Update app.js

app.js

@@ -25,7 +25,6 @@ const log = (tag, ...args) => {
 let callLogs = [];
 
 // ─── UTILITY: GLOBAL TELNYX COMMAND DISPATCHER ───
-// STRICTLY FIXED: https://api.telnyx.com/v2/calls/...
 const sendTelnyxCmd = async (callId, action, body = {}) => {
   try {
     const response = await fetch(`https://api.telnyx.com/v2/calls/${callId}/actions/${action}`, {
@@ -214,6 +213,7 @@ app.get("/health", (req, res) => res.json({ status: "ok", uptime: process.uptime
 app.get("/play-recording/:call_control_id", async (req, res) => {
   const logEntry = callLogs.find(l => l.call_control_id === req.params.call_control_id);
   if (!logEntry || !logEntry.raw_telnyx_url) return res.sendStatus(404);
+
   try {
     const s3Res = await fetch(logEntry.raw_telnyx_url);
     if (!s3Res.ok) return res.sendStatus(s3Res.status);
@@ -229,7 +229,8 @@ app.get("/play-recording/:call_control_id", async (req, res) => {
 // ─── DIAL OUT ROUTE ────────────────────────────────────────
 app.post("/dial", async (req, res) => {
   const { to } = req.body;
-  const voice = "Algieba"; 
+  // SETTING ENCELADUS AS THE DEFAULT VOICE
+  const voice = "Enceladus"; 
   log("DIAL", `Dialing ${to}...`);
 
   try {
@@ -306,7 +307,7 @@ app.post("/webhook/call", async (req, res) => {
 
         callLogs.push({
           id: randomUUID(), call_control_id: callId, target_number: payload?.from,
-          voice_used: "Algieba", status: "in-progress", last_transcript: "", recording_url: null,
+          voice_used: "Enceladus", status: "in-progress", last_transcript: "", recording_url: null,
           raw_telnyx_url: null, is_playing: false, timers: {}, created_at: new Date().toISOString()
         });
         await sendTelnyxCmd(callId, "answer");
@@ -319,7 +320,6 @@ app.post("/webhook/call", async (req, res) => {
 
       await sendTelnyxCmd(callId, "record_start", { format: "mp3", channels: "dual", play_beep: false });
       
-      // Open WebSocket purely for the VAD staller
       const wsUrl = `wss://${req.headers.host}/audio-stream`;
       log("STREAM", "Connecting WebSocket for early turn detection...");
       await sendTelnyxCmd(callId, "streaming_start", { 
@@ -327,7 +327,6 @@ app.post("/webhook/call", async (req, res) => {
         stream_track: "inbound_track" 
       });
 
-      // Start Native STT to handle the actual text
       log("STT", "Activating native Telnyx real-time transcription engine...");
       await sendTelnyxCmd(callId, "transcription_start", {
         transcription_engine: "Telnyx",
@@ -352,7 +351,7 @@ app.post("/webhook/call", async (req, res) => {
       const data = payload?.transcription_data;
       if (!data) break;
 
-      const voice = entry ? entry.voice_used : "Algieba";
+      const voice = entry ? entry.voice_used : "Enceladus";
 
       // 1. BARGE-IN (User is currently speaking)
       if (!data.is_final) {
@@ -371,7 +370,7 @@ app.post("/webhook/call", async (req, res) => {
         log("TRANSCRIPT", `User said: "${transcript}"`);
         updateLog(callId, { last_transcript: transcript });
 
-        // 3. FETCH FULL AI RESPONSE
+        // FETCH FULL AI RESPONSE
         try {
           const aiRes = await fetch(`${AI_SERVER_URL}/api/process-text-turn`, {
             method: "POST",
@@ -382,7 +381,7 @@ app.post("/webhook/call", async (req, res) => {
           if (aiRes.ok) {
             const aiData = await aiRes.json();
             if (aiData.speak_url && entry && entry.status !== "completed") {
-              log("AI_REPLY", `Playing full reply.`);
+              log("AI_REPLY", `Playing full reply: ${aiData.raw_text}`);
               updateLog(callId, { is_playing: true });
               await sendTelnyxCmd(callId, "playback_start", {
                 audio_url: aiData.speak_url,
@@ -414,9 +413,9 @@ app.post("/webhook/call", async (req, res) => {
   res.sendStatus(200);
 });
 
