@glydeunity/voice-sdk 1.2.0 → 1.2.1

package/dist/voice-sdk.es.js

@@ -1,4 +1,4 @@
-const h = `
+const d = `
 class AudioCaptureProcessor extends AudioWorkletProcessor {
   constructor() {
     super();
@@ -33,7 +33,7 @@ class AudioCaptureProcessor extends AudioWorkletProcessor {
 }
 
 registerProcessor('audio-capture-processor', AudioCaptureProcessor);
-`, p = `
+`, u = `
 class AudioPlaybackProcessor extends AudioWorkletProcessor {
   constructor() {
     super();
@@ -131,7 +131,7 @@ class AudioPlaybackProcessor extends AudioWorkletProcessor {
 
 registerProcessor('audio-playback-processor', AudioPlaybackProcessor);
 `;
-class u {
+class g {
   config;
   unityUrl;
   active = !1;
@@ -174,15 +174,15 @@ class u {
    * @returns Voice configuration including system prompt, tools, and Deepgram settings
    */
   async fetchConfig() {
-    const e = `${this.unityUrl}/api/unity/voice/config/${this.config.contextType}`, t = this.config.contextId ? `${e}/${this.config.contextId}` : e, s = await fetch(t, {
+    const e = `${this.unityUrl}/api/unity/voice/config/${this.config.contextType}`, t = this.config.contextId ? `${e}/${this.config.contextId}` : e, i = await fetch(t, {
       method: "GET",
       headers: this.getAuthHeaders()
     });
-    if (!s.ok) {
-      const a = await s.json();
+    if (!i.ok) {
+      const a = await i.json();
       throw new Error(a.error?.message || a.message || "Failed to fetch voice config");
     }
-    const { data: o } = await s.json();
+    const { data: o } = await i.json();
     return o;
   }
   /**
@@ -193,27 +193,29 @@ class u {
       this.active = !0;
       try {
         this.config.systemPrompt || (this.serverConfig = await this.fetchConfig(), console.log("[GlydeVoice] Fetched config:", this.serverConfig));
-        const e = await fetch(`${this.unityUrl}/api/unity/voice/auth`, {
+        const e = {
+          context_id: this.config.contextId,
+          domain: typeof window < "u" ? window.location.hostname : "localhost"
+        };
+        this.config.systemPrompt && (e.system_prompt = this.config.systemPrompt), this.config.deepgramConfig && (e.deepgram_config = this.config.deepgramConfig);
+        const t = await fetch(`${this.unityUrl}/api/unity/voice/auth`, {
           method: "POST",
           headers: this.getAuthHeaders(),
-          body: JSON.stringify({
-            context_id: this.config.contextId,
-            domain: typeof window < "u" ? window.location.hostname : "localhost"
-          })
+          body: JSON.stringify(e)
         });
-        if (!e.ok) {
-          const i = await e.json();
-          throw new Error(i.error?.message || i.message || "Failed to authenticate voice session");
+        if (!t.ok) {
+          const s = await t.json();
+          throw new Error(s.error?.message || s.message || "Failed to authenticate voice session");
         }
-        const { data: t } = await e.json(), { token: s, agent_config: o } = t, a = this.config.systemPrompt || this.serverConfig?.system_prompt || o.instructions || "You are a helpful AI assistant.";
+        const { data: i } = await t.json(), { token: o, agent_config: a, deepgram_config: r } = i, l = this.config.systemPrompt || a.instructions || this.serverConfig?.system_prompt || "You are a helpful AI assistant.";
         await this.initializeAudio();
-        const r = "wss://agent.deepgram.com/v1/agent/converse";
-        this.ws = new WebSocket(r, ["bearer", s]), this.ws.onopen = () => {
-          const i = this.config.deepgramConfig || this.serverConfig?.deepgram_config || {
+        const c = "wss://agent.deepgram.com/v1/agent/converse";
