preventing segmentation fault

src/app.py

@@ -207,11 +207,9 @@ llm = LlamaCPP(
     temperature=0.7,
     max_new_tokens=256,
     context_window=2048,
-    n_batch=512,      # Added batch size limit
-    n_ctx=2048,       # Explicit context window
-    verbose=False,    # Reduce logging
-    n_threads=4,      # Limit threads
-    last_n_tokens_size=256  # Limit token history
+    verbose=False    # Reduce logging
+    # n_batch and n_threads are not valid parameters for LlamaCPP and should not be used.
+    # If you encounter segmentation faults, try reducing context_window or check your system resources.
 )
 print("LLM initialized successfully")
 
@@ -560,6 +558,13 @@ with gr.Blocks(
         queue=False
     )
     
+    def cleanup_memory():
+        """Release unused memory (placeholder for future memory management)."""
+        import gc
+        gc.collect()
+        if torch.cuda.is_available():
+            torch.cuda.empty_cache()
+
     def process_text_input(text, history):
         """Process text input with memory management."""
         if not text: