@lokutor/sdk 1.1.35 → 1.1.37

package/dist/index.d.mts

@@ -236,10 +236,14 @@ declare class VoiceAgentClient {
     private enableAudio;
     private currentGeneration;
     private listeners;
+    private configConfirmed;
     private isUserDisconnect;
     private reconnecting;
     private reconnectAttempts;
     private maxReconnectAttempts;
+    private reconnectTimer;
+    private serverPlaybackRate;
+    private serverInputRate;
     private serverUrl;
     constructor(config: LokutorConfig & {
         prompt: string;
@@ -346,6 +350,18 @@ declare class VoiceAgentClient {
      * Update the system prompt mid-conversation
      */
     updatePrompt(newPrompt: string): void;
+    /**
+     * Change the voice style mid-conversation
+     */
+    updateVoice(voice: VoiceStyle): void;
+    /**
+     * Change the language mid-conversation
+     */
+    updateLanguage(language: Language): void;
+    /**
+     * Update tool definitions mid-conversation
+     */
+    updateTools(tools: ToolDefinition[]): void;
     /**
      * Get full conversation transcript
      */

package/dist/index.d.ts

@@ -236,10 +236,14 @@ declare class VoiceAgentClient {
     private enableAudio;
     private currentGeneration;
     private listeners;
+    private configConfirmed;
     private isUserDisconnect;
     private reconnecting;
     private reconnectAttempts;
     private maxReconnectAttempts;
+    private reconnectTimer;
+    private serverPlaybackRate;
+    private serverInputRate;
     private serverUrl;
     constructor(config: LokutorConfig & {
         prompt: string;
@@ -346,6 +350,18 @@ declare class VoiceAgentClient {
      * Update the system prompt mid-conversation
      */
     updatePrompt(newPrompt: string): void;
+    /**
+     * Change the voice style mid-conversation
+     */
+    updateVoice(voice: VoiceStyle): void;
+    /**
+     * Change the language mid-conversation
+     */
+    updateLanguage(language: Language): void;
+    /**
+     * Update tool definitions mid-conversation
+     */
+    updateTools(tools: ToolDefinition[]): void;
     /**
      * Get full conversation transcript
      */

