@glydeunity/voice-sdk 1.3.3 → 1.3.4

package/dist/voice-sdk.es.js

@@ -226,7 +226,7 @@ class y {
         }
         this.ws = new WebSocket(l, ["bearer", i]), this.ws.onopen = () => {
           const s = r || {
-            think: { provider: { type: "open_ai", model: "gpt-5-nano" } },
+            think: { provider: { type: "open_ai", model: "gpt-4.1-nano" } },
             speak: { provider: { type: "deepgram", model: "aura-2-thalia-en" } },
             listen: { provider: { type: "deepgram", version: "v2", model: "flux-general-en" } }
           }, h = {
@@ -251,7 +251,7 @@ class y {
                 provider: { type: "deepgram", version: "v2", model: "flux-general-en" }
               },
               think: {
-                provider: s.think?.provider || { type: "open_ai", model: "gpt-5-nano" },
+                provider: s.think?.provider || { type: "open_ai", model: "gpt-4.1-nano" },
                 // Functions come from server config - no client-side defaults
                 ...s.think?.functions && { functions: s.think.functions }
               },

package/dist/voice-sdk.umd.js

@@ -130,7 +130,7 @@ class AudioPlaybackProcessor extends AudioWorkletProcessor {
 }
 
 registerProcessor('audio-playback-processor', AudioPlaybackProcessor);
-`;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;sessionContext={};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}/${this.config.contextId}`:e,o=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!o.ok){const a=await o.json();throw new Error(a.error?.message||a.message||"Failed to fetch voice config")}const{data:i}=await o.json();return i}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e={context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"};this.config.systemPrompt&&(e.system_prompt=this.config.systemPrompt),this.config.deepgramConfig&&(e.deepgram_config=this.config.deepgramConfig);const t=await fetch(`${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:o}=await t.json(),{token:i,agent_config:a,deepgram_config:n}=o;this.setSessionContext({clientUuid:a?.client_uuid,contextId:this.config.contextId,contextType:this.config.contextType,currentJobUuid:a?.job_uuid});const d=this.config.systemPrompt||a.instructions||this.serverConfig?.system_prompt||"You are a helpful AI assistant.";await this.initializeAudio();let h="wss://agent.deepgram.com/v1/agent/converse";const r=this.config.deepgramConfig||n||this.serverConfig?.deepgram_config;if(r?.tags&&r.tags.length>0){const s=new URLSearchParams;r.tags.forEach(u=>s.append("tag",u)),h+=`?${s.toString()}`}this.ws=new WebSocket(h,["bearer",i]),this.ws.onopen=()=>{const s=r||{think:{provider:{type:"open_ai",model:"gpt-5-nano"}},speak:{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:{provider:{type:"deepgram",version:"v2",model:"flux-general-en"}}},u={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-5-nano"},...s.think?.functions&&{functions:s.think.functions}},greeting:"Hi! I'm excited you chose to speak with me. Are you ready to start?"}};s.tags&&s.tags.length>0&&(u.tags=s.tags),this.ws.send(JSON.stringify(u)),this.emit({type:"open",payload:{config:a,serverConfig:this.serverConfig}})};const g=d;this.ws.onmessage=s=>{if(typeof s.data=="string"){try{if(JSON.parse(s.data).type==="SettingsApplied"){const y={type:"UpdatePrompt",prompt:g};this.ws.send(JSON.stringify(y)),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(l),t=this.createWorkletBlobUrl(p);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=o=>{const{type:i}=o.data;(i==="cleared"||i==="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 o=t.role==="assistant"?"agent":"user";this.config.onTranscript&&this.config.onTranscript(t.content,o),this.emit({type:"transcript",payload:{text:t.content,role:o}}),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;case"FunctionCallRequest":this.handleFunctionCallRequest(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 o=t-t%2;if(o===0)return;const i=o===t?e:e.slice(0,o),a=new Int16Array(i),n=new Float32Array(a.length);for(let r=0;r<a.length;r++)n[r]=a[r]/32768;const d=this.resample24kTo48k(n);!this.isAgentSpeaking&&!this.agentAudioDoneReceived&&(this.isAgentSpeaking=!0,this.emit({type:"agent_speaking",payload:!0}));const h=new Float32Array(d);this.playbackWorkletNode.port.postMessage({type:"audio",data:h},[h.buffer])}resample24kTo48k(e){const t=e.length*2,o=new Float32Array(t);for(let a=0;a<e.length-1;a++){const n=e[a],d=e[a+1];o[a*2]=n,o[a*2+1]=(n+d)/2}const i=e.length-1;return o[i*2]=e[i],o[i*2+1]=e[i],o}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;sessionContext={};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}/${this.config.contextId}`:e,o=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!o.ok){const a=await o.json();throw new Error(a.error?.message||a.message||"Failed to fetch voice config")}const{data:i}=await o.json();return i}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e={context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"};this.config.systemPrompt&&(e.system_prompt=this.config.systemPrompt),this.config.deepgramConfig&&(e.deepgram_config=this.config.deepgramConfig);const t=await fetch(`${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:o}=await t.json(),{token:i,agent_config:a,deepgram_config:n}=o;this.setSessionContext({clientUuid:a?.client_uuid,contextId:this.config.contextId,contextType:this.config.contextType,currentJobUuid:a?.job_uuid});const d=this.config.systemPrompt||a.instructions||this.serverConfig?.system_prompt||"You are a helpful AI assistant.";await this.initializeAudio();let h="wss://agent.deepgram.com/v1/agent/converse";const r=this.config.deepgramConfig||n||this.serverConfig?.deepgram_config;if(r?.tags&&r.tags.length>0){const s=new URLSearchParams;r.tags.forEach(u=>s.append("tag",u)),h+=`?${s.toString()}`}this.ws=new WebSocket(h,["bearer",i]),this.ws.onopen=()=>{const s=r||{think:{provider:{type:"open_ai",model:"gpt-4.1-nano"}},speak:{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:{provider:{type:"deepgram",version:"v2",model:"flux-general-en"}}},u={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-4.1-nano"},...s.think?.functions&&{functions:s.think.functions}},greeting:"Hi! I'm excited you chose to speak with me. Are you ready to start?"}};s.tags&&s.tags.length>0&&(u.tags=s.tags),this.ws.send(JSON.stringify(u)),this.emit({type:"open",payload:{config:a,serverConfig:this.serverConfig}})};const g=d;this.ws.onmessage=s=>{if(typeof s.data=="string"){try{if(JSON.parse(s.data).type==="SettingsApplied"){const y={type:"UpdatePrompt",prompt:g};this.ws.send(JSON.stringify(y)),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(l),t=this.createWorkletBlobUrl(p);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=o=>{const{type:i}=o.data;(i==="cleared"||i==="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 o=t.role==="assistant"?"agent":"user";this.config.onTranscript&&this.config.onTranscript(t.content,o),this.emit({type:"transcript",payload:{text:t.content,role:o}}),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;case"FunctionCallRequest":this.handleFunctionCallRequest(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 o=t-t%2;if(o===0)return;const i=o===t?e:e.slice(0,o),a=new Int16Array(i),n=new Float32Array(a.length);for(let r=0;r<a.length;r++)n[r]=a[r]/32768;const d=this.resample24kTo48k(n);!this.isAgentSpeaking&&!this.agentAudioDoneReceived&&(this.isAgentSpeaking=!0,this.emit({type:"agent_speaking",payload:!0}));const h=new Float32Array(d);this.playbackWorkletNode.port.postMessage({type:"audio",data:h},[h.buffer])}resample24kTo48k(e){const t=e.length*2,o=new Float32Array(t);for(let a=0;a<e.length-1;a++){const n=e[a],d=e[a+1];o[a*2]=n,o[a*2+1]=(n+d)/2}const i=e.length-1;return o[i*2]=e[i],o[i*2+1]=e[i],o}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>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@glydeunity/voice-sdk",
-  "version": "1.3.3",
+  "version": "1.3.4",
   "description": "GLYDE Voice Agent SDK - AI-powered voice interactions for web applications",
   "type": "module",
   "main": "./dist/voice-sdk.umd.js",