@absolutejs/voice 0.0.22-beta.594 → 0.0.22-beta.596

package/dist/index.js

@@ -3723,6 +3723,8 @@ var EXTENDED_VENDOR_COMMIT_SILENCE_THRESHOLD_MS = 200;
 var MAX_VENDOR_COMMIT_GRACE_MS = 1200;
 var STT_RECONNECT_FLAP_WINDOW_MS = 4000;
 var MAX_STT_RECONNECTS_IN_FLAP_WINDOW = 3;
+var STT_HEALTH_STALE_MS = 6000;
+var STT_HEALTH_SPEECH_GAP_MS = 2000;
 var DEFAULT_FORMAT = {
   channels: 1,
   container: "raw",
@@ -3972,6 +3974,8 @@ var createVoiceSession = (options) => {
   let activeAdapterGeneration = 0;
   let sttReconnectCount = 0;
   let lastSttReconnectAt = 0;
+  let lastSpeechEnergyAt = 0;
+  let sttHealthPhaseStart = 0;
   let activeTTSTurnId;
   let assistantSpeechEndsAt = 0;
   let lastAssistantAudioAt = 0;
@@ -4639,13 +4643,11 @@ var createVoiceSession = (options) => {
     if (!didComplete) {
       return;
     }
-    if (shouldInvokeOnComplete) {
-      await options.route.onComplete({
-        api,
-        context: options.context,
-        session
-      });
-    }
+    const onCompletePromise = shouldInvokeOnComplete ? Promise.resolve(options.route.onComplete({
+      api,
+      context: options.context,
+      session
+    })) : Promise.resolve();
     if (disposition === "completed") {
       await drainAssistantSpeech(lastTtsSendAt);
       await assistantAudioQueue;
@@ -4708,6 +4710,7 @@ var createVoiceSession = (options) => {
         session
       });
     }
+    await onCompletePromise;
     await options.route.onCallEnd?.({
       api,
       context: options.context,
@@ -6304,6 +6307,30 @@ var createVoiceSession = (options) => {
       } else {
         clearSilenceTimer();
       }
+      const nowMs = Date.now();
+      if (nowMs - lastSpeechEnergyAt > STT_HEALTH_SPEECH_GAP_MS) {
+        sttHealthPhaseStart = nowMs;
+      }
+      lastSpeechEnergyAt = nowMs;
+      const lastTranscriptAt = latest.currentTurn.lastTranscriptAt ?? 0;
+      if (!options.realtime && sttSession && lastTranscriptAt < sttHealthPhaseStart && nowMs - sttHealthPhaseStart >= STT_HEALTH_STALE_MS) {
+        sttReconnectCount = nowMs - lastSttReconnectAt < STT_RECONNECT_FLAP_WINDOW_MS ? sttReconnectCount + 1 : 1;
+        lastSttReconnectAt = nowMs;
+        sttHealthPhaseStart = nowMs;
+        if (sttReconnectCount <= MAX_STT_RECONNECTS_IN_FLAP_WINDOW) {
+          await appendTrace({
+            payload: {
+              action: "stt-health-reconnect",
+              attempt: sttReconnectCount,
+              reason: `no transcript for ${STT_HEALTH_STALE_MS}ms of continuous speech`
+            },
+            session: latest,
+            type: "session.error"
+          });
+          await closeAdapter("stt stale; health-reconnect");
+          return;
+        }
+      }
     } else if (speechDetected) {
       backchannelDriver?.noteSilence();
       const currentSession = await readSession();

package/dist/testing/index.js

@@ -5950,6 +5950,8 @@ var EXTENDED_VENDOR_COMMIT_SILENCE_THRESHOLD_MS = 200;
 var MAX_VENDOR_COMMIT_GRACE_MS = 1200;
 var STT_RECONNECT_FLAP_WINDOW_MS = 4000;
 var MAX_STT_RECONNECTS_IN_FLAP_WINDOW = 3;
+var STT_HEALTH_STALE_MS = 6000;
+var STT_HEALTH_SPEECH_GAP_MS = 2000;
 var DEFAULT_FORMAT = {
   channels: 1,
   container: "raw",
@@ -6199,6 +6201,8 @@ var createVoiceSession = (options) => {
   let activeAdapterGeneration = 0;
   let sttReconnectCount = 0;
   let lastSttReconnectAt = 0;
+  let lastSpeechEnergyAt = 0;
+  let sttHealthPhaseStart = 0;
   let activeTTSTurnId;
   let assistantSpeechEndsAt = 0;
   let lastAssistantAudioAt = 0;
@@ -6866,13 +6870,11 @@ var createVoiceSession = (options) => {
     if (!didComplete) {
       return;
     }
-    if (shouldInvokeOnComplete) {
-      await options.route.onComplete({
-        api,
-        context: options.context,
-        session
-      });
-    }
+    const onCompletePromise = shouldInvokeOnComplete ? Promise.resolve(options.route.onComplete({
+      api,
+      context: options.context,
+      session
+    })) : Promise.resolve();
     if (disposition === "completed") {
       await drainAssistantSpeech(lastTtsSendAt);
       await assistantAudioQueue;
@@ -6935,6 +6937,7 @@ var createVoiceSession = (options) => {
         session
       });
     }
+    await onCompletePromise;
     await options.route.onCallEnd?.({
       api,
       context: options.context,
@@ -8531,6 +8534,30 @@ var createVoiceSession = (options) => {
       } else {
         clearSilenceTimer();
       }
+      const nowMs = Date.now();
+      if (nowMs - lastSpeechEnergyAt > STT_HEALTH_SPEECH_GAP_MS) {
+        sttHealthPhaseStart = nowMs;
+      }
+      lastSpeechEnergyAt = nowMs;
+      const lastTranscriptAt = latest.currentTurn.lastTranscriptAt ?? 0;
+      if (!options.realtime && sttSession && lastTranscriptAt < sttHealthPhaseStart && nowMs - sttHealthPhaseStart >= STT_HEALTH_STALE_MS) {
+        sttReconnectCount = nowMs - lastSttReconnectAt < STT_RECONNECT_FLAP_WINDOW_MS ? sttReconnectCount + 1 : 1;
+        lastSttReconnectAt = nowMs;
+        sttHealthPhaseStart = nowMs;
+        if (sttReconnectCount <= MAX_STT_RECONNECTS_IN_FLAP_WINDOW) {
+          await appendTrace({
+            payload: {
+              action: "stt-health-reconnect",
+              attempt: sttReconnectCount,
+              reason: `no transcript for ${STT_HEALTH_STALE_MS}ms of continuous speech`
+            },
+            session: latest,
+            type: "session.error"
+          });
+          await closeAdapter("stt stale; health-reconnect");
+          return;
+        }
+      }
     } else if (speechDetected) {
       backchannelDriver?.noteSilence();
       const currentSession = await readSession();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@absolutejs/voice",
-  "version": "0.0.22-beta.594",
+  "version": "0.0.22-beta.596",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",