@absolutejs/voice 0.0.22-beta.623 → 0.0.22-beta.625

package/dist/core/types.d.ts

@@ -810,6 +810,21 @@ export type VoicePluginConfig<TContext = unknown, TSession extends VoiceSessionR
         }) => string | Promise<string>);
         reason?: string;
     };
+    /** Gentle re-engagement before a silent call is given up on. When the caller
+     *  makes no progress (no committed turn / user partial) for `afterMs`, the
+     *  assistant speaks `line` ("still there?") instead of ending, and the close
+     *  deadline (stuckCallClose) is pushed out — up to `maxReprompts` times (default
+     *  1). After the budget is spent the normal close path runs. Caller speech
+     *  resets the budget. Assistant speech does NOT, so this can only extend a
+     *  bounded amount. `afterMs` should be < stuckCallClose.afterMs so the nudge
+     *  fires first. */
+    idleReprompt?: {
+        afterMs: number;
+        line: string | ((input: {
+            session: TSession;
+        }) => string | Promise<string>);
+        maxReprompts?: number;
+    };
     languageStrategy?: VoiceLanguageStrategy;
     lexicon?: VoiceLexiconEntry[] | VoiceLexiconResolver<TContext>;
     phraseHints?: VoicePhraseHint[] | VoicePhraseHintResolver<TContext>;
@@ -979,6 +994,21 @@ export type CreateVoiceSessionOptions<TContext = unknown, TSession extends Voice
         }) => string | Promise<string>);
         reason?: string;
     };
+    /** Gentle re-engagement before a silent call is given up on. When the caller
+     *  makes no progress (no committed turn / user partial) for `afterMs`, the
+     *  assistant speaks `line` ("still there?") instead of ending, and the close
+     *  deadline (stuckCallClose) is pushed out — up to `maxReprompts` times (default
+     *  1). After the budget is spent the normal close path runs. Caller speech
+     *  resets the budget. Assistant speech does NOT, so this can only extend a
+     *  bounded amount. `afterMs` should be < stuckCallClose.afterMs so the nudge
+     *  fires first. */
+    idleReprompt?: {
+        afterMs: number;
+        line: string | ((input: {
+            session: TSession;
+        }) => string | Promise<string>);
+        maxReprompts?: number;
+    };
     stt?: STTAdapter;
     realtime?: RealtimeAdapter;
     realtimeInputFormat?: AudioFormat;

