hot-fix: memory

agents/ensemble_team.py

@@ -4,6 +4,7 @@ import torch
 import psutil # Ensure psutil is imported here as well
 import GPUtil
 from datetime import datetime, timedelta
+import gc # Import garbage collector
 
 logger = logging.getLogger(__name__)
 
@@ -40,7 +41,7 @@ class EnsembleMonitorAgent:
             self.alerts.append(alert_msg)
             logger.warning(alert_msg)
             
-        logger.debug(f"Updated metrics for '{model_id}': {metrics}")
+        logger.info(f"Updated metrics for '{model_id}': {metrics}")
     
     def get_performance_summary(self):
         logger.info("Generating performance summary for all models.")
@@ -74,7 +75,7 @@ class WeightOptimizationAgent:
         
         # Keep history windowed
         self.prediction_history = [p for p in self.prediction_history if timestamp - p["timestamp"] < self.performance_window]
-        logger.debug(f"Prediction history length: {len(self.prediction_history)}")
+        logger.info(f"Prediction history length: {len(self.prediction_history)}")
 
         # In a real scenario, this would involve a more complex optimization logic
         # For now, it just logs the history length.
@@ -98,6 +99,19 @@ class SystemHealthAgent:
             "percent": mem.percent
         }
 
+        # Holy moly, been at 99% for hours whoops
+        if mem.percent > 90: 
+            logger.warning(f"CRITICAL: System memory usage is at {mem.percent}%. Attempting to clear memory cache...")
+            gc.collect() 
+            logger.info("Garbage collection triggered. Re-checking memory usage...")
+            mem_after_gc = psutil.virtual_memory()
+            self.health_metrics["memory_usage_after_gc"] = {
+                "total": mem_after_gc.total,
+                "available": mem_after_gc.available,
+                "percent": mem_after_gc.percent
+            }
+            logger.info(f"Memory usage after GC: {mem_after_gc.percent}%")
+
         gpu_info = []
         try:
             gpus = GPUtil.getGPUs()