+        this.ws = new WebSocket(c, ["bearer", o]), this.ws.onopen = () => {
+          const s = this.config.deepgramConfig || r || this.serverConfig?.deepgram_config || {
             think: { provider: { type: "open_ai", model: "gpt-4o-mini" } },
             speak: { provider: { type: "deepgram", model: "aura-2-thalia-en" } },
             listen: { provider: { type: "deepgram", model: "nova-2", version: "latest" } }
-          }, n = {
+          }, h = {
             type: "Settings",
             audio: {
               input: {
@@ -228,15 +230,15 @@ class u {
             },
             agent: {
               language: "en",
-              speak: i.speak || {
+              speak: s.speak || {
                 provider: { type: "deepgram", model: "aura-2-thalia-en" }
               },
-              listen: i.listen || {
+              listen: s.listen || {
                 provider: { type: "deepgram", version: "v2", model: "flux-general-en" }
               },
               think: {
-                provider: i.think?.provider || { type: "open_ai", model: "gpt-4o-mini" },
-                functions: i.think?.functions || [
+                provider: s.think?.provider || { type: "open_ai", model: "gpt-4o-mini" },
+                functions: s.think?.functions || [
                   {
                     name: "end_conversation",
                     description: "End the conversation when stop phrases are detected.",
@@ -253,25 +255,25 @@ class u {
               greeting: "Hi! I'm ready to speak with you. How can I help you today?"
             }
           };
-          this.ws.send(JSON.stringify(n)), this.emit({ type: "open", payload: { config: o, serverConfig: this.serverConfig } });
+          this.ws.send(JSON.stringify(h)), this.emit({ type: "open", payload: { config: a, serverConfig: this.serverConfig } });
         };
-        const l = a;
-        this.ws.onmessage = (i) => {
-          if (typeof i.data == "string") {
+        const n = l;
+        this.ws.onmessage = (s) => {
+          if (typeof s.data == "string") {
             try {
-              if (JSON.parse(i.data).type === "SettingsApplied") {
-                const c = {
+              if (JSON.parse(s.data).type === "SettingsApplied") {
+                const p = {
                   type: "UpdatePrompt",
-                  prompt: l
+                  prompt: n
                 };
-                this.ws.send(JSON.stringify(c)), this.startMicrophone();
+                this.ws.send(JSON.stringify(p)), this.startMicrophone();
               }
             } catch {
             }
-            this.handleTextMessage(i.data);
-          } else i.data instanceof Blob ? this.handleAudioData(i.data) : i.data instanceof ArrayBuffer && this.handleAudioBuffer(i.data);
-        }, this.ws.onerror = (i) => {
-          console.error("[GlydeVoice] WebSocket error:", i), this.emit({ type: "error", payload: i });
+            this.handleTextMessage(s.data);
+          } else s.data instanceof Blob ? this.handleAudioData(s.data) : s.data instanceof ArrayBuffer && this.handleAudioBuffer(s.data);
+        }, this.ws.onerror = (s) => {
+          console.error("[GlydeVoice] WebSocket error:", s), this.emit({ type: "error", payload: s });
         }, this.ws.onclose = () => {
           this.cleanup(), this.emit({ type: "close" });
         }, this.renderUI();
@@ -296,7 +298,7 @@ class u {
    */
   async initializeAudio() {
     this.audioContext = new AudioContext({ sampleRate: this.inputSampleRate });
-    const e = this.createWorkletBlobUrl(h), t = this.createWorkletBlobUrl(p);
