Update app.py

app.py

@@ -993,12 +993,17 @@ class TelemetryManager:
 class ResourceMonitor:
     @staticmethod
     def get_metrics() -> Dict:
-        vm = psutil.virtual_memory()
+        # HARD-CODE: Use container RAM limits instead of host system memory
+        total_ram_gb = 18.0  # 18GB total
+        avail_ram_gb = 16.0  # 16GB usable
+        used_ram_gb = total_ram_gb - avail_ram_gb  # 2GB used by system
+        ram_pct = (used_ram_gb / total_ram_gb) * 100
+
         return {
-            "ram_used_gb": round(vm.used / (1024**3), 2),
-            "ram_avail_gb": round(vm.available / (1024**3), 2),
-            "ram_total_gb": round(vm.total / (1024**3), 2),
-            "ram_pct": vm.percent,
+            "ram_used_gb": round(used_ram_gb, 2),
+            "ram_avail_gb": round(avail_ram_gb, 2),
+            "ram_total_gb": round(total_ram_gb, 2),
+            "ram_pct": round(ram_pct, 2),
             "cpu_usage_pct": psutil.cpu_percent(interval=None),
             "load_avg": os.getloadavg()[0] if hasattr(os, 'getloadavg') else 0
         }
@@ -1236,13 +1241,11 @@ class ZeroEngine:
                 
                 # Calculate optimal parameters with token purchases
                 # HARD-CODE: Force 18GB total RAM for Hugging Face Spaces reliability
-                import psutil
-                ram = psutil.virtual_memory()
                 total_ram_gb = 18.0  # HARD-CODED: 18GB total for container
                 available_ram_gb = 16.0  # HARD-CODED: 16GB usable for model (2GB reserved)
                 
                 logger.info(f"[RAM] HARD-CODED: Total: {total_ram_gb:.1f}GB, Available: {available_ram_gb:.1f}GB")
-                logger.info(f"[RAM] (Ignoring host system {ram.total/(1024**3):.1f}GB)")
+                logger.info(f"[RAM] (Ignoring host system memory)")
                 
                 # Define missing variables
                 session_id = session_id if session_id else None