Update app.py

app.py

@@ -31,7 +31,10 @@ def askme(symptoms, question):
     inputs = tokenizer(prompt, return_tensors="pt").to(device)  # Ensure inputs are on CUDA device
     outputs = model.generate(**inputs, max_new_tokens=200, use_cache=True)
     response_text = tokenizer.batch_decode(outputs)[0].strip()
-    return answer
+    # Remove system messages and content
+    response_text = response_text.replace(sys_message, "").replace(content, "").strip()
+
+    return response_text
 
 # Example usage
 symptoms = '''\