package/dist/index.js

@@ -4231,6 +4231,45 @@ var createVoiceSession = (options) => {
     clearStuckCloseWatchdog();
     stuckCloseWatchdog = setTimeout(fireStuckClose, stuckCloseAfterMs);
   };
+  const idleRepromptConfig = options.idleReprompt;
+  const idleRepromptAfterMs = idleRepromptConfig && idleRepromptConfig.afterMs > 0 ? idleRepromptConfig.afterMs : undefined;
+  const idleRepromptMax = Math.max(1, idleRepromptConfig?.maxReprompts ?? 1);
+  let idleRepromptsUsed = 0;
+  let idleRepromptWatchdog = null;
+  const clearIdleRepromptWatchdog = () => {
+    if (idleRepromptWatchdog) {
+      clearTimeout(idleRepromptWatchdog);
+      idleRepromptWatchdog = null;
+    }
+  };
+  const kickIdleRepromptWatchdog = (resetBudget = false) => {
+    if (idleRepromptAfterMs === undefined) {
+      return;
+    }
+    if (resetBudget) {
+      idleRepromptsUsed = 0;
+    }
+    clearIdleRepromptWatchdog();
+    if (idleRepromptsUsed >= idleRepromptMax) {
+      return;
+    }
+    idleRepromptWatchdog = setTimeout(() => {
+      idleRepromptWatchdog = null;
+      if (idleRepromptConfig === undefined || idleRepromptsUsed >= idleRepromptMax) {
+        return;
+      }
+      runSerial("idle-reprompt", async () => {
+        const snapshot = await readSession();
+        if (snapshot.status === "completed" || snapshot.status === "failed" || snapshot.call?.endedAt) {
+          return;
+        }
+        idleRepromptsUsed += 1;
+        await speakResolvedLine(idleRepromptConfig.line, snapshot);
+        kickStuckCloseWatchdog();
+        kickIdleRepromptWatchdog();
+      });
+    }, idleRepromptAfterMs);
+  };
   const recordingConfig = options.recording;
   const recordingChannels = new Set(recordingConfig?.channels ?? ["assistant", "user"]);
   const recordingMaxBytes = recordingConfig?.maxBytesPerChannel ?? 50 * 1024 * 1024;
@@ -4728,6 +4767,7 @@ var createVoiceSession = (options) => {
     });
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     await closeTTSSession("failed");
     await closeAdapter("failed");
@@ -4838,6 +4878,7 @@ var createVoiceSession = (options) => {
     });
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     await closeTTSSession("complete");
     await closeAdapter("complete");
@@ -5317,6 +5358,7 @@ var createVoiceSession = (options) => {
   const handlePartial = async (transcript) => {
     if (transcript.text.trim()) {
       kickStuckCloseWatchdog();
+      kickIdleRepromptWatchdog(true);
     }
     if (activeTTSTurnId !== undefined) {
       const triggeringText = transcript.text.trim();
@@ -5793,6 +5835,7 @@ var createVoiceSession = (options) => {
   const completeTurn = async (session, turn) => {
     console.error(`[voice] completeTurn ENTER session=${options.id} turn=${turn.id} textLen=${turn.text?.length ?? 0}`);
     kickStuckCloseWatchdog();
+    kickIdleRepromptWatchdog(true);
     const liveOpsControl = await options.liveOps?.getControl(options.id);
     if (liveOpsControl?.assistantPaused || liveOpsControl?.operatorTakeover) {
       await appendTrace({
@@ -6495,6 +6538,7 @@ var createVoiceSession = (options) => {
     warmTTSSession();
     kickCallSilenceWatchdog();
     kickStuckCloseWatchdog();
+    kickIdleRepromptWatchdog(true);
     startAmdEvaluationTimer();
     if (options.greeting && session.turns.length === 0) {
       await speakResolvedLine(options.greeting, session);
@@ -6641,6 +6685,7 @@ var createVoiceSession = (options) => {
     clearSilenceTimer();
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     if (options.noiseSuppressor?.close) {
       try {
@@ -25442,6 +25487,9 @@ var createTwilioMediaStreamBridge = (socket, options) => {
       stt: options.stt,
       sttFallback: resolveSTTFallbackConfig(options.sttFallback),
       sttLifecycle: options.sttLifecycle ?? runtimePreset.sttLifecycle,
+      sttRecoveryLine: options.sttRecoveryLine,
+      stuckCallClose: options.stuckCallClose,
+      idleReprompt: options.idleReprompt,
       ...options.semanticTurnDetector ? { semanticTurnDetector: options.semanticTurnDetector } : {},
       ...options.bargeInDetector ? { bargeInDetector: options.bargeInDetector } : {},
       ...options.fillerPhrases ? { fillerPhrases: options.fillerPhrases } : {},
@@ -39759,6 +39807,7 @@ var voice = (config) => {
       resumeGreeting: config.resumeGreeting,
       sttRecoveryLine: config.sttRecoveryLine,
       stuckCallClose: config.stuckCallClose,
+      idleReprompt: config.idleReprompt,
       handoff: config.handoff,
       languageStrategy: config.languageStrategy,
       lexicon,

