@glydeunity/voice-sdk 1.3.3 → 1.3.5

package/dist/index.d.ts

@@ -47,12 +47,30 @@ export declare interface DeepgramAgentConfig {
  * Function call request from Deepgram Voice Agent
  * @see https://developers.deepgram.com/docs/voice-agent-function-call-request
  */
+/**
+ * Individual function call within a FunctionCallRequest
+ * @see https://developers.deepgram.com/docs/voice-agent-function-call-request
+ */
+export declare interface FunctionCall {
+    /** Unique identifier for the function call */
+    id: string;
+    /** Name of the function to execute */
+    name: string;
+    /** JSON string containing function arguments */
+    arguments: string;
+    /** If true, client must execute and respond; if false, server handles it */
+    client_side?: boolean;
+}
+
+/**
+ * Function call request message from Deepgram Voice Agent
+ * Contains an array of functions to be executed
+ * @see https://developers.deepgram.com/docs/voice-agent-function-call-request
+ */
 export declare interface FunctionCallRequest {
     type: 'FunctionCallRequest';
-    function_name: string;
-    function_call_id: string;
-    input: Record<string, unknown>;
-    client_side?: boolean;
+    /** Array of function calls to execute */
+    functions: FunctionCall[];
 }
 
 /**
@@ -195,7 +213,10 @@ export declare class GlydeVoice {
      * Handle a function call request from Deepgram Voice Agent
      * Routes function execution through the Unity voice function endpoint for proper authentication
      *
-     * @param request - The function call request from Deepgram
+     * Deepgram sends an array of functions in each request, so we process each one
+     * and send individual responses back.
+     *
+     * @param request - The function call request from Deepgram containing functions array
      * @see https://developers.deepgram.com/docs/voice-agents-function-calling
      */
     private handleFunctionCallRequest;