package/dist/index.js

@@ -408,6 +408,9 @@ var BrowserAudioManager = class {
       console.warn("AudioContext not initialized");
       return;
     }
+    if (this.audioContext.state === "suspended") {
+      this.audioContext.resume();
+    }
     if (pcm16Data.length % 2 !== 0) {
       console.warn(`Discarding odd-length PCM buffer (${pcm16Data.length} bytes)`);
       return;
@@ -656,11 +659,16 @@ var VoiceAgentClient = class {
   enableAudio = false;
   currentGeneration = 0;
   listeners = {};
+  configConfirmed = false;
   // Connection resilience
   isUserDisconnect = false;
   reconnecting = false;
   reconnectAttempts = 0;
   maxReconnectAttempts = 5;
+  reconnectTimer = null;
+  // Runtime state
+  serverPlaybackRate = 44100;
+  serverInputRate = 16e3;
   serverUrl;
   constructor(config) {
     this.apiKey = config.apiKey;
@@ -759,6 +767,7 @@ var VoiceAgentClient = class {
         };
         this.ws.onclose = (event) => {
           this.isConnected = false;
+          this.configConfirmed = false;
           const diagnostic = {
             code: event.code,
             reason: event.reason,
@@ -787,12 +796,19 @@ var VoiceAgentClient = class {
             const backoffDelay = Math.min(1e3 * Math.pow(2, this.reconnectAttempts), 1e4);
             console.warn(`Connection lost. Reconnecting in ${backoffDelay}ms (attempt ${this.reconnectAttempts}/${this.maxReconnectAttempts})`);
             if (this.onStatus) this.onStatus("reconnecting");
-            setTimeout(() => {
-              this.connect().catch((e) => console.error("Reconnect failed", e));
+            if (this.reconnectTimer) clearTimeout(this.reconnectTimer);
+            this.reconnectTimer = setTimeout(() => {
+              this.reconnectTimer = null;
+              this.connect().then(() => {
+                this.emit("reconnected");
+              }).catch((e) => console.error("Reconnect failed", e));
             }, backoffDelay);
-          } else {
+          } else if (this.isUserDisconnect) {
             console.log("Disconnected");
             if (this.onStatus) this.onStatus("disconnected");
+          } else {
+            console.log("Max reconnection attempts reached");
+            if (this.onStatus) this.onStatus("disconnected");
           }
         };
       } catch (err) {
@@ -824,9 +840,13 @@ var VoiceAgentClient = class {
    */
   sendConfig() {
     if (!this.ws || !this.isConnected) return;
-    this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
+    this.configConfirmed = false;
     this.ws.send(JSON.stringify({ type: "voice", data: this.voice }));
     this.ws.send(JSON.stringify({ type: "language", data: this.language }));
+    if (this.tools && this.tools.length > 0) {
+      this.ws.send(JSON.stringify({ type: "tools", data: this.tools }));
+    }
+    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", {
@@ -836,9 +856,6 @@ var VoiceAgentClient = class {
       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 }));
-    }
     console.log(`\u2699\uFE0F Configured: voice=${this.voice}, language=${this.language}, visemes=${this.wantVisemes}, tools=${this.tools.length}`);
   }
   /**
@@ -885,6 +902,11 @@ var VoiceAgentClient = class {
             this.handleBinaryMessage(buffer, msg.generation);
           }
           break;
+        case "server_rates":
+          if (msg.playback) this.serverPlaybackRate = msg.playback;
+          if (msg.input) this.serverInputRate = msg.input;
+          sdkTrace("server_rates", { playback: this.serverPlaybackRate, input: this.serverInputRate });
+          break;
         case "transcript":
           const role = msg.role === "user" ? "user" : "agent";
           this.messages.push({
@@ -903,12 +925,17 @@ var VoiceAgentClient = class {
           }
           break;
         case "status":
+          if (msg.data === "prompt_set") {
+            this.configConfirmed = true;
+          }
           if (msg.data === "thinking") {
             const newGen = msg.generation || 0;
             if (newGen > this.currentGeneration) {
-              console.log(`\u{1F9E0} New thought (Gen ${newGen}) - Clearing audio queue`);
+              console.log(`\u{1F9E0} New thought (Gen ${newGen})`);
               this.currentGeneration = newGen;
-              if (this.audioManager) this.audioManager.stopPlayback();
+              if (this.audioManager) {
+                this.audioManager.stopPlayback();
+              }
             }
           }
           if (msg.data === "interrupted" && this.audioManager) {
@@ -1019,6 +1046,10 @@ var VoiceAgentClient = class {
    */
   disconnect() {
     this.isUserDisconnect = true;
+    if (this.reconnectTimer) {
+      clearTimeout(this.reconnectTimer);
+      this.reconnectTimer = null;
+    }
     if (this.ws) {
       this.ws.close();
       this.ws = null;
@@ -1027,6 +1058,7 @@ var VoiceAgentClient = class {
       this.audioManager.cleanup();
     }
     this.isConnected = false;
+    this.configConfirmed = false;
     this.reconnecting = false;
     this.reconnectAttempts = 0;
   }
@@ -1133,6 +1165,36 @@ var VoiceAgentClient = class {
       console.warn("Not connected - prompt will be updated on next connection");
     }
   }
+  /**
+   * Change the voice style mid-conversation
+   */
+  updateVoice(voice) {
+    this.voice = voice;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "voice", data: voice }));
+      console.log(`\u2699\uFE0F Updated voice: ${voice}`);
+    }
+  }
+  /**
+   * Change the language mid-conversation
+   */
+  updateLanguage(language) {
+    this.language = language;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "language", data: language }));
+      console.log(`\u2699\uFE0F Updated language: ${language}`);
+    }
+  }
+  /**
+   * Update tool definitions mid-conversation
+   */
+  updateTools(tools) {
+    this.tools = tools;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "tools", data: tools }));
+      console.log(`\u2699\uFE0F Updated ${tools.length} tool(s)`);
+    }
+  }
   /**
    * Get full conversation transcript
    */

