Tune web voice VAD for short utterances

packages/app/src/contexts/voice-context.tsx

@@ -8,6 +8,10 @@ import { activateKeepAwakeAsync, deactivateKeepAwake } from "expo-keep-awake";
 
 const VOICE_CONVERSATION_ID_STORAGE_KEY = "@paseo:voice-conversation-id";
 const KEEP_AWAKE_TAG = "paseo:voice";
+const VOICE_VAD_VOLUME_THRESHOLD = 0.18;
+const VOICE_VAD_SILENCE_DURATION_MS = 1400;
+const VOICE_VAD_SPEECH_CONFIRMATION_MS = 120;
+const VOICE_VAD_DETECTION_GRACE_PERIOD_MS = 700;
 
 interface VoiceContextValue {
   isVoiceMode: boolean;
@@ -123,10 +127,10 @@ export function VoiceProvider({ children }: VoiceProviderProps) {
         console.error("[Voice] Cannot handle error - setMessages not available from SessionState");
       }
     },
-    volumeThreshold: 0.3,
+    volumeThreshold: VOICE_VAD_VOLUME_THRESHOLD,
-    silenceDuration: 2000,
+    silenceDuration: VOICE_VAD_SILENCE_DURATION_MS,
-    speechConfirmationDuration: 300,
+    speechConfirmationDuration: VOICE_VAD_SPEECH_CONFIRMATION_MS,
-    detectionGracePeriod: 200,
+    detectionGracePeriod: VOICE_VAD_DETECTION_GRACE_PERIOD_MS,
   });
 
   useEffect(() => {

packages/app/src/voice/speech-segmenter.test.ts

@@ -88,4 +88,37 @@ describe("SpeechSegmenter", () => {
     expect(lastCall.isLast).toBe(true);
     expect(lastCall.audioData.length).toBeGreaterThan(0);
   });
+
+  it("keeps detection alive across a brief pause and confirms short utterances", () => {
+    const onSpeechStart = vi.fn();
+    const detectingChanges: boolean[] = [];
+
+    const segmenter = new SpeechSegmenter(
+      {
+        enableContinuousStreaming: false,
+        volumeThreshold: 0.18,
+        silenceDurationMs: 1400,
+        speechConfirmationMs: 120,
+        detectionGracePeriodMs: 700,
+        minChunkDurationMs: 100,
+        pcmSampleRate: 1000,
+      },
+      {
+        onSpeechStart,
+        onDetectingChange: (v) => detectingChanges.push(v),
+      }
+    );
+
+    const t0 = 20_000;
+
+    segmenter.pushVolumeLevel(0.4, t0);
+    segmenter.pushPcmChunk(mkPcmBytes(20));
+    segmenter.pushVolumeLevel(0.0, t0 + 80);
+    segmenter.pushPcmChunk(mkPcmBytes(20));
+    segmenter.pushVolumeLevel(0.4, t0 + 140);
+    segmenter.pushPcmChunk(mkPcmBytes(20));
+
+    expect(detectingChanges).toContain(true);
+    expect(onSpeechStart).toHaveBeenCalledTimes(1);
+  });
 });