package/dist/voice-sdk.es.js

@@ -176,15 +176,15 @@ class y {
    * @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, o = await fetch(t, {
+    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 (!o.ok) {
-      const a = await o.json();
-      throw new Error(a.error?.message || a.message || "Failed to fetch voice config");
+    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 o.json();
+    const { data: i } = await s.json();
     return i;
   }
   /**
@@ -206,27 +206,27 @@ class y {
           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 a = await t.json();
+          throw new Error(a.error?.message || a.message || "Failed to authenticate voice session");
         }
-        const { data: o } = await t.json(), { token: i, agent_config: a, deepgram_config: n } = o;
+        const { data: s } = await t.json(), { token: i, agent_config: o, deepgram_config: n } = s;
         this.setSessionContext({
-          clientUuid: a?.client_uuid,
+          clientUuid: o?.client_uuid,
           contextId: this.config.contextId,
           contextType: this.config.contextType,
-          currentJobUuid: a?.job_uuid
+          currentJobUuid: o?.job_uuid
         });
-        const c = this.config.systemPrompt || a.instructions || this.serverConfig?.system_prompt || "You are a helpful AI assistant.";
+        const c = this.config.systemPrompt || o.instructions || this.serverConfig?.system_prompt || "You are a helpful AI assistant.";
         await this.initializeAudio();
         let l = "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((h) => s.append("tag", h)), l += `?${s.toString()}`;
+          const a = new URLSearchParams();
+          r.tags.forEach((h) => a.append("tag", h)), l += `?${a.toString()}`;
         }
         this.ws = new WebSocket(l, ["bearer", i]), this.ws.onopen = () => {
-          const s = r || {
-            think: { provider: { type: "open_ai", model: "gpt-5-nano" } },
+          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 = {
@@ -244,39 +244,39 @@ class y {
             },
             agent: {
               language: "en",
-              speak: s.speak || {
+              speak: a.speak || {
                 provider: { type: "deepgram", model: "aura-2-thalia-en" }
               },
-              listen: s.listen || {
+              listen: a.listen || {
                 provider: { type: "deepgram", version: "v2", model: "flux-general-en" }
               },
               think: {
-                provider: s.think?.provider || { type: "open_ai", model: "gpt-5-nano" },
+                provider: a.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 }
+                ...a.think?.functions && { functions: a.think.functions }
               },
               greeting: "Hi! I'm excited you chose to speak with me. Are you ready to start?"
             }
           };
-          s.tags && s.tags.length > 0 && (h.tags = s.tags), this.ws.send(JSON.stringify(h)), this.emit({ type: "open", payload: { config: a, serverConfig: this.serverConfig } });
+          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 d = c;
-        this.ws.onmessage = (s) => {
-          if (typeof s.data == "string") {
+        const u = c;
+        this.ws.onmessage = (a) => {
+          if (typeof a.data == "string") {
             try {
-              if (JSON.parse(s.data).type === "SettingsApplied") {
-                const u = {
+              if (JSON.parse(a.data).type === "SettingsApplied") {
+                const d = {
                   type: "UpdatePrompt",
-                  prompt: d
+                  prompt: u
                 };
-                this.ws.send(JSON.stringify(u)), this.startMicrophone();
+                this.ws.send(JSON.stringify(d)), 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.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();
@@ -310,8 +310,8 @@ class y {
     } 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;
+    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 }));
     };
   }
@@ -335,8 +335,8 @@ class y {
           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);
+            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":
@@ -372,11 +372,11 @@ class y {
     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 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 c = this.resample24kTo48k(n);
     !this.isAgentSpeaking && !this.agentAudioDoneReceived && (this.isAgentSpeaking = !0, this.emit({ type: "agent_speaking", payload: !0 }));
     const l = new Float32Array(c);
@@ -389,13 +389,13 @@ class y {
    * Resample audio from 24kHz to 48kHz using linear interpolation
    */
   resample24kTo48k(e) {
-    const t = e.length * 2, o = new Float32Array(t);
-    for (let a = 0; a < e.length - 1; a++) {
-      const n = e[a], c = e[a + 1];
-      o[a * 2] = n, o[a * 2 + 1] = (n + c) / 2;
+    const t = e.length * 2, s = new Float32Array(t);
+    for (let o = 0; o < e.length - 1; o++) {
+      const n = e[o], c = e[o + 1];
+      s[o * 2] = n, s[o * 2 + 1] = (n + c) / 2;
     }
     const i = e.length - 1;
-    return o[i * 2] = e[i], o[i * 2 + 1] = e[i], o;
+    return s[i * 2] = e[i], s[i * 2 + 1] = e[i], s;
   }
   /**
    * Clear the playback buffer (for interruption handling)
@@ -506,26 +506,37 @@ class y {
    * Handle a function call request from Deepgram Voice Agent
    * Routes function execution through the Unity voice function endpoint for proper authentication
    * 
-   * @param request - The function call request from Deepgram
+   * Deepgram sends an array of functions in each request, so we process each one
+   * and send individual responses back.
+   * 
+   * @param request - The function call request from Deepgram containing functions array
    * @see https://developers.deepgram.com/docs/voice-agents-function-calling
    */
   async handleFunctionCallRequest(e) {
-    console.log("[GlydeVoice] Function call request:", e.function_name, e.input);
-    let t;
-    try {
-      e.function_name === "end_conversation" ? t = await this.handleEndConversation(e.input) : t = await this.executeVoiceFunction(e.function_name, e.function_call_id, e.input);
-    } catch (i) {
-      console.error("[GlydeVoice] Function call error:", i), t = JSON.stringify({
-        error: "Function execution failed",
-        details: i instanceof Error ? i.message : String(i)
-      });
+    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",
+        function_call_id: t.id,
+        output: 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");
     }
-    const o = {
-      type: "FunctionCallResponse",
-      function_call_id: e.function_call_id,
-      output: t
-    };
-    this.ws && this.ws.readyState === WebSocket.OPEN ? (this.ws.send(JSON.stringify(o)), console.log("[GlydeVoice] Function response sent:", e.function_name)) : console.error("[GlydeVoice] Cannot send function response - WebSocket not open");
   }
   /**
    * Execute a voice function through the Unity API with proper authentication
@@ -537,7 +548,7 @@ class y {
    * @param input - Function input parameters
    * @returns JSON string with function result
    */
-  async executeVoiceFunction(e, t, o) {
+  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`, {
@@ -546,7 +557,7 @@ class y {
         body: JSON.stringify({
           function_name: e,
           function_call_id: t,
-          input: o,
+          input: s,
           context: {
             context_id: this.sessionContext.contextId,
             context_type: this.sessionContext.contextType,
@@ -558,9 +569,9 @@ class y {
         const n = await i.json().catch(() => ({}));
         throw new Error(n.error?.message || `Function call failed: ${i.status}`);
       }
-      const a = await i.json();
-      if (a.success && a.data?.output)
-        return typeof a.data.output == "string" ? a.data.output : JSON.stringify(a.data.output);
+      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({