package/dist/index.mjs

@@ -361,6 +361,9 @@ var BrowserAudioManager = class {
       console.warn("AudioContext not initialized");
       return;
     }
+    if (this.audioContext.state === "suspended") {
+      this.audioContext.resume();
+    }
     if (pcm16Data.length % 2 !== 0) {
       console.warn(`Discarding odd-length PCM buffer (${pcm16Data.length} bytes)`);
       return;
@@ -609,11 +612,16 @@ var VoiceAgentClient = class {
   enableAudio = false;
   currentGeneration = 0;
   listeners = {};
+  configConfirmed = false;
   // Connection resilience
   isUserDisconnect = false;
   reconnecting = false;
   reconnectAttempts = 0;
   maxReconnectAttempts = 5;
+  reconnectTimer = null;
+  // Runtime state
+  serverPlaybackRate = 44100;
+  serverInputRate = 16e3;
   serverUrl;
   constructor(config) {
     this.apiKey = config.apiKey;
@@ -712,6 +720,7 @@ var VoiceAgentClient = class {
         };
         this.ws.onclose = (event) => {
           this.isConnected = false;
+          this.configConfirmed = false;
           const diagnostic = {
             code: event.code,
             reason: event.reason,
@@ -740,12 +749,19 @@ var VoiceAgentClient = class {
             const backoffDelay = Math.min(1e3 * Math.pow(2, this.reconnectAttempts), 1e4);
             console.warn(`Connection lost. Reconnecting in ${backoffDelay}ms (attempt ${this.reconnectAttempts}/${this.maxReconnectAttempts})`);
             if (this.onStatus) this.onStatus("reconnecting");
-            setTimeout(() => {
-              this.connect().catch((e) => console.error("Reconnect failed", e));
+            if (this.reconnectTimer) clearTimeout(this.reconnectTimer);
+            this.reconnectTimer = setTimeout(() => {
+              this.reconnectTimer = null;
+              this.connect().then(() => {
+                this.emit("reconnected");
+              }).catch((e) => console.error("Reconnect failed", e));
             }, backoffDelay);
-          } else {
+          } else if (this.isUserDisconnect) {
             console.log("Disconnected");
             if (this.onStatus) this.onStatus("disconnected");
+          } else {
+            console.log("Max reconnection attempts reached");
+            if (this.onStatus) this.onStatus("disconnected");
           }
         };
       } catch (err) {
@@ -777,9 +793,13 @@ var VoiceAgentClient = class {
    */
   sendConfig() {
     if (!this.ws || !this.isConnected) return;
-    this.ws.send(JSON.stringify({ type: "visemes", data: this.wantVisemes }));
+    this.configConfirmed = false;
     this.ws.send(JSON.stringify({ type: "voice", data: this.voice }));
     this.ws.send(JSON.stringify({ type: "language", data: this.language }));
+    if (this.tools && this.tools.length > 0) {
+      this.ws.send(JSON.stringify({ type: "tools", data: this.tools }));
+    }
+    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", {
@@ -789,9 +809,6 @@ var VoiceAgentClient = class {
       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 }));
-    }
     console.log(`\u2699\uFE0F Configured: voice=${this.voice}, language=${this.language}, visemes=${this.wantVisemes}, tools=${this.tools.length}`);
   }
   /**
@@ -838,6 +855,11 @@ var VoiceAgentClient = class {
             this.handleBinaryMessage(buffer, msg.generation);
           }
           break;
+        case "server_rates":
+          if (msg.playback) this.serverPlaybackRate = msg.playback;
+          if (msg.input) this.serverInputRate = msg.input;
+          sdkTrace("server_rates", { playback: this.serverPlaybackRate, input: this.serverInputRate });
+          break;
         case "transcript":
           const role = msg.role === "user" ? "user" : "agent";
           this.messages.push({
@@ -856,12 +878,17 @@ var VoiceAgentClient = class {
           }
           break;
         case "status":
+          if (msg.data === "prompt_set") {
+            this.configConfirmed = true;
+          }
           if (msg.data === "thinking") {
             const newGen = msg.generation || 0;
             if (newGen > this.currentGeneration) {
-              console.log(`\u{1F9E0} New thought (Gen ${newGen}) - Clearing audio queue`);
+              console.log(`\u{1F9E0} New thought (Gen ${newGen})`);
               this.currentGeneration = newGen;
-              if (this.audioManager) this.audioManager.stopPlayback();
+              if (this.audioManager) {
+                this.audioManager.stopPlayback();
+              }
             }
           }
           if (msg.data === "interrupted" && this.audioManager) {
@@ -972,6 +999,10 @@ var VoiceAgentClient = class {
    */
   disconnect() {
     this.isUserDisconnect = true;
+    if (this.reconnectTimer) {
+      clearTimeout(this.reconnectTimer);
+      this.reconnectTimer = null;
+    }
     if (this.ws) {
       this.ws.close();
       this.ws = null;
@@ -980,6 +1011,7 @@ var VoiceAgentClient = class {
       this.audioManager.cleanup();
     }
     this.isConnected = false;
+    this.configConfirmed = false;
     this.reconnecting = false;
     this.reconnectAttempts = 0;
   }
@@ -1086,6 +1118,36 @@ var VoiceAgentClient = class {
       console.warn("Not connected - prompt will be updated on next connection");
     }
   }
+  /**
+   * Change the voice style mid-conversation
+   */
+  updateVoice(voice) {
+    this.voice = voice;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "voice", data: voice }));
+      console.log(`\u2699\uFE0F Updated voice: ${voice}`);
+    }
+  }
+  /**
+   * Change the language mid-conversation
+   */
+  updateLanguage(language) {
+    this.language = language;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "language", data: language }));
+      console.log(`\u2699\uFE0F Updated language: ${language}`);
+    }
+  }
+  /**
+   * Update tool definitions mid-conversation
+   */
+  updateTools(tools) {
+    this.tools = tools;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: "tools", data: tools }));
+      console.log(`\u2699\uFE0F Updated ${tools.length} tool(s)`);
+    }
+  }
   /**
    * Get full conversation transcript
    */

