worst bug with visual orchestration

src/gui_2.py

@@ -253,6 +253,12 @@ class App:
   try:
    self.perf_monitor.start_frame()
    # Process GUI task queue
+   # DEBUG: Check if tasks exist before processing
+   if hasattr(self, 'controller') and hasattr(self.controller, '_pending_gui_tasks'):
+       pending_count = len(self.controller._pending_gui_tasks)
+       if pending_count > 0:
+           sys.stderr.write(f"[DEBUG gui_2] _gui_func: found {pending_count} pending tasks\n")
+           sys.stderr.flush()
    self._process_pending_gui_tasks()
    self._process_pending_history_adds()
    self._render_track_proposal_modal()

src/orchestrator_pm.py

@@ -74,6 +74,12 @@ def generate_tracks(user_request: str, project_config: dict[str, Any], file_item
  # Set custom system prompt for this call
  old_system_prompt = ai_client._custom_system_prompt
  ai_client.set_custom_system_prompt(system_prompt or "")
+ # Ensure we use the current provider from ai_client state
+ # Import ai_client module-level to access globals
+ import src.ai_client as ai_client_module
+ current_provider = ai_client_module._provider
+ current_model = ai_client_module._model
+ ai_client.set_provider(current_provider, current_model)
  try:
  # 3. Call Tier 1 Model (Strategic - Pro)
  # Note: We use gemini-1.5-pro or similar high-reasoning model for Tier 1