Persist user history

app.py

@@ -211,10 +211,10 @@ def chat_with_model(messages):
 
     if in_think:
         output_text += "*"
-        updated_messages[-1]["content"] = output_text
 
-    torch.cuda.empty_cache()
     updated_messages[-1]["content"] = output_text
+    patient_conversations[current_id] = updated_messages  # <- SAVE the full conversation including model outputs
+    torch.cuda.empty_cache()
     return updated_messages