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@lokutor/sdk",
 
-  "version": "1.1.35",
+  "version": "1.1.37",
   "description": "JavaScript/TypeScript SDK for Lokutor Real-time Voice AI",
   "main": "./dist/index.js",
   "module": "./dist/index.mjs",

package/src/browser-audio.ts

@@ -230,6 +230,10 @@ export class BrowserAudioManager {
       return;
     }
 
+    if (this.audioContext.state === 'suspended') {
+      this.audioContext.resume();
+    }
+
     if (pcm16Data.length % 2 !== 0) {
       console.warn(`Discarding odd-length PCM buffer (${pcm16Data.length} bytes)`);
       return;

package/src/client.ts

@@ -166,12 +166,18 @@ export class VoiceAgentClient {
   private enableAudio: boolean = false;
   private currentGeneration: number = 0;
   private listeners: Record<string, Function[]> = {};
+  private configConfirmed: boolean = false;
 
   // Connection resilience
   private isUserDisconnect: boolean = false;
   private reconnecting: boolean = false;
   private reconnectAttempts: number = 0;
   private maxReconnectAttempts: number = 5;
+  private reconnectTimer: ReturnType<typeof setTimeout> | null = null;
+
+  // Runtime state
+  private serverPlaybackRate: number = 44100;
+  private serverInputRate: number = 16000;
 
   private serverUrl: string;
 
@@ -296,6 +302,7 @@ export class VoiceAgentClient {
 
         this.ws.onclose = (event) => {
           this.isConnected = false;
+          this.configConfirmed = false;
           const diagnostic = {
             code: event.code,
             reason: event.reason,
@@ -331,12 +338,19 @@ export class VoiceAgentClient {
 
             if (this.onStatus) this.onStatus('reconnecting');
 
-            setTimeout(() => {
-              this.connect().catch(e => console.error("Reconnect failed", e));
+            if (this.reconnectTimer) clearTimeout(this.reconnectTimer);
+            this.reconnectTimer = setTimeout(() => {
+              this.reconnectTimer = null;
+              this.connect().then(() => {
+                this.emit('reconnected');
+              }).catch(e => console.error("Reconnect failed", e));
             }, backoffDelay);
-          } else {
+          } else if (this.isUserDisconnect) {
             console.log('Disconnected');
             if (this.onStatus) this.onStatus('disconnected');
+          } else {
+            console.log('Max reconnection attempts reached');
+            if (this.onStatus) this.onStatus('disconnected');
           }
         };
 
@@ -374,14 +388,18 @@ export class VoiceAgentClient {
   private sendConfig() {
     if (!this.ws || !this.isConnected) return;
 
-    // Send feature/config flags first so the first generated response uses them.
-    this.ws.send(JSON.stringify({ type: 'visemes', data: this.wantVisemes }));
+    this.configConfirmed = false;
+
+    // Send feature/config flags — order matches the test agent for consistency.
     this.ws.send(JSON.stringify({ type: 'voice', data: this.voice }));
     this.ws.send(JSON.stringify({ type: 'language', data: this.language }));
+    if (this.tools && this.tools.length > 0) {
+      this.ws.send(JSON.stringify({ type: 'tools', data: this.tools }));
+    }
+    this.ws.send(JSON.stringify({ type: 'visemes', data: this.wantVisemes }));
     this.ws.send(JSON.stringify({ type: 'prompt', data: this.prompt }));
 
     // Inform the server of our sample rates for echo cancellation.
-    // These match the SDK defaults: 16 kHz mic input, 44.1 kHz speaker output.
     this.ws.send(JSON.stringify({ type: 'rates', playback: 44100, input: 16000 }));
 
     sdkTrace('ws.send.config', {
@@ -392,10 +410,6 @@ export class VoiceAgentClient {
       tools: this.tools?.length || 0
     });
 
-    if (this.tools && this.tools.length > 0) {
-      this.ws.send(JSON.stringify({ type: 'tools', data: this.tools }));
-    }
-
     console.log(`⚙️ Configured: voice=${this.voice}, language=${this.language}, visemes=${this.wantVisemes}, tools=${this.tools.length}`);
   }
 
@@ -445,6 +459,11 @@ export class VoiceAgentClient {
             this.handleBinaryMessage(buffer, msg.generation);
           }
           break;
+        case 'server_rates':
+          if (msg.playback) this.serverPlaybackRate = msg.playback;
+          if (msg.input) this.serverInputRate = msg.input;
+          sdkTrace('server_rates', { playback: this.serverPlaybackRate, input: this.serverInputRate });
+          break;
         case 'transcript':
           const role = msg.role === 'user' ? 'user' : 'agent';
           // Store in history
@@ -465,12 +484,17 @@ export class VoiceAgentClient {
           }
           break;
         case 'status':
+          if (msg.data === 'prompt_set') {
+            this.configConfirmed = true;
+          }
           if (msg.data === 'thinking') {
             const newGen = msg.generation || 0;
             if (newGen > this.currentGeneration) {
-              console.log(`🧠 New thought (Gen ${newGen}) - Clearing audio queue`);
+              console.log(`🧠 New thought (Gen ${newGen})`);
               this.currentGeneration = newGen;
-              if (this.audioManager) this.audioManager.stopPlayback();
+              if (this.audioManager) {
+                this.audioManager.stopPlayback();
+              }
             }
           }
           if (msg.data === 'interrupted' && this.audioManager) {
@@ -590,6 +614,10 @@ export class VoiceAgentClient {
    */
   public disconnect() {
     this.isUserDisconnect = true;
+    if (this.reconnectTimer) {
+      clearTimeout(this.reconnectTimer);
+      this.reconnectTimer = null;
+    }
     if (this.ws) {
       this.ws.close();
       this.ws = null;
@@ -598,6 +626,7 @@ export class VoiceAgentClient {
       this.audioManager.cleanup();
     }
     this.isConnected = false;
+    this.configConfirmed = false;
     this.reconnecting = false;
     this.reconnectAttempts = 0;
   }
@@ -716,6 +745,39 @@ export class VoiceAgentClient {
     }
   }
 
+  /**
+   * Change the voice style mid-conversation
+   */
+  public updateVoice(voice: VoiceStyle) {
+    this.voice = voice;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: 'voice', data: voice }));
+      console.log(`⚙️ Updated voice: ${voice}`);
+    }
+  }
+
+  /**
+   * Change the language mid-conversation
+   */
+  public updateLanguage(language: Language) {
+    this.language = language;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: 'language', data: language }));
+      console.log(`⚙️ Updated language: ${language}`);
+    }
+  }
+
+  /**
+   * Update tool definitions mid-conversation
+   */
+  public updateTools(tools: ToolDefinition[]) {
+    this.tools = tools;
+    if (this.ws && this.ws.readyState === WebSocket.OPEN && this.isConnected) {
+      this.ws.send(JSON.stringify({ type: 'tools', data: tools }));
+      console.log(`⚙️ Updated ${tools.length} tool(s)`);
+    }
+  }
+
   /**
    * Get full conversation transcript
    */