+    const e = this.createWorkletBlobUrl(d), t = this.createWorkletBlobUrl(u);
     try {
       await Promise.all([
         this.audioContext.audioWorklet.addModule(e),
@@ -305,8 +307,8 @@ class u {
     } finally {
       URL.revokeObjectURL(e), URL.revokeObjectURL(t);
     }
-    this.playbackWorkletNode = new AudioWorkletNode(this.audioContext, "audio-playback-processor"), this.playbackWorkletNode.connect(this.audioContext.destination), this.playbackWorkletNode.port.onmessage = (s) => {
-      const { type: o } = s.data;
+    this.playbackWorkletNode = new AudioWorkletNode(this.audioContext, "audio-playback-processor"), this.playbackWorkletNode.connect(this.audioContext.destination), this.playbackWorkletNode.port.onmessage = (i) => {
+      const { type: o } = i.data;
       (o === "cleared" || o === "bufferEmpty") && (this.isAgentSpeaking = !1, this.agentAudioDoneReceived = !1, this.emit({ type: "agent_speaking", payload: !1 }));
     };
   }
@@ -330,8 +332,8 @@ class u {
           break;
         case "ConversationText":
           if (t.content && t.content.trim()) {
-            const s = t.role === "assistant" ? "agent" : "user";
-            this.config.onTranscript && this.config.onTranscript(t.content, s), this.emit({ type: "transcript", payload: { text: t.content, role: s } }), this.saveTranscript(t.content, t.role);
+            const i = t.role === "assistant" ? "agent" : "user";
+            this.config.onTranscript && this.config.onTranscript(t.content, i), this.emit({ type: "transcript", payload: { text: t.content, role: i } }), this.saveTranscript(t.content, t.role);
           }
           break;
         case "AgentStartedSpeaking":
@@ -364,30 +366,30 @@ class u {
     this.audioContext.state === "suspended" && this.audioContext.resume();
     const t = e.byteLength;
     if (t === 0) return;
-    const s = t - t % 2;
-    if (s === 0) return;
-    const o = s === t ? e : e.slice(0, s), a = new Int16Array(o), r = new Float32Array(a.length);
+    const i = t - t % 2;
+    if (i === 0) return;
+    const o = i === t ? e : e.slice(0, i), a = new Int16Array(o), r = new Float32Array(a.length);
     for (let n = 0; n < a.length; n++)
       r[n] = a[n] / 32768;
     const l = this.resample24kTo48k(r);
     !this.isAgentSpeaking && !this.agentAudioDoneReceived && (this.isAgentSpeaking = !0, this.emit({ type: "agent_speaking", payload: !0 }));
-    const i = new Float32Array(l);
+    const c = new Float32Array(l);
     this.playbackWorkletNode.port.postMessage({
       type: "audio",
-      data: i
-    }, [i.buffer]);
+      data: c
+    }, [c.buffer]);
   }
   /**
    * Resample audio from 24kHz to 48kHz using linear interpolation
    */
   resample24kTo48k(e) {
-    const t = e.length * 2, s = new Float32Array(t);
+    const t = e.length * 2, i = new Float32Array(t);
     for (let a = 0; a < e.length - 1; a++) {
       const r = e[a], l = e[a + 1];
-      s[a * 2] = r, s[a * 2 + 1] = (r + l) / 2;
+      i[a * 2] = r, i[a * 2 + 1] = (r + l) / 2;
     }
     const o = e.length - 1;
-    return s[o * 2] = e[o], s[o * 2 + 1] = e[o], s;
+    return i[o * 2] = e[o], i[o * 2 + 1] = e[o], i;
   }
   /**
    * Clear the playback buffer (for interruption handling)
@@ -496,5 +498,5 @@ class u {
   }
 }
 export {
-  u as GlydeVoice
+  g as GlydeVoice
 };

package/dist/voice-sdk.umd.js

@@ -1,4 +1,4 @@
-(function(l,c){typeof exports=="object"&&typeof module<"u"?c(exports):typeof define=="function"&&define.amd?define(["exports"],c):(l=typeof globalThis<"u"?globalThis:l||self,c(l.GlydeVoice={}))})(this,(function(l){"use strict";const c=`
