.prePrompt,
-      voiceId: config.voiceId ?? SESSION_DEFAULTS.voiceId,
-      languageCode: config.languageCode ?? SESSION_DEFAULTS.languageCode,
-      inputFormat: config.inputFormat ?? SESSION_DEFAULTS.inputFormat,
-      outputFormat: config.outputFormat ?? SESSION_DEFAULTS.outputFormat,
-      sampleRate: config.sampleRate ?? SESSION_DEFAULTS.sampleRate,
-      metadata: config.metadata ?? {}
-    };
-    this.sessionConfig = resolvedConfig;
     this.logger.info("Starting session...");
     return new Promise((resolve, reject) => {
       const onSessionStarted = (event) => {
@@ -705,16 +685,17 @@ var LiveSpeechClient = class {
       };
       this.on("sessionStarted", onSessionStarted);
       this.on("error", onError);
-      this.connection.send({
-        action: "startSession",
-        prePrompt: resolvedConfig.prePrompt,
-        voiceId: resolvedConfig.voiceId,
-        languageCode: resolvedConfig.languageCode,
-        inputFormat: resolvedConfig.inputFormat,
-        outputFormat: resolvedConfig.outputFormat,
-        sampleRate: resolvedConfig.sampleRate,
-        metadata: resolvedConfig.metadata
-      });
+      const startMessage = {
+        action: "startSession"
+      };
+      if (config?.prePrompt) {
+        startMessage.prePrompt = config.prePrompt;
+      }
+      if (config?.language) {
+        startMessage.language = config.language;
+      }
+      startMessage.pipelineMode = config?.pipelineMode ?? "live";
+      this.connection.send(startMessage);
     });
   }
   /**
@@ -726,6 +707,9 @@ var LiveSpeechClient = class {
       return;
     }
     this.logger.info("Ending session...");
+    if (this.isStreaming) {
+      this.audioEnd();
+    }
     return new Promise((resolve) => {
       const onSessionEnded = () => {
         this.off("sessionEnded", onSessionEnded);
@@ -736,28 +720,49 @@ var LiveSpeechClient = class {
     });
   }
   /**
-   * Send audio data
+   * Start audio streaming session
    */
-  sendAudio(data, options) {
+  audioStart() {
     if (!this.isConnected) {
       throw new Error("Not connected");
     }
     if (!this.sessionId) {
       throw new Error("No active session. Call startSession() first.");
     }
+    if (this.isStreaming) {
+      throw new Error("Already streaming. Call audioEnd() first.");
+    }
+    this.logger.info("Starting audio stream...");
+    this.connection.send({ action: "audioStart" });
+    this.isStreaming = true;
+  }
+  /**
+   * Send audio chunk (PCM16 base64 encoded)
+   */
+  sendAudioChunk(data) {
+    if (!this.isConnected) {
+      throw new Error("Not connected");
+    }
+    if (!this.isStreaming) {
+      throw new Error("Not streaming. Call audioStart() first.");
+    }
     const base64Data = this.audioEncoder.encode(data);
-    const format = options?.format ?? this.sessionConfig?.inputFormat ?? SESSION_DEFAULTS.inputFormat;
-    const sampleRate = this.sessionConfig?.sampleRate ?? SESSION_DEFAULTS.sampleRate;
-    const audioMessage = {
-      action: "audio",
-      data: base64Data,
-      format,
-      sampleRate
-    };
-    if (options?.isFinal !== void 0) {
-      audioMessage.isFinal = options.isFinal;
+    this.connection.send({
+      action: "audioChunk",
+      data: base64Data
+    });
+  }
+  /**
+   * End audio streaming session
+   */
+  audioEnd() {
+    if (!this.isStreaming) {
+      this.logger.warn("Not streaming");
+      return;
     }
-    this.connection.send(audioMessage);
+    this.logger.info("Ending audio stream...");
+    this.connection.send({ action: "audioEnd" });
+    this.isStreaming = false;
   }
   // ==================== Event System ====================
   /**
@@ -779,13 +784,7 @@ var LiveSpeechClient = class {
     }
   }
   /**
-   * Set transcript handler (simplified)
-   */
-  setTranscriptHandler(handler) {
-    this.transcriptHandler = handler;
-  }
-  /**
-   * Set response handler (simplified)
+   * Set response handler
    */
   setResponseHandler(handler) {
     this.responseHandler = handler;
@@ -796,6 +795,12 @@ var LiveSpeechClient = class {
   setAudioHandler(handler) {
     this.audioHandler = handler;
   }
+  /**
+   * Set user transcript handler
+   */
+  setUserTranscriptHandler(handler) {
+    this.userTranscriptHandler = handler;
+  }
   /**
    * Set error handler (simplified)
    */
@@ -825,7 +830,7 @@ var LiveSpeechClient = class {
   }
   handleDisconnected(code, _reason) {
     this.sessionId = null;
-    this.sessionConfig = null;
+    this.isStreaming = false;
     const event = {
       type: "disconnected",
       reason: code === 1e3 ? "normal" : "error",
@@ -867,25 +872,19 @@ var LiveSpeechClient = class {
         break;
       case "sessionEnded":
         this.sessionId = null;
-        this.sessionConfig = null;
+        this.isStreaming = false;
         this.emit("sessionEnded", {
           type: "sessionEnded",
           sessionId: message.sessionId,
           timestamp: message.timestamp
         });
         break;
-      case "transcript": {
-        const transcriptEvent = {
-          type: "transcript",
-          text: message.text,
-          isFinal: message.isFinal,
+      case "ready": {
+        const readyEvent = {
+          type: "ready",
           timestamp: message.timestamp
         };
-        if (message.confidence !== void 0) {
-          transcriptEvent.confidence = message.confidence;
-        }
-        this.emit("transcript", transcriptEvent);
-        this.transcriptHandler?.(message.text, message.isFinal);
+        this.emit("ready", readyEvent);
         break;
       }
       case "response": {
@@ -912,8 +911,26 @@ var LiveSpeechClient = class {
         this.audioHandler?.(audioData);
         break;
       }
+      case "userTranscript": {
+        const userTranscriptEvent = {
+          type: "userTranscript",
+          text: message.text,
+          timestamp: message.timestamp
+        };
+        this.emit("userTranscript", userTranscriptEvent);
+        this.userTranscriptHandler?.(message.text);
+        break;
+      }
+      case "turnComplete": {
+        const turnCompleteEvent = {
+          type: "turnComplete",
+          timestamp: message.timestamp
+        };
+        this.emit("turnComplete", turnCompleteEvent);
+        break;
+      }
       case "error":
-        this.handleError(message.code, message.message, message.details);
+        this.handleError(message.code, message.message);
         break;
       default:
         this.logger.warn("Unknown message type:", message.type);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@drawdream/livespeech",
-  "version": "0.1.0",
+  "version": "0.1.2",
   "description": "Real-time speech-to-speech AI conversation SDK",
   "main": "dist/index.js",
   "module": "dist/index.mjs",