chat history via final_chat_response

app.py

@@ -106,6 +106,7 @@ def predict(
         stop_sequences=[user_name.rstrip(), assistant_name.rstrip()],
     )
 
+  final_chat_response = None
   for i, response in enumerate(iterator):
     if response.token.special:
       continue
@@ -122,19 +123,18 @@ def predict(
       history[-1] = partial_words
 
     chat = [(history[i].strip(), history[i + 1].strip()) for i in range(0, len(history) - 1, 2)]
+    final_chat_response = chat
     yield chat, history, None, None, None, []
 
-  chat = [(history[i].strip(), history[i + 1].strip()) for i in range(0, len(history) - 1, 2)]
-  
   # Pinecone context retrieval
   top_context_list = ta.retrieve_contexts_from_pinecone(user_question=inputs, topk=NUM_ANSWERS_GENERATED)
   # yield chat, history, top_context_list[0], top_context_list[1], top_context_list[2], []
-  yield chat, history, top_context_list[0], top_context_list[1], top_context_list[2], []
+  yield final_chat_response, history, top_context_list[0], top_context_list[1], top_context_list[2], []
 
   # run CLIP
   images_list = ta.clip_text_to_image(inputs)
   # yield chat, history, top_context_list[0], top_context_list[1], top_context_list[2], images_list
-  yield chat, history, top_context_list[0], top_context_list[1], top_context_list[2], images_list
+  yield final_chat_response, history, top_context_list[0], top_context_list[1], top_context_list[2], images_list
 
 
 def reset_textbox():