@absolutejs/voice 0.0.22-beta.553 → 0.0.22-beta.554

package/dist/core/types.d.ts

@@ -916,6 +916,13 @@ export type CreateVoiceSessionOptions<TContext = unknown, TSession extends Voice
     fillerPhrases?: ReadonlyArray<string>;
     /** Milliseconds after turn-commit before the filler fires. Default 250ms — short enough to feel instant, long enough to skip if the LLM is very fast. */
     fillerDelayMs?: number;
+    /**
+     * Default spoken ack if the model returns ONLY tool calls (no text) and the
+     * turn isn't ending. Without this, the caller hears total silence after
+     * their turn and assumes the line dropped. Default is "Sorry, one moment."
+     * Set to "" to opt out entirely.
+     */
+    defaultSilentTurnAck?: string;
     assistantMode?: import("./assistantMode").VoiceAssistantMode;
     modalities?: ReadonlyArray<"audio" | "text">;
     prosody?: VoiceTTSProsody;

package/dist/index.js

@@ -5464,6 +5464,53 @@ var createVoiceSession = (options) => {
         });
       }
     }
+    const audioWasSent = Boolean(streamResult?.streamed) || Boolean(output?.assistantText?.trim());
+    const turnIsEnding = Boolean(output?.complete) || Boolean(output?.transfer) || Boolean(output?.escalate) || Boolean(output?.voicemail) || Boolean(output?.noAnswer);
+    if (!audioWasSent && !turnIsEnding) {
+      const fallback = typeof options.defaultSilentTurnAck === "string" ? options.defaultSilentTurnAck : "Sorry, one moment.";
+      if (fallback.trim() && options.tts) {
+        try {
+          const activeTTSSession = await ensureTTSSession();
+          if (activeTTSSession) {
+            fillerToken += 1;
+            if (fillerTimer) {
+              clearTimeout(fillerTimer);
+              fillerTimer = null;
+            }
+            if (fillerActive) {
+              await cancelActiveTTS("filler-superseded").catch(() => {});
+              fillerActive = false;
+            }
+            activeTTSTurnId = turn.id;
+            await activeTTSSession.send(fallback);
+            await appendTrace({
+              payload: {
+                assistantMode: resolveVoiceAssistantMode(options),
+                fallback: true,
+                realtimeConfigured: Boolean(options.realtime),
+                reason: "model-returned-no-text",
+                text: fallback,
+                ttsConfigured: Boolean(options.tts)
+              },
+              session,
+              turnId: turn.id,
+              type: "turn.assistant"
+            });
+            if (options.costAccountant) {
+              options.costAccountant.recordTTS({
+                characters: fallback.length
+              });
+            }
+          }
+        } catch (error) {
+          logger.warn("voice default-silent-turn-ack fallback send failed", {
+            error: toError(error).message,
+            sessionId: options.id,
+            turnId: turn.id
+          });
+        }
+      }
+    }
     if (output?.result !== undefined) {
       await writeSession((currentSession) => {
         setTurnResult(currentSession, turn.id, {
@@ -24608,6 +24655,7 @@ var createTwilioMediaStreamBridge = (socket, options) => {
       ...options.semanticTurnDetector ? { semanticTurnDetector: options.semanticTurnDetector } : {},
       ...options.fillerPhrases ? { fillerPhrases: options.fillerPhrases } : {},
       ...options.fillerDelayMs !== undefined ? { fillerDelayMs: options.fillerDelayMs } : {},
+      ...options.defaultSilentTurnAck !== undefined ? { defaultSilentTurnAck: options.defaultSilentTurnAck } : {},
       trace: options.trace,
       tts: options.tts,
       turnDetection

package/dist/telephony/twilio.d.ts

@@ -132,6 +132,13 @@ export type TwilioMediaStreamBridgeOptions<TContext = unknown, TSession extends
     fillerPhrases?: ReadonlyArray<string>;
     /** Milliseconds after turn-commit before the filler fires. Default 250ms. */
     fillerDelayMs?: number;
+    /**
+     * Default spoken ack if the model returns ONLY tool calls (no text) and
+     * the turn isn't ending. Without this, the caller hears silence and
+     * assumes the line dropped. Default "Sorry, one moment." — set to ""
+     * to opt out. See CreateVoiceSessionOptions for full semantics.
+     */
+    defaultSilentTurnAck?: string;
 };
 export type TwilioMediaStreamBridge = {
     close: (reason?: string) => Promise<void>;

package/dist/testing/index.js

@@ -7281,6 +7281,53 @@ var createVoiceSession = (options) => {
         });
       }
     }
+    const audioWasSent = Boolean(streamResult?.streamed) || Boolean(output?.assistantText?.trim());
+    const turnIsEnding = Boolean(output?.complete) || Boolean(output?.transfer) || Boolean(output?.escalate) || Boolean(output?.voicemail) || Boolean(output?.noAnswer);
+    if (!audioWasSent && !turnIsEnding) {
+      const fallback = typeof options.defaultSilentTurnAck === "string" ? options.defaultSilentTurnAck : "Sorry, one moment.";
+      if (fallback.trim() && options.tts) {
+        try {
+          const activeTTSSession = await ensureTTSSession();
+          if (activeTTSSession) {
+            fillerToken += 1;
+            if (fillerTimer) {
+              clearTimeout(fillerTimer);
+              fillerTimer = null;
+            }
+            if (fillerActive) {
+              await cancelActiveTTS("filler-superseded").catch(() => {});
+              fillerActive = false;
+            }
+            activeTTSTurnId = turn.id;
+            await activeTTSSession.send(fallback);
+            await appendTrace({
+              payload: {
+                assistantMode: resolveVoiceAssistantMode(options),
+                fallback: true,
+                realtimeConfigured: Boolean(options.realtime),
+                reason: "model-returned-no-text",
+                text: fallback,
+                ttsConfigured: Boolean(options.tts)
+              },
+              session,
+              turnId: turn.id,
+              type: "turn.assistant"
+            });
+            if (options.costAccountant) {
+              options.costAccountant.recordTTS({
+                characters: fallback.length
+              });
+            }
+          }
+        } catch (error) {
+          logger.warn("voice default-silent-turn-ack fallback send failed", {
+            error: toError(error).message,
+            sessionId: options.id,
+            turnId: turn.id
+          });
+        }
+      }
+    }
     if (output?.result !== undefined) {
       await writeSession((currentSession) => {
         setTurnResult(currentSession, turn.id, {
@@ -13144,6 +13191,7 @@ var createTwilioMediaStreamBridge = (socket, options) => {
       ...options.semanticTurnDetector ? { semanticTurnDetector: options.semanticTurnDetector } : {},
       ...options.fillerPhrases ? { fillerPhrases: options.fillerPhrases } : {},
       ...options.fillerDelayMs !== undefined ? { fillerDelayMs: options.fillerDelayMs } : {},
+      ...options.defaultSilentTurnAck !== undefined ? { defaultSilentTurnAck: options.defaultSilentTurnAck } : {},
       trace: options.trace,
       tts: options.tts,
       turnDetection

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@absolutejs/voice",
-	"version": "0.0.22-beta.553",
+	"version": "0.0.22-beta.554",
 	"description": "Voice primitives and Elysia plugin for AbsoluteJS",
 	"repository": {
 		"type": "git",