From 759c7fc18e4883a1068944793c0ea4bb310b119d Mon Sep 17 00:00:00 2001
From: Sebastian <19554889+sebslight@users.noreply.github.com>
Date: Mon, 16 Feb 2026 22:32:42 -0500
Subject: [PATCH] revert(voice-call): remove cached inbound greeting

---
 extensions/voice-call/src/providers/twilio.ts | 11 -----
 extensions/voice-call/src/runtime.ts          | 29 --------------
 extensions/voice-call/src/webhook.ts          | 40 ++++---------------
 3 files changed, 7 insertions(+), 73 deletions(-)

diff --git a/extensions/voice-call/src/providers/twilio.ts b/extensions/voice-call/src/providers/twilio.ts
index 43af8e299a3..245c5e2bc3b 100644
--- a/extensions/voice-call/src/providers/twilio.ts
+++ b/extensions/voice-call/src/providers/twilio.ts
@@ -62,17 +62,6 @@ export class TwilioProvider implements VoiceCallProvider {
 
   /** Map of call SID to stream SID for media streams */
   private callStreamMap = new Map<string, string>();
-  /** Pre-generated greeting audio for instant inbound playback */
-  private cachedGreetingAudio: Buffer | null = null;
-
-  setCachedGreetingAudio(audio: Buffer): void {
-    this.cachedGreetingAudio = audio;
-    console.log(`[voice-call] Cached greeting audio: ${audio.length} bytes`);
-  }
-
-  getCachedGreetingAudio(): Buffer | null {
-    return this.cachedGreetingAudio;
-  }
   /** Per-call tokens for media stream authentication */
   private streamAuthTokens = new Map<string, string>();
 
diff --git a/extensions/voice-call/src/runtime.ts b/extensions/voice-call/src/runtime.ts
index 72365b9ccb5..811a9074037 100644
--- a/extensions/voice-call/src/runtime.ts
+++ b/extensions/voice-call/src/runtime.ts
@@ -187,35 +187,6 @@ export async function createVoiceCallRuntime(params: {
       twilioProvider.setMediaStreamHandler(mediaHandler);
       log.info("[voice-call] Media stream handler wired to provider");
     }
-
-    // Pre-cache inbound greeting TTS for instant playback on connect
-    if (config.inboundGreeting && ttsRuntime?.textToSpeechTelephony) {
-      try {
-        const greetingTts = createTelephonyTtsProvider({
-          coreConfig,
-          ttsOverride: config.tts,
-          runtime: ttsRuntime,
-        });
-        greetingTts
-          .synthesizeForTelephony(config.inboundGreeting)
-          .then((audio) => {
-            twilioProvider.setCachedGreetingAudio(audio);
-          })
-          .catch((err) => {
-            log.warn(
-              `[voice-call] Failed to pre-cache greeting: ${
-                err instanceof Error ? err.message : String(err)
-              }`,
-            );
-          });
-      } catch (err) {
-        log.warn(
-          `[voice-call] Failed to init greeting TTS: ${
-            err instanceof Error ? err.message : String(err)
-          }`,
-        );
-      }
-    }
   }
 
   manager.initialize(provider, webhookUrl);
diff --git a/extensions/voice-call/src/webhook.ts b/extensions/voice-call/src/webhook.ts
index aa43f218396..4574c77bcb4 100644
--- a/extensions/voice-call/src/webhook.ts
+++ b/extensions/voice-call/src/webhook.ts
@@ -142,39 +142,13 @@ export class VoiceCallWebhookServer {
           (this.provider as TwilioProvider).registerCallStream(callId, streamSid);
         }
 
-        // Try instant cached greeting for inbound calls (pre-generated at startup)
-        const cachedAudio =
-          this.provider.name === "twilio"
-            ? (this.provider as TwilioProvider).getCachedGreetingAudio()
-            : null;
-        const call = this.manager.getCallByProviderCallId(callId);
-        if (cachedAudio && call?.metadata?.initialMessage && call.direction === "inbound") {
-          console.log(`[voice-call] Playing cached greeting (${cachedAudio.length} bytes)`);
-          delete call.metadata.initialMessage; // prevent re-speaking via fallback
-          const handler = this.mediaStreamHandler!;
-          const CHUNK_SIZE = 160;
-          const CHUNK_DELAY_MS = 20;
-          void (async () => {
-            const { chunkAudio } = await import("./telephony-audio.js");
-            await handler.queueTts(streamSid, async (signal) => {
-              for (const chunk of chunkAudio(cachedAudio, CHUNK_SIZE)) {
-                if (signal.aborted) break;
-                handler.sendAudio(streamSid, chunk);
-                await new Promise((r) => setTimeout(r, CHUNK_DELAY_MS));
-              }
-              if (!signal.aborted) {
-                handler.sendMark(streamSid, `greeting-${Date.now()}`);
-              }
-            });
-          })().catch((err) => console.warn("[voice-call] Cached greeting playback failed:", err));
-        } else {
-          // Fallback: original path with reduced delay
-          setTimeout(() => {
-            this.manager.speakInitialMessage(callId).catch((err) => {
-              console.warn(`[voice-call] Failed to speak initial message:`, err);
-            });
-          }, 100);
-        }
+        // Speak initial message if one was provided when call was initiated
+        // Use setTimeout to allow stream setup to complete
+        setTimeout(() => {
+          this.manager.speakInitialMessage(callId).catch((err) => {
+            console.warn(`[voice-call] Failed to speak initial message:`, err);
+          });
+        }, 500);
       },
       onDisconnect: (callId) => {
         console.log(`[voice-call] Media stream disconnected: ${callId}`);