package/dist/voice-sdk.umd.js

@@ -130,11 +130,11 @@ 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}/${this.config.contextId}`:e,s=await fetch(t,{method:"GET",headers:this.getAuthHeaders()});if(!s.ok){const o=await s.json();throw new Error(o.error?.message||o.message||"Failed to fetch voice config")}const{data:i}=await s.json();return i}async start(){if(!this.active){this.active=!0;try{this.config.systemPrompt||(this.serverConfig=await this.fetchConfig(),console.log("[GlydeVoice] Fetched config:",this.serverConfig));const e={context_id:this.config.contextId,domain:typeof window<"u"?window.location.hostname:"localhost"};this.config.systemPrompt&&(e.system_prompt=this.config.systemPrompt),this.config.deepgramConfig&&(e.deepgram_config=this.config.deepgramConfig);const t=await fetch(`${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: ${this.config.contextType}</p>
           <button onclick="this.closest('div').remove()">Close</button>
         </div>
-      `)}async handleFunctionCallRequest(e){console.log("[GlydeVoice] Function call request:",e.function_name,e.input);let t;try{e.function_name==="end_conversation"?t=await this.handleEndConversation(e.input):t=await this.executeVoiceFunction(e.function_name,e.function_call_id,e.input)}catch(i){console.error("[GlydeVoice] Function call error:",i),t=JSON.stringify({error:"Function execution failed",details:i instanceof Error?i.message:String(i)})}const o={type:"FunctionCallResponse",function_call_id:e.function_call_id,output:t};this.ws&&this.ws.readyState===WebSocket.OPEN?(this.ws.send(JSON.stringify(o)),console.log("[GlydeVoice] Function response sent:",e.function_name)):console.error("[GlydeVoice] Cannot send function response - WebSocket not open")}async executeVoiceFunction(e,t,o){console.log("[GlydeVoice] Executing voice function via Unity API:",e);try{const i=await fetch(`${this.unityUrl}/api/unity/voice/function`,{method:"POST",headers:this.getAuthHeaders(),body:JSON.stringify({function_name:e,function_call_id:t,input:o,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 a=await i.json();if(a.success&&a.data?.output)return typeof a.data.output=="string"?a.data.output:JSON.stringify(a.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: ${t}`),setTimeout(()=>{this.stop()},2e3),JSON.stringify({success:!0,message:"Conversation ending. Say goodbye to the user.",trigger_phrase:t})}setSessionContext(e){this.sessionContext={...this.sessionContext,...e},console.log("[GlydeVoice] Session context updated:",{hasContextId:!!e.contextId,contextType:e.contextType,hasJobUuid:!!e.currentJobUuid})}}c.GlydeVoice=f,Object.defineProperty(c,Symbol.toStringTag,{value:"Module"})}));
+      `)}async handleFunctionCallRequest(e){for(const t of e.functions){console.log("[GlydeVoice] Function call request:",t.name,t.arguments);let s={};try{s=t.arguments?JSON.parse(t.arguments):{}}catch(n){console.warn("[GlydeVoice] Failed to parse function arguments:",n)}let i;try{t.name==="end_conversation"?i=await this.handleEndConversation(s):i=await this.executeVoiceFunction(t.name,t.id,s)}catch(n){console.error("[GlydeVoice] Function call error:",n),i=JSON.stringify({error:"Function execution failed",details:n instanceof Error?n.message:String(n)})}const o={type:"FunctionCallResponse",function_call_id:t.id,output:i};this.ws&&this.ws.readyState===WebSocket.OPEN?(this.ws.send(JSON.stringify(o)),console.log("[GlydeVoice] Function response sent:",t.name)):console.error("[GlydeVoice] Cannot send function response - WebSocket not open")}}async executeVoiceFunction(e,t,s){console.log("[GlydeVoice] Executing voice function via Unity API:",e);try{const i=await fetch(`${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"})}));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@glydeunity/voice-sdk",
-  "version": "1.3.3",
+  "version": "1.3.5",
   "description": "GLYDE Voice Agent SDK - AI-powered voice interactions for web applications",
   "type": "module",
   "main": "./dist/voice-sdk.umd.js",