Update app.py

app.py

@@ -1494,7 +1494,7 @@ class Hive:
 
         if self.lite_mode:
             prompt = f"<|user|>\n{message}</s>\n<|assistant|>\n"
-            full_reply = "".join(list(self.chat_stream(prompt, max_new_tokens, temp)))
+            full_reply = "".join(list(self.chat_stream(prompt, max_new_tokens=max_new_tokens, temperature=temp)))
             return full_reply
         
         kk = k if k is not None else (self.retrieval_k if hasattr(self, 'retrieval_k') else 6)
@@ -1665,7 +1665,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
             
             if hive_instance.lite_mode:
                 # Lite mode: direct, non-streaming response.
-                reply = hive_instance.chat(sanitized_m, eff, current_user_id)
+                reply = hive_instance.chat(sanitized_m, eff, current_user_id, max_new_tokens=256, temperature=0.7)
                 messages_hist.append({"role": "assistant", "content": reply or "[No response from model]"})
                 yield messages_hist, gr.Textbox(value="", interactive=True)
             else:
@@ -1782,7 +1782,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
             """Waits for ASR/TTS models and enables voice-related UI elements."""
             bootstrap_instance.voice_ready.wait()
             bootstrap_instance.hive_ready.wait() # Also wait for full core for voice features
-            hive_instance = get_hive_instance() # type: ignore
+            hive_instance = get_hive_instance(bootstrap_instance)
 
             voice_ready = not hive_instance.lite_mode and hasattr(hive_instance, 'asr_service') and hasattr(hive_instance, 'tts_service')
             video_ready = not hive_instance.lite_mode and hasattr(hive_instance, 'video_service') and CFG["VIDEO_ENABLED"] # type: ignore
@@ -1803,7 +1803,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
         demo.load(wait_for_voice_features, None, [voice_status_md, ptt_audio_in, ptt_transcript, ptt_transcribe_btn, ptt_chat_btn, vocal_chat_btn, enroll_audio, enroll_btn, who_btn, camera_status_md, video_out], show_progress="hidden")
         def stream_video():
             """Streams video frames from the VideoService to the UI."""
-            hive_instance = get_hive_instance()
+            hive_instance = get_hive_instance(bootstrap_instance)
             if not (
                 hive_instance and not hive_instance.lite_mode and 
                 hasattr(hive_instance, 'video_service') and hive_instance.video_service and
@@ -1821,7 +1821,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
         demo.load(stream_video, None, video_out)
 
         def do_online_update():
-            hive_instance = get_hive_instance() # type: ignore
+            hive_instance = get_hive_instance(bootstrap_instance)
             if hive_instance.lite_mode: return "Online features are disabled in Lite Mode." # type: ignore
             return "Added %s new summaries to curves." % (hive_instance.online_update().get("added",0))
 
@@ -1834,7 +1834,7 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
 
         # This function is now the core of the hands-free mode, using the new VADService.
         def process_vocal_chat_stream(stream, state, uid, role, mode, chatbot_history, request: gr.Request): # type: ignore
-            now = time.time()
+            now = time.time() # type: ignore
             hive_instance = get_hive_instance() # type: ignore
             if hive_instance.lite_mode or not hasattr(hive_instance, 'vad_service') or not hive_instance.vad_service: # type: ignore
                 return None, state, chatbot_history, "VAD service not ready."
@@ -2097,6 +2097,20 @@ def launch_ui(bootstrap_instance: "Bootstrap"):
         share=os.getenv("GRADIO_SHARE", "false").lower() == "true"
     ); return demo
 
+def get_hive_instance(bootstrap_instance: "Bootstrap"):
+    """
+    Global function to safely get the current Hive instance.
+    It prioritizes the full instance if ready, otherwise falls back to the lite one.
+    """
+    global HIVE_INSTANCE
+    if bootstrap_instance.hive_ready.is_set() and bootstrap_instance.hive_instance:
+        if HIVE_INSTANCE is None or HIVE_INSTANCE.lite_mode:
+            HIVE_INSTANCE = bootstrap_instance.hive_instance
+    elif HIVE_INSTANCE is None and bootstrap_instance.lite_core_ready.is_set():
+        HIVE_INSTANCE = bootstrap_instance.hive_lite_instance
+    return HIVE_INSTANCE
+
+
 class Bootstrap:
     """Handles the entire application startup sequence cleanly."""
     def __init__(self, config: Dict):
@@ -2223,7 +2237,7 @@ class Bootstrap:
     def _init_full_core(self):
         """Initializes all features of the full Hive core."""
         logging.info("Initializing Full Hive Core...") # Added logging
-        # --- Pre-launch heavy model loading in parallel ---
+        # This is now correctly calling the global get_hive_instance
         llm_thread = threading.Thread(target=lambda: get_hive_instance(lite=False, caps=self.caps), daemon=True)
         asr_thread = threading.Thread(target=get_asr, daemon=True)
         tts_thread = threading.Thread(target=lambda: get_tts(CFG["TTS_LANG"]), daemon=True)