@glydeunity/voice-sdk 1.3.6 → 1.4.0

package/dist/voice-sdk.umd.js

@@ -1,140 +0,0 @@
-(function(c,l){typeof exports=="object"&&typeof module<"u"?l(exports):typeof define=="function"&&define.amd?define(["exports"],l):(c=typeof globalThis<"u"?globalThis:c||self,l(c.GlydeVoice={}))})(this,(function(c){"use strict";const l=`
-class AudioCaptureProcessor extends AudioWorkletProcessor {
-  constructor() {
-    super();
-    this.bufferSize = 4096;
-    this.buffer = new Float32Array(this.bufferSize);
-    this.bufferIndex = 0;
-  }
-
-  process(inputs) {
-    const input = inputs[0];
-    if (!input || !input[0]) return true;
-
-    const samples = input[0];
-    
-    for (let i = 0; i < samples.length; i++) {
-      this.buffer[this.bufferIndex++] = samples[i];
-      
-      if (this.bufferIndex >= this.bufferSize) {
-        const pcm16 = new Int16Array(this.bufferSize);
-        for (let j = 0; j < this.bufferSize; j++) {
-          const s = Math.max(-1, Math.min(1, this.buffer[j]));
-          pcm16[j] = s < 0 ? s * 0x8000 : s * 0x7FFF;
-        }
-        
-        this.port.postMessage(pcm16.buffer, [pcm16.buffer]);
-        this.bufferIndex = 0;
-      }
-    }
-
-    return true;
-  }
-}
-
-registerProcessor('audio-capture-processor', AudioCaptureProcessor);
-`,p=`
-class AudioPlaybackProcessor extends AudioWorkletProcessor {
-  constructor() {
-    super();
-    
-    this.bufferSize = 48000 * 15;
-    this.buffer = new Float32Array(this.bufferSize);
-    this.writeIndex = 0;
-    this.readIndex = 0;
-    this.samplesAvailable = 0;
-    this.isPlaying = false;
-    
-    this.port.onmessage = (event) => {
-      const { type, data } = event.data;
-      
-      switch (type) {
-        case 'audio':
-          const audioData = data instanceof Float32Array ? data : new Float32Array(data);
-          this.writeAudio(audioData);
-          break;
-        case 'clear':
-          this.clearBuffer();
-          break;
-      }
-    };
-  }
-
-  writeAudio(samples) {
-    if (!samples || samples.length === 0) return;
-    
-    const samplesToWrite = samples.length;
-    
-    if (this.samplesAvailable + samplesToWrite > this.bufferSize) {
-      const overflow = (this.samplesAvailable + samplesToWrite) - this.bufferSize;
-      this.readIndex = (this.readIndex + overflow) % this.bufferSize;
-      this.samplesAvailable -= overflow;
-    }
-    
-    for (let i = 0; i < samplesToWrite; i++) {
-      this.buffer[this.writeIndex] = samples[i];
-      this.writeIndex = (this.writeIndex + 1) % this.bufferSize;
-    }
-    
-    this.samplesAvailable += samplesToWrite;
-    this.isPlaying = true;
-  }
-
-  clearBuffer() {
-    this.readIndex = 0;
-    this.writeIndex = 0;
-    this.samplesAvailable = 0;
-    this.isPlaying = false;
-    this.port.postMessage({ type: 'cleared' });
-  }
-
-  process(inputs, outputs) {
-    const output = outputs[0];
-    if (!output || !output[0]) return true;
-    
-    const outputChannel = output[0];
-    const samplesToRead = outputChannel.length;
-    
-    if (this.samplesAvailable >= samplesToRead) {
-      for (let i = 0; i < samplesToRead; i++) {
-        outputChannel[i] = this.buffer[this.readIndex];
-        this.readIndex = (this.readIndex + 1) % this.bufferSize;
-      }
-      this.samplesAvailable -= samplesToRead;
-    } else if (this.samplesAvailable > 0) {
-      let i = 0;
-      while (this.samplesAvailable > 0 && i < samplesToRead) {
-        outputChannel[i] = this.buffer[this.readIndex];
-        this.readIndex = (this.readIndex + 1) % this.bufferSize;
-        this.samplesAvailable--;
-        i++;
-      }
-      while (i < samplesToRead) {
-        outputChannel[i] = 0;
-        i++;
-      }
-      
-      if (this.isPlaying) {
-        this.isPlaying = false;
-        this.port.postMessage({ type: 'bufferEmpty' });
-      }
-    } else {
-      for (let i = 0; i < samplesToRead; i++) {
-        outputChannel[i] = 0;
-      }
-      this.isPlaying = false;
-    }
-    
-    return true;
-  }
-}
-
-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}/${this.config.contextId}`:e,s=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!s.ok){const o=await s.json();throw new Error(o.error?.message||o.message||"Failed to fetch voice config")}const{data:i}=await s.json();return i}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e={context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"};this.config.systemPrompt&&(e.system_prompt=this.config.systemPrompt),this.config.deepgramConfig&&(e.deepgram_config=this.config.deepgramConfig);const t=await fetch(`${this.unityUrl}/api/unity/voice/auth`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify(e)});if(!t.ok){const a=await t.json();throw new Error(a.error?.message||a.message||"Failed to authenticate voice session")}const{data:s}=await t.json(),{token:i,agent_config:o,deepgram_config:n}=s;this.setSessionContext({clientUuid:o?.client_uuid,contextId:this.config.contextId,contextType:this.config.contextType,currentJobUuid:o?.job_uuid});const d=this.config.systemPrompt||o.instructions||this.serverConfig?.system_prompt||"You are a helpful AI assistant.";await this.initializeAudio();let u="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 a=new URLSearchParams;r.tags.forEach(h=>a.append("tag",h)),u+=`?