@absolutejs/voice 0.0.22-beta.595 → 0.0.22-beta.597

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",
@@ -3755,6 +3757,8 @@ var getBufferedAudioDurationMs = (chunks) => chunks.reduce((total, chunk) => tot
 var STREAM_SENTENCE_BOUNDARY = /[.!?\u2026]['")\]]*\s/;
 var STREAM_CLAUSE_BOUNDARY = /[,;:]\s/g;
 var MAX_TTS_CHUNK_CHARS = 320;
+var STREAM_SENTENCE_END = /[.!?\u2026]['")\]]*$/;
+var STREAM_IDLE_FLUSH_MS = 350;
 var nextSpeakableBoundary = (buffer) => {
   const match = STREAM_SENTENCE_BOUNDARY.exec(buffer);
   return match ? match.index + match[0].length : -1;
@@ -3972,6 +3976,8 @@ 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;
@@ -5444,6 +5450,7 @@ var createVoiceSession = (options) => {
     let charsSent = 0;
     let started = false;
     let streamed = false;
+    let idleFlushTimer = null;
     let sendChain = Promise.resolve();
     let ttsSessionRequest = null;
     const ttsStartedAt = Date.now();
@@ -5503,8 +5510,23 @@ var createVoiceSession = (options) => {
         }
       })();
     };
+    const clearIdleFlush = () => {
+      if (idleFlushTimer) {
+        clearTimeout(idleFlushTimer);
+        idleFlushTimer = null;
+      }
+    };
+    const flushOnIdle = () => {
+      idleFlushTimer = null;
+      const pending = buffer.trim();
+      if (pending && STREAM_SENTENCE_END.test(pending)) {
+        flush(buffer);
+        buffer = "";
+      }
+    };
     return {
       finish: async () => {
+        clearIdleFlush();
         if (buffer.trim()) {
           flush(buffer);
         }
@@ -5550,6 +5572,10 @@ var createVoiceSession = (options) => {
           flush(buffer.slice(0, cut));
           buffer = buffer.slice(cut);
         }
+        clearIdleFlush();
+        if (buffer.trim()) {
+          idleFlushTimer = setTimeout(flushOnIdle, STREAM_IDLE_FLUSH_MS);
+        }
       }
     };
   };
@@ -6303,6 +6329,30 @@ 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",
@@ -5982,6 +5984,8 @@ var getBufferedAudioDurationMs = (chunks) => chunks.reduce((total, chunk) => tot
 var STREAM_SENTENCE_BOUNDARY = /[.!?\u2026]['")\]]*\s/;
 var STREAM_CLAUSE_BOUNDARY = /[,;:]\s/g;
 var MAX_TTS_CHUNK_CHARS = 320;
+var STREAM_SENTENCE_END = /[.!?\u2026]['")\]]*$/;
+var STREAM_IDLE_FLUSH_MS = 350;
 var nextSpeakableBoundary = (buffer) => {
   const match = STREAM_SENTENCE_BOUNDARY.exec(buffer);
   return match ? match.index + match[0].length : -1;
@@ -6199,6 +6203,8 @@ 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;
@@ -7671,6 +7677,7 @@ var createVoiceSession = (options) => {
     let charsSent = 0;
     let started = false;
     let streamed = false;
+    let idleFlushTimer = null;
     let sendChain = Promise.resolve();
     let ttsSessionRequest = null;
     const ttsStartedAt = Date.now();
@@ -7730,8 +7737,23 @@ var createVoiceSession = (options) => {
         }
       })();
     };
+    const clearIdleFlush = () => {
+      if (idleFlushTimer) {
+        clearTimeout(idleFlushTimer);
+        idleFlushTimer = null;
+      }
+    };
+    const flushOnIdle = () => {
+      idleFlushTimer = null;
+      const pending = buffer.trim();
+      if (pending && STREAM_SENTENCE_END.test(pending)) {
+        flush(buffer);
+        buffer = "";
+      }
+    };
     return {
       finish: async () => {
+        clearIdleFlush();
         if (buffer.trim()) {
           flush(buffer);
         }
@@ -7777,6 +7799,10 @@ var createVoiceSession = (options) => {
           flush(buffer.slice(0, cut));
           buffer = buffer.slice(cut);
         }
+        clearIdleFlush();
+        if (buffer.trim()) {
+          idleFlushTimer = setTimeout(flushOnIdle, STREAM_IDLE_FLUSH_MS);
+        }
       }
     };
   };
@@ -8530,6 +8556,30 @@ 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.595",
+  "version": "0.0.22-beta.597",
   "description": "Voice primitives and Elysia plugin for AbsoluteJS",
   "repository": {
     "type": "git",