fix: do not fire no-speech reprompt while caller is still mid-utterance with live partial ASR

services/ai_voice_runtime_service/media_runtime.py

@@ -956,6 +956,11 @@ class AudioSocketMediaRuntime:
                 bool(partial.is_final),
                 transcript_text[:160],
             )
+            # Fresh speech content is still arriving, so push the
+            # no-speech silence-timeout deadline forward instead of
+            # interrupting a caller who is actively mid-utterance.
+            if actor.state == "listening":
+                actor.listening_since_monotonic = time.monotonic()
         actor.partial_transcript = transcript_text
         self._update_stable_partial_transcript(actor, transcript_text, provider_stable=bool(partial.is_stable or partial.is_final))
         intent = self._detect_early_intent(transcript_text)