package/dist/testing/index.js

@@ -6551,6 +6551,45 @@ var createVoiceSession = (options) => {
     clearStuckCloseWatchdog();
     stuckCloseWatchdog = setTimeout(fireStuckClose, stuckCloseAfterMs);
   };
+  const idleRepromptConfig = options.idleReprompt;
+  const idleRepromptAfterMs = idleRepromptConfig && idleRepromptConfig.afterMs > 0 ? idleRepromptConfig.afterMs : undefined;
+  const idleRepromptMax = Math.max(1, idleRepromptConfig?.maxReprompts ?? 1);
+  let idleRepromptsUsed = 0;
+  let idleRepromptWatchdog = null;
+  const clearIdleRepromptWatchdog = () => {
+    if (idleRepromptWatchdog) {
+      clearTimeout(idleRepromptWatchdog);
+      idleRepromptWatchdog = null;
+    }
+  };
+  const kickIdleRepromptWatchdog = (resetBudget = false) => {
+    if (idleRepromptAfterMs === undefined) {
+      return;
+    }
+    if (resetBudget) {
+      idleRepromptsUsed = 0;
+    }
+    clearIdleRepromptWatchdog();
+    if (idleRepromptsUsed >= idleRepromptMax) {
+      return;
+    }
+    idleRepromptWatchdog = setTimeout(() => {
+      idleRepromptWatchdog = null;
+      if (idleRepromptConfig === undefined || idleRepromptsUsed >= idleRepromptMax) {
+        return;
+      }
+      runSerial("idle-reprompt", async () => {
+        const snapshot = await readSession();
+        if (snapshot.status === "completed" || snapshot.status === "failed" || snapshot.call?.endedAt) {
+          return;
+        }
+        idleRepromptsUsed += 1;
+        await speakResolvedLine(idleRepromptConfig.line, snapshot);
+        kickStuckCloseWatchdog();
+        kickIdleRepromptWatchdog();
+      });
+    }, idleRepromptAfterMs);
+  };
   const recordingConfig = options.recording;
   const recordingChannels = new Set(recordingConfig?.channels ?? ["assistant", "user"]);
   const recordingMaxBytes = recordingConfig?.maxBytesPerChannel ?? 50 * 1024 * 1024;
@@ -7048,6 +7087,7 @@ var createVoiceSession = (options) => {
     });
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     await closeTTSSession("failed");
     await closeAdapter("failed");
@@ -7158,6 +7198,7 @@ var createVoiceSession = (options) => {
     });
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     await closeTTSSession("complete");
     await closeAdapter("complete");
@@ -7637,6 +7678,7 @@ var createVoiceSession = (options) => {
   const handlePartial = async (transcript) => {
     if (transcript.text.trim()) {
       kickStuckCloseWatchdog();
+      kickIdleRepromptWatchdog(true);
     }
     if (activeTTSTurnId !== undefined) {
       const triggeringText = transcript.text.trim();
@@ -8113,6 +8155,7 @@ var createVoiceSession = (options) => {
   const completeTurn = async (session, turn) => {
     console.error(`[voice] completeTurn ENTER session=${options.id} turn=${turn.id} textLen=${turn.text?.length ?? 0}`);
     kickStuckCloseWatchdog();
+    kickIdleRepromptWatchdog(true);
     const liveOpsControl = await options.liveOps?.getControl(options.id);
     if (liveOpsControl?.assistantPaused || liveOpsControl?.operatorTakeover) {
       await appendTrace({
@@ -8815,6 +8858,7 @@ var createVoiceSession = (options) => {
     warmTTSSession();
     kickCallSilenceWatchdog();
     kickStuckCloseWatchdog();
+    kickIdleRepromptWatchdog(true);
     startAmdEvaluationTimer();
     if (options.greeting && session.turns.length === 0) {
       await speakResolvedLine(options.greeting, session);
@@ -8961,6 +9005,7 @@ var createVoiceSession = (options) => {
     clearSilenceTimer();
     clearCallSilenceWatchdog();
     clearStuckCloseWatchdog();
+    clearIdleRepromptWatchdog();
     clearAmdEvaluationTimer();
     if (options.noiseSuppressor?.close) {
       try {
@@ -14408,6 +14453,9 @@ var createTwilioMediaStreamBridge = (socket, options) => {
       stt: options.stt,
       sttFallback: resolveSTTFallbackConfig(options.sttFallback),
       sttLifecycle: options.sttLifecycle ?? runtimePreset.sttLifecycle,
+      sttRecoveryLine: options.sttRecoveryLine,
+      stuckCallClose: options.stuckCallClose,
+      idleReprompt: options.idleReprompt,
       ...options.semanticTurnDetector ? { semanticTurnDetector: options.semanticTurnDetector } : {},
       ...options.bargeInDetector ? { bargeInDetector: options.bargeInDetector } : {},
       ...options.fillerPhrases ? { fillerPhrases: options.fillerPhrases } : {},

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.623",
+  "version": "0.0.22-beta.625",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",