fix: add STT timeout, move finally restart to thread, guard exit on recording

- Set OpenAI client timeout=30s in transcribe_audio() — default 600s
  blocks _voice_processing for 10 min if Groq/OpenAI stalls
- Move _voice_start_recording in _voice_stop_and_transcribe finally
  block to a daemon thread (same pattern as Ctrl+B handler and
  process_loop)
- Add _should_exit guard at top of _voice_start_recording so all 4
  call sites respect shutdown without individual checks

cli.py

@@ -3544,6 +3544,8 @@ class HermesCLI:
 
     def _voice_start_recording(self):
         """Start capturing audio from the microphone."""
+        if getattr(self, '_should_exit', False):
+            return
         from tools.voice_mode import AudioRecorder, check_voice_requirements
 
         reqs = check_voice_requirements()
@@ -3691,12 +3693,14 @@ class HermesCLI:
             # (When transcript IS submitted, process_loop handles restart
             # after chat() completes.)
             if self._voice_continuous and not submitted and not self._voice_recording:
-                try:
-                    self._voice_start_recording()
-                    if hasattr(self, '_app') and self._app:
-                        self._app.invalidate()
-                except Exception:
-                    pass
+                def _restart_recording():
+                    try:
+                        self._voice_start_recording()
+                        if hasattr(self, '_app') and self._app:
+                            self._app.invalidate()
+                    except Exception:
+                        pass
+                threading.Thread(target=_restart_recording, daemon=True).start()
 
     def _voice_speak_response(self, text: str):
         """Speak the agent's response aloud using TTS (runs in background thread)."""

tools/transcription_tools.py

@@ -150,7 +150,7 @@ def transcribe_audio(file_path: str, model: Optional[str] = None) -> Dict[str, A
     try:
         from openai import OpenAI, APIError, APIConnectionError, APITimeoutError
 
-        client = OpenAI(api_key=api_key, base_url=base_url)
+        client = OpenAI(api_key=api_key, base_url=base_url, timeout=30)
 
         with open(file_path, "rb") as audio_file:
             transcription = client.audio.transcriptions.create(