don't remove <br> only reduce newlines

app.py

@@ -5,6 +5,7 @@ import retrieval
 # UNCOMMENT ONLY WHEN RUNNING LOCALLY (not on Spaces)
 # from dotenv import load_dotenv
 from text_generation import Client, InferenceAPIClient
+from typing import List, Tuple
 
 # load API keys from globally-availabe .env file
 # SECRETS_FILEPATH = "/mnt/project/chatbotai/huggingface_cache/internal_api_keys.env"
@@ -106,7 +107,7 @@ def predict(
         stop_sequences=[user_name.rstrip(), assistant_name.rstrip()],
     )
 
-  final_chat_response = None
+  chat_response = None
   for i, response in enumerate(iterator):
     if response.token.special:
       continue
@@ -123,33 +124,33 @@ 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
+    chat_response = chat
     yield chat, history, None, None, None, []
   
-  print("1. final_chat_response:", final_chat_response)
-  print("1. history:", history)
-  
-  # nested list, lame.
-  cleaned_final_chat_response = []
-  for human_chat, bot_chat in final_chat_response:
-    human_chat = human_chat.replace("<br>", "")
-    human_chat = human_chat.replace("\n\n", "\n")
-    bot_chat = bot_chat.replace("<br>", "")
-    bot_chat = bot_chat.replace("\n\n", "\n")
-    cleaned_final_chat_response.append( (human_chat, bot_chat) )
-  
-  print(">>>>>>>>>>>> KEYY. final_chat_response -- AFTER CLEANING -----:", cleaned_final_chat_response)
-
+  cleaned_final_chat_response = clean_chat_response(chat_response)
   # 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 cleaned_final_chat_response, history, top_context_list[0], top_context_list[1], top_context_list[2], []
+  
+  cleaned_final_chat_response = clean_chat_response(chat_response)
 
   # 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 cleaned_final_chat_response, history, top_context_list[0], top_context_list[1], top_context_list[2], images_list
 
+def clean_chat_response(chat: List[Tuple[str, str]]) -> List[Tuple[str, str]]:
+  ''' Not perfect, but much better at removing all the crazy newlines. '''
+  cleaned_chat = []
+  for human_chat, bot_chat in chat:
+    # human_chat = human_chat.replace("<br>", "")
+    human_chat = human_chat.replace("\n\n", "\n")
+    # bot_chat = bot_chat.replace("<br>", "")
+    bot_chat = bot_chat.replace("\n\n", "\n")
+    cleaned_chat.append( (human_chat, bot_chat) )
+  return cleaned_chat
+  
 
 def reset_textbox():
   return gr.update(value="")