+(function(n,c){typeof exports=="object"&&typeof module<"u"?c(exports):typeof define=="function"&&define.amd?define(["exports"],c):(n=typeof globalThis<"u"?globalThis:n||self,c(n.GlydeVoice={}))})(this,(function(n){"use strict";const c=`
 class AudioCaptureProcessor extends AudioWorkletProcessor {
   constructor() {
     super();
@@ -33,7 +33,7 @@ class AudioCaptureProcessor extends AudioWorkletProcessor {
 }
 
 registerProcessor('audio-capture-processor', AudioCaptureProcessor);
-`,d=`
+`,u=`
 class AudioPlaybackProcessor extends AudioWorkletProcessor {
   constructor() {
     super();
@@ -130,11 +130,11 @@ class AudioPlaybackProcessor extends AudioWorkletProcessor {
 }
 
 registerProcessor('audio-playback-processor', AudioPlaybackProcessor);
-`;class p{config;unityUrl;active=!1;serverConfig=null;ws=null;audioContext=null;mediaStream=null;captureWorkletNode=null;playbackWorkletNode=null;isMuted=!1;outputSampleRate=24e3;inputSampleRate=48e3;isAgentSpeaking=!1;agentAudioDoneReceived=!1;constructor(e){this.config=e,this.unityUrl=e.unityBaseUrl||"https://api.glydeunity.com",!e.publishableKey&&!e.apiKey&&!e.authToken&&console.warn("[GlydeVoice] No authentication method provided. One of publishableKey, apiKey, or authToken is required.")}getAuthHeaders(){const e={"Content-Type":"application/json"};return this.config.publishableKey&&(e["x-publishable-key"]=this.config.publishableKey),this.config.apiKey&&(e["x-api-key"]=this.config.apiKey),this.config.authToken&&(e.Authorization=`Bearer ${this.config.authToken}`),e}async fetchConfig(){const e=`${this.unityUrl}/api/unity/voice/config/${this.config.contextType}`,t=this.config.contextId?`${e}/${this.config.contextId}`:e,s=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!s.ok){const o=await s.json();throw new Error(o.error?.message||o.message||"Failed to fetch voice config")}const{data:a}=await s.json();return a}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e=await fetch(`${this.unityUrl}/api/unity/voice/auth`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify({context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"})});if(!e.ok){const i=await e.json();throw new Error(i.error?.message||i.message||"Failed to authenticate voice session")}const{data:t}=await e.json(),{token:s,agent_config:a}=t,o=this.config.systemPrompt||this.serverConfig?.system_prompt||a.instructions||"You are a helpful AI assistant.";await this.initializeAudio();const r="wss://agent.deepgram.com/v1/agent/converse";this.ws=new WebSocket(r,["bearer",s]),this.ws.onopen=()=>{const i=this.config.deepgramConfig||this.serverConfig?.deepgram_config||{think:{provider:{type:"open_ai",model:"gpt-4o-mini"}},speak:{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:{provider:{type:"deepgram",model:"nova-2",version:"latest"}}},n={type:"Settings",audio:{input:{encoding:"linear16",sample_rate:this.inputSampleRate},output:{encoding:"linear16",sample_rate:this.outputSampleRate,container:"none"}},agent:{language:"en",speak:i.speak||{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:i.listen||{provider:{type:"deepgram",version:"v2",model:"flux-general-en"}},think:{provider:i.think?.provider||{type:"open_ai",model:"gpt-4o-mini"},functions:i.think?.functions||[{name:"end_conversation",description:"End the conversation when stop phrases are detected.",parameters:{type:"object",properties:{item:{type:"string",description:"The phrase that triggered end of conversation"}},required:["item"]}}]},greeting:"Hi! I'm ready to speak with you. How can I help you today?"