Update app.py

app.py

@@ -1234,12 +1234,21 @@ class ZeroEngine:
                         logger.warning(f"[BOOT] Cleanup warning: {e}")
                 
                 # Calculate optimal parameters with token purchases
-                # Use actual system RAM detection
+                # FIX: Use container RAM limits, not host system memory
                 import psutil
                 ram = psutil.virtual_memory()
                 total_ram_gb = ram.total / (1024**3)
                 available_ram_gb = ram.available / (1024**3)
-                logger.info(f"[RAM] Total: {total_ram_gb:.1f}GB, Available: {available_ram_gb:.1f}GB")
+                
+                # CRITICAL FIX: Force realistic container limits for Hugging Face Spaces
+                # The host shows 123.8GB but container only has 16GB total
+                if total_ram_gb > 50.0:  # Host system memory detected
+                    logger.warning(f"[RAM] Host system memory detected ({total_ram_gb:.1f}GB), forcing container limits")
+                    total_ram_gb = 16.0  # Container limit
+                    available_ram_gb = min(available_ram_gb, 11.0)  # Conservative available RAM
+                    logger.info(f"[RAM] FORCED: Total: {total_ram_gb:.1f}GB, Available: {available_ram_gb:.1f}GB")
+                else:
+                    logger.info(f"[RAM] Total: {total_ram_gb:.1f}GB, Available: {available_ram_gb:.1f}GB")
                 
                 # CPU-OPTIMIZED BATCH CALCULATION - Very aggressive for 16GB RAM
                 # Base calculation: use more RAM for batching on CPU