Update app.py

app.py

@@ -92,7 +92,6 @@ def generate(
         num_beams=1,
         repetition_penalty=repetition_penalty,
     )
-    save_json("user", message)
     
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()
@@ -101,7 +100,8 @@ def generate(
     for text in streamer:
         outputs.append(text)
         yield "".join(outputs)
-
+        
+    save_json("user", message)
     save_json("assistant", "".join(outputs))