}};this.ws.send(JSON.stringify(n)),this.emit({type:"open",payload:{config:a,serverConfig:this.serverConfig}})};const h=o;this.ws.onmessage=i=>{if(typeof i.data=="string"){try{if(JSON.parse(i.data).type==="SettingsApplied"){const u={type:"UpdatePrompt",prompt:h};this.ws.send(JSON.stringify(u)),this.startMicrophone()}}catch{}this.handleTextMessage(i.data)}else i.data instanceof Blob?this.handleAudioData(i.data):i.data instanceof ArrayBuffer&&this.handleAudioBuffer(i.data)},this.ws.onerror=i=>{console.error("[GlydeVoice] WebSocket error:",i),this.emit({type:"error",payload:i})},this.ws.onclose=()=>{this.cleanup(),this.emit({type:"close"})},this.renderUI()}catch(e){throw console.error("[GlydeVoice] Error starting session:",e),this.active=!1,this.emit({type:"error",payload:e}),e}}}createWorkletBlobUrl(e){const t=new Blob([e],{type:"application/javascript"});return URL.createObjectURL(t)}async initializeAudio(){this.audioContext=new AudioContext({sampleRate:this.inputSampleRate});const e=this.createWorkletBlobUrl(c),t=this.createWorkletBlobUrl(d);try{await Promise.all([this.audioContext.audioWorklet.addModule(e),this.audioContext.audioWorklet.addModule(t)])}finally{URL.revokeObjectURL(e),URL.revokeObjectURL(t)}this.playbackWorkletNode=new AudioWorkletNode(this.audioContext,"audio-playback-processor"),this.playbackWorkletNode.connect(this.audioContext.destination),this.playbackWorkletNode.port.onmessage=s=>{const{type:a}=s.data;(a==="cleared"||a==="bufferEmpty")&&(this.isAgentSpeaking=!1,this.agentAudioDoneReceived=!1,this.emit({type:"agent_speaking",payload:!1}))}}handleTextMessage(e){try{const t=JSON.parse(e);switch(t.type){case"Welcome":this.emit({type:"ready"});break;case"SettingsApplied":break;case"UserStartedSpeaking":this.emit({type:"user_speaking",payload:!0}),this.clearPlaybackBuffer(),this.isAgentSpeaking=!1,this.agentAudioDoneReceived=!1;break;case"UserStoppedSpeaking":this.emit({type:"user_speaking",payload:!1});break;case"ConversationText":if(t.content&&t.content.trim()){const s=t.role==="assistant"?"agent":"user";this.config.onTranscript&&this.config.onTranscript(t.content,s),this.emit({type:"transcript",payload:{text:t.content,role:s}}),this.saveTranscript(t.content,t.role)}break;case"AgentStartedSpeaking":this.isAgentSpeaking=!0,this.agentAudioDoneReceived=!1,this.emit({type:"agent_speaking",payload:!0});break;case"AgentAudioDone":this.agentAudioDoneReceived=!0;break;case"Error":console.error("[GlydeVoice] Agent error:",t),this.emit({type:"error",payload:t});break}}catch(t){console.error("[GlydeVoice] Failed to parse message:",t)}}async handleAudioData(e){const t=await e.arrayBuffer();this.handleAudioBuffer(t)}handleAudioBuffer(e){if(!this.playbackWorkletNode||!this.audioContext)return;this.audioContext.state==="suspended"&&this.audioContext.resume();const t=e.byteLength;if(t===0)return;const s=t-t%2;if(s===0)return;const a=s===t?e:e.slice(0,s),o=new Int16Array(a),r=new Float32Array(o.length);for(let n=0;n<o.length;n++)r[n]=o[n]/32768;const h=this.resample24kTo48k(r);!this.isAgentSpeaking&&!this.agentAudioDoneReceived&&(this.isAgentSpeaking=!0,this.emit({type:"agent_speaking",payload:!0}));const i=new Float32Array(h);this.playbackWorkletNode.port.postMessage({type:"audio",data:i},[i.buffer])}resample24kTo48k(e){const t=e.length*2,s=new Float32Array(t);for(let o=0;o<e.length-1;o++){const r=e[o],h=e[o+1];s[o*2]=r,s[o*2+1]=(r+h)/2}const a=e.length-1;return s[a*2]=e[a],s[a*2+1]=e[a],s}clearPlaybackBuffer(){this.playbackWorkletNode&&this.playbackWorkletNode.port.postMessage({type:"clear"})}async startMicrophone(){if(!this.audioContext)throw new Error("Audio context not initialized");try{this.mediaStream=await navigator.mediaDevices.getUserMedia({audio:{channelCount:1,sampleRate:this.inputSampleRate,echoCancellation:!0,noiseSuppression:!0}});const e=this.audioContext.createMediaStreamSource(this.mediaStream);this.captureWorkletNode=new AudioWorkletNode(this.audioContext,"audio-capture-processor"),this.captureWorkletNode.port.onmessage=t=>{!this.active||!this.ws||this.ws.readyState!