${a.toString()}`}this.ws=new WebSocket(u,["bearer",i]),this.ws.onopen=()=>{const a=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"}}},h={type:"Settings",audio:{input:{encoding:"linear16",sample_rate:this.inputSampleRate},output:{encoding:"linear16",sample_rate:this.outputSampleRate,container:"none"}},agent:{language:"en",speak:a.speak||{provider:{type:"deepgram",model:"aura-2-thalia-en"}},listen:a.listen||{provider:{type:"deepgram",version:"v2",model:"flux-general-en"}},think:{provider:a.think?.provider||{type:"open_ai",model:"gpt-4.1-nano"},...a.think?.functions&&{functions:a.think.functions}},greeting:"Hi! I'm excited you chose to speak with me. Are you ready to start?"}};a.tags&&a.tags.length>0&&(h.tags=a.tags),this.ws.send(JSON.stringify(h)),this.emit({type:"open",payload:{config:o,serverConfig:this.serverConfig}})};const g=d;this.ws.onmessage=a=>{if(typeof a.data=="string"){try{if(JSON.parse(a.data).type==="SettingsApplied"){const y={type:"UpdatePrompt",prompt:g};this.ws.send(JSON.stringify(y)),this.startMicrophone()}}catch{}this.handleTextMessage(a.data)}else a.data instanceof Blob?this.handleAudioData(a.data):a.data instanceof ArrayBuffer&&this.handleAudioBuffer(a.data)},this.ws.onerror=a=>{console.error("[GlydeVoice] WebSocket error:",a),this.emit({type:"error",payload:a})},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=s=>{const{type:i}=s.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 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;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 s=t-t%2;if(s===0)return;const i=s===t?e:e.slice(0,s),o=new Int16Array(i),n=new Float32Array(o.length);for(let r=0;r<o.length;r++)n[r]=o[r]/32768;const d=this.resample24kTo48k(n);!this.isAgentSpeaking&&!this.agentAudioDoneReceived&&(this.isAgentSpeaking=!0,this.emit({type:"agent_speaking",payload:!0}));const u=new Float32Array(d);this.playbackWorkletNode.port.postMessage({type:"audio",data:u},[u.buffer])}resample24kTo48k(e){const t=e.length*2,s=new Float32Array(t);for(let o=0;o<e.length-1;o++){const n=e[o],d=e[o+1];s[o*2]=n,s[o*2+1]=(n+d)/2}const i=e.length-1;return s[i*2]=e[i],s[i*2+1]=e[i],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=`
-        <div style="padding: 20px; border: 1px solid #ccc; border-radius: 8px; background: #fff;">
-          <h3>Glyde Voice Agent</h3>
-          <p>Status: Active</p>
-          <p>Context: ${this.config.contextType}</p>
-          <button onclick="this.closest('div').remove()">Close</button>
-        </div>
-      `)}async handleFunctionCallRequest(e){for(const t of e.functions){console.log("[GlydeVoice] Function call request:",t.name,t.arguments);let s={};try{s=t.arguments?JSON.parse(t.arguments):{}}catch(n){console.warn("[GlydeVoice] Failed to parse function arguments:",n)}let i;try{t.name==="end_conversation"?i=await this.handleEndConversation(s):i=await this.executeVoiceFunction(t.name,t.id,s)}catch(n){console.error("[GlydeVoice] Function call error:",n),i=JSON.stringify({error:"Function execution failed",details:n instanceof Error?n.message:String(n)})}const o={type:"FunctionCallResponse",id:t.id,name:t.name,content:i};this.ws&&this.ws.readyState===WebSocket.OPEN?(this.ws.send(JSON.stringify(o)),console.log("[GlydeVoice] Function response sent:",t.name)):console.error("[GlydeVoice] Cannot send function response - WebSocket not open")}}async executeVoiceFunction(e,t,s){console.log("[GlydeVoice] Executing voice function via Unity API:",e);try{const i=await fetch(`${this.unityUrl}/api/unity/voice/function`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify({function_name:e,function_call_id:t,input:s,context:{context_id:this.sessionContext.contextId,context_type:this.sessionContext.contextType,current_job_uuid:this.sessionContext.currentJobUuid}})});if(!i.ok){const n=await i.json().catch(()=>({}));throw new Error(n.error?.message||`Function call failed: ${i.status}`)}const o=await i.json();if(o.success&&o.data?.output)return typeof o.data.output=="string"?o.data.output:JSON.stringify(o.data.output);throw new Error("Invalid response from voice function endpoint")}catch(i){return console.error("[GlydeVoice] Voice function error:",i),JSON.stringify({success:!1,error:i instanceof Error?i.message:"Function execution failed",fallback_message:"I apologize, but I'm having trouble with that request right now. Is there something else I can help you with?"})}}async handleEndConversation(e){const t=e.item||"user request";return console.log(`[GlydeVoice] End conversation triggered by: ${t}`),setTimeout(()=>{this.stop()},2e3),JSON.stringify({success:!0,message:"Conversation ending. Say goodbye to the user.",trigger_phrase:t})}setSessionContext(e){this.sessionContext={...this.sessionContext,...e},console.log("[GlydeVoice] Session context updated:",{hasContextId:!!e.contextId,contextType:e.contextType,hasJobUuid:!!e.currentJobUuid})}}c.GlydeVoice=f,Object.defineProperty(c,Symbol.toStringTag,{value:"Module"})}));