-// ─── WEBSOCKET (Strictly for Early Staller & Barge-In) ───
+// ─── WEBSOCKET (VAD & Early Staller) ───
 wss.on("connection", (ws, req) => {
-  log("WS", `VAD WebSocket connected from ${req.socket.remoteAddress}`);
+  log("WS", `VAD WebSocket connected`);
   
   let isSpeaking = false;
   let silenceFrames = 0;
@@ -425,8 +424,9 @@ wss.on("connection", (ws, req) => {
   let hasGreeted = false;
   let totalNoiseFrames = 0;
 
-  const SILENCE_THRESHOLD_FRAMES = 25; 
-  const VOLUME_THRESHOLD = 500; 
+  // INCREASED THRESHOLDS FOR A MORE PATIENT AI
+  const SILENCE_THRESHOLD_FRAMES = 37; // ~750ms of silence needed before declaring a turn end
+  const VOLUME_THRESHOLD = 500; // Require louder noise to prevent false barge-ins
 
   ws.on("message", async (raw) => {
     try {
@@ -434,14 +434,13 @@ wss.on("connection", (ws, req) => {
 
       if (msg.event === "start") {
         callControlId = msg.start?.call_control_id;
-        log("WS_STREAM", `Receiving audio for call ${callControlId}`);
         
         // Smart Initial Greeting
         setTimeout(async () => {
           if (!hasGreeted && callControlId) {
              const entry = callLogs.find(l => l.call_control_id === callControlId);
              if (entry && entry.status === "in-progress") {
-               const voice = entry.voice_used || "Algieba";
+               const voice = entry.voice_used || "Enceladus";
                log("SMART_GREETING", "No immediate human audio detected. Saying Hello.");
                entry.is_playing = true;
                hasGreeted = true;
@@ -459,7 +458,6 @@ wss.on("connection", (ws, req) => {
         const pcm16 = decodeMuLaw(mulawBuffer);
         const rms = calculateRMS(pcm16);
 
-        // ── 1. NOISE DETECTED (Human is speaking) ──
         if (rms > VOLUME_THRESHOLD) {
           hasGreeted = true; 
           silenceFrames = 0;
@@ -469,7 +467,6 @@ wss.on("connection", (ws, req) => {
             isSpeaking = true;
             hasPlayedFiller = false;
             
-            // BARGE IN: If AI is talking, tell it to shut up
             const entry = callLogs.find(l => l.call_control_id === callControlId);
             if (entry && entry.is_playing) {
               log("BARGE-IN", "User is speaking. Interrupting AI playback.");
@@ -478,21 +475,20 @@ wss.on("connection", (ws, req) => {
             }
           }
         } 
-        
-        // ── 2. SILENCE DETECTED ──
         else if (isSpeaking) {
           silenceFrames++;
 
-          // ── 3. EARLY STALL TRIGGER (500ms silence) ──
           if (silenceFrames > SILENCE_THRESHOLD_FRAMES && !hasPlayedFiller) {
             hasPlayedFiller = true;
             isSpeaking = false; 
             
-            // Only play filler if the human actually spoke a sentence
-            if (totalNoiseFrames > 15) {
-              log("VAD_EARLY_TURN", "500ms silence detected. Playing early staller...");
+            // LAZY BACKCHANNELING:
+            // Only play a filler word if they spoke continuously for ~800ms
+            // This prevents it from saying "Okay" to an IVR breath or a short "Yeah".
+            if (totalNoiseFrames > 40) {
+              log("VAD_EARLY_TURN", "Long sentence detected. Playing early staller...");
               const entry = callLogs.find(l => l.call_control_id === callControlId);
-              const voice = entry ? entry.voice_used : "Algieba";
+              const voice = entry ? entry.voice_used : "Enceladus";
 
               const fillers = ["Right.", "Got it.", "Okay.", "Yeah.", "Ah, okay."];
               const filler = fillers[Math.floor(Math.random() * fillers.length)];
@@ -517,5 +513,5 @@ requiredEnv.forEach((key) => {
 });
 
 server.listen(PORT, "0.0.0.0", () => {
-  log("STARTUP", `Telecom Server running on port ${PORT}`);
+  log("STARTUP", `Elite Telecom Server running on port ${PORT}`);
 });