==WebSocket.OPEN||this.isMuted||this.ws.send(t.data)},e.connect(this.captureWorkletNode),this.emit({type:"microphone_ready"})}catch(e){throw console.error("[GlydeVoice] Microphone error:",e),e}}async saveTranscript(e,t){if(!(!this.config.contextId||!e))try{await fetch(`${this.unityUrl}/api/unity/voice/transcript`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify({context_id:this.config.contextId,content:e,role:t==="assistant"?"assistant":"user"})})}catch{}}setMuted(e){this.isMuted=e}getMuted(){return this.isMuted}isActive(){return this.active}getServerConfig(){return this.serverConfig}stop(){this.active=!1,this.cleanup()}cleanup(){this.captureWorkletNode&&(this.captureWorkletNode.disconnect(),this.captureWorkletNode.port.close(),this.captureWorkletNode=null),this.playbackWorkletNode&&(this.playbackWorkletNode.disconnect(),this.playbackWorkletNode.port.close(),this.playbackWorkletNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(e=>e.stop()),this.mediaStream=null),this.audioContext&&(this.audioContext.close(),this.audioContext=null),this.ws&&(this.ws.readyState===WebSocket.OPEN&&this.ws.close(),this.ws=null)}emit(e){this.config.onEvent&&this.config.onEvent(e)}renderUI(){if(!this.config.container)return;const e=typeof this.config.container=="string"?document.querySelector(this.config.container):this.config.container;e&&(e.innerHTML=`
+`;class f{config;unityUrl;active=!1;serverConfig=null;ws=null;audioContext=null;mediaStream=null;captureWorkletNode=null;playbackWorkletNode=null;isMuted=!1;outputSampleRate=24e3;inputSampleRate=48e3;isAgentSpeaking=!1;agentAudioDoneReceived=!1;constructor(e){this.config=e,this.unityUrl=e.unityBaseUrl||"https://api.glydeunity.com",!e.publishableKey&&!e.apiKey&&!e.authToken&&console.warn("[GlydeVoice] No authentication method provided. One of publishableKey, apiKey, or authToken is required.")}getAuthHeaders(){const e={"Content-Type":"application/json"};return this.config.publishableKey&&(e["x-publishable-key"]=this.config.publishableKey),this.config.apiKey&&(e["x-api-key"]=this.config.apiKey),this.config.authToken&&(e.Authorization=`Bearer ${this.config.authToken}`),e}async fetchConfig(){const e=`${this.unityUrl}/api/unity/voice/config/${this.config.contextType}`,t=this.config.contextId?`${e}/${this.config.contextId}`:e,i=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!i.ok){const o=await i.json();throw new Error(o.error?.message||o.message||"Failed to fetch voice config")}const{data:a}=await i.json();return a}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e={context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"};this.config.systemPrompt&&(e.system_prompt=this.config.systemPrompt),this.config.deepgramConfig&&(e.deepgram_config=this.config.deepgramConfig);const t=await fetch(`${this.unityUrl}/api/unity/voice/auth`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify(e)});if(!t.ok){const s=await t.json();throw new Error(s.error?.message||s.message||"Failed to authenticate voice session")}const{data:i}=await t.json(),{token:a,agent_config:o,deepgram_config:r}=i,h=this.config.systemPrompt||o.instructions||this.serverConfig?.system_prompt||"You are a helpful AI assistant.";await this.initializeAudio();const d="wss://agent.deepgram.com/v1/agent/converse";this.ws=new WebSocket(d,["bearer",a]),this.ws.onopen=()=>{const s=this.config.deepgramConfig||r||this.serverConfig?.deepgram_config||{think:{provider:{type:"open_ai",model:"gpt-4o-mini"}},speak:{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:{provider:{type:"deepgram",model:"nova-2",version:"latest"}}},p={type:"Settings",audio:{input:{encoding:"linear16",sample_rate:this.inputSampleRate},output:{encoding:"linear16",sample_rate:this.outputSampleRate,container:"none"}},agent:{language:"en",speak:s.speak||{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:s.listen||{provider:{type:"deepgram",version:"v2",model:"flux-general-en"}},think:{provider:s.think?.provider||{type:"open_ai",model:"gpt-4o-mini"},functions:s.think?.functions||[{name:"end_conversation",description:"End the conversation when stop phrases are detected.",parameters:{type:"object",properties:{item:{type:"string",description:"The phrase that triggered end of conversation"}},required:["item"]}}]},greeting:"Hi! I'm ready to speak with you. How can I help you today?"}};this.ws.send(JSON.stringify(p)),this.emit({type:"open",payload:{config:o,serverConfig:this.serverConfig}})};const l=h;this.ws.onmessage=s=>{if(typeof s.data=="string"){try{if(JSON.parse(s.data).type==="SettingsApplied"){const g={type:"UpdatePrompt",prompt:l};this.ws.send(JSON.stringify(g)),this.startMicrophone()}}catch{}this.handleTextMessage(s.data)}else s.data instanceof Blob?this.handleAudioData(s.data):s.data instanceof ArrayBuffer&&this.handleAudioBuffer(s.data)},this.ws.onerror=s=>{console.error("[GlydeVoice] WebSocket error:",s),this.emit({type:"error",payload:s})},this.ws.onclose=()=>{this.cleanup(),this.emit({type:"close"})},this.renderUI()}catch(e){throw console.error("[GlydeVoice] Error starting session:",e),this.active=!1,this.emit({type:"error",payload:e}),e}}}createWorkletBlobUrl(e){const t=new Blob([e],{type:"application/javascript"});return URL.createObjectURL(t)}async initializeAudio(){this.audioContext=new AudioContext({sampleRate:this.inputSampleRate});const e=this.createWorkletBlobUrl(c),t=this.createWorkletBlobUrl(u);try{await Promise.all([this.audioContext.audioWorklet.addModule(e),this.audioContext.audioWorklet.addModule(t)])}finally{URL.revokeObjectURL(e),URL.revokeObjectURL(t)}this.playbackWorkletNode=new AudioWorkletNode(this.audioContext,"audio-playback-processor"),this.playbackWorkletNode.connect(this.audioContext.destination),this.playbackWorkletNode.port.onmessage=i=>{const{type:a}=i.data;(a==="cleared"||a==="bufferEmpty")&&(this.isAgentSpeaking=!1,this.agentAudioDoneReceived=!1,this.emit({type:"agent_speaking",payload:!1}))}}handleTextMessage(e){try{const t=JSON.parse(e);switch(t.type){case"Welcome":this.emit({type:"ready"});break;case"SettingsApplied":break;case"UserStartedSpeaking":this.emit({type:"user_speaking",payload:!0}),this.clearPlaybackBuffer(),this.isAgentSpeaking=!1,this.agentAudioDoneReceived=!1;break;case"UserStoppedSpeaking":this.emit({type:"user_speaking",payload:!1});break;case"ConversationText":if(t.content&&t.content.trim()){const i=t.role==="assistant"?"agent":"user";this.config.onTranscript&&this.config.onTranscript(t.content,i),this.emit({type:"transcript",payload:{text:t.content,role:i}}),this.saveTranscript(t.content,t.role)}break;case"AgentStartedSpeaking":this.isAgentSpeaking=!0,this.agentAudioDoneReceived=!1,this.emit({type:"agent_speaking",payload:!0});break;case"AgentAudioDone":this.agentAudioDoneReceived=!0;break;case"Error":console.error("[GlydeVoice] Agent error:",t),this.emit({type:"error",payload:t});break}}catch(t){console.error("[GlydeVoice] Failed to parse message:",t)}}async handleAudioData(e){const t=await e.arrayBuffer();this.handleAudioBuffer(t)}handleAudioBuffer(e){if(!this.playbackWorkletNode||!this.audioContext)return;this.audioContext.state==="suspended"&&this.audioContext.resume();const t=e.byteLength;if(t===0)return;const i=t-t%2;if(i===0)return;const a=i===t?e:e.slice(0,i),o=new Int16Array(a),r=new Float32Array(o.length);for(let l=0;l<o.length;l++)r[l]=o[l]/32768;const h=this.resample24kTo48k(r);!this.isAgentSpeaking&&!this.agentAudioDoneReceived&&(this.isAgentSpeaking=!0,this.emit({type:"agent_speaking",payload:!0}));const d=new Float32Array(h);this.playbackWorkletNode.port.postMessage({type:"audio",data:d},[d.buffer])}resample24kTo48k(e){const t=e.length*2,i=new Float32Array(t);for(let o=0;o<e.length-1;o++){const r=e[o],h=e[o+1];i[o*2]=r,i[o*2+1]=(r+h)/2}const a=e.length-1;return i[a*2]=e[a],i[a*2+1]=e[a],i}clearPlaybackBuffer(){this.playbackWorkletNode&&this.playbackWorkletNode.port.postMessage({type:"clear"})}async startMicrophone(){if(!this.audioContext)throw new Error("Audio context not initialized");try{this.mediaStream=await navigator.mediaDevices.getUserMedia({audio:{channelCount:1,sampleRate:this.inputSampleRate,echoCancellation:!0,noiseSuppression:!0}});const e=this.audioContext.createMediaStreamSource(this.mediaStream);this.captureWorkletNode=new AudioWorkletNode(this.audioContext,"audio-capture-processor"),this.captureWorkletNode.port.onmessage=t=>{!this.active||!this.ws||this.ws.readyState!==WebSocket.OPEN||this.isMuted||this.ws.send(t.data)},e.connect(this.captureWorkletNode),this.emit({type:"microphone_ready"})}catch(e){throw console.error("[GlydeVoice] Microphone error:",e),e}}async saveTranscript(e,t){if(!(!this.config.contextId||!e))try{await fetch(`${this.unityUrl}/api/unity/voice/transcript`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify({context_id:this.config.contextId,content:e,role:t==="assistant"?"assistant":"user"})})}catch{}}setMuted(e){this.isMuted=e}getMuted(){return this.isMuted}isActive(){return this.active}getServerConfig(){return this.serverConfig}stop(){this.active=!1,this.cleanup()}cleanup(){this.captureWorkletNode&&(this.captureWorkletNode.disconnect(),this.captureWorkletNode.port.close(),this.captureWorkletNode=null),this.playbackWorkletNode&&(this.playbackWorkletNode.disconnect(),this.playbackWorkletNode.port.close(),this.playbackWorkletNode=null),this.mediaStream&&(this.mediaStream.getTracks().forEach(e=>e.stop()),this.mediaStream=null),this.audioContext&&(this.audioContext.close(),this.audioContext=null),this.ws&&(this.ws.readyState===WebSocket.OPEN&&this.ws.close(),this.ws=null)}emit(e){this.config.onEvent&&this.config.onEvent(e)}renderUI(){if(!this.config.container)return;const e=typeof this.config.container=="string"?document.querySelector(this.config.container):this.config.container;e&&(e.innerHTML=`
         <div style="padding: 20px; border: 1px solid #ccc; border-radius: 8px; background: #fff;">
           <h3>Glyde Voice Agent</h3>
           <p>Status: Active</p>
           <p>Context: ${this.config.contextType}</p>
           <button onclick="this.closest('div').remove()">Close</button>
         </div>
-      `)}}l.GlydeVoice=p,Object.defineProperty(l,Symbol.toStringTag,{value:"Module"})}));
+      `)}}n.GlydeVoice=f,Object.defineProperty(n,Symbol.toStringTag,{value:"Module"})}));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@glydeunity/voice-sdk",
-  "version": "1.2.0",
+  "version": "1.2.1",
   "description": "GLYDE Voice Agent SDK - AI-powered voice interactions for web applications",
   "type": "module",
   "main": "./dist/voice-sdk.umd.js",