add: spinner for querying document, increment in number of sources referenced

qa/qa.py

@@ -5,10 +5,10 @@ from typing import Text, Union
 
 multiple_files = True
 
-def query_pipeline(index:VectorStore, query:Text, stream_answer:bool=False)->Text:
+def query_pipeline(index:VectorStore, query:Text, stream_answer:bool=False, n_sources:int=5)->Text:
     """This function reproduces the querying pipeline considering a given input index."""
     # retrieving the most relevant pieces of information within the knowledge base
-    sources = search_docs(index, query=query)
+    sources = search_docs(index, query=query, k=n_sources)
     # getting the answer, all at once
     answer = get_answer(sources, query=query, stream_answer=stream_answer)["output_text"]
     
@@ -94,7 +94,8 @@ def parse_file(file:Union[PDFFile, DocxFile, TxtFile, CodeFile]) -> None:
 # def document_embedding_pipeline(file:Union[PDFFile, DocxFile, TxtFile, CodeFile]) -> None:  
 
 def qa_main():
-    """Main function for the QA app."""    
+    """Main function for the QA app."""
+    st.title("Chat with a file 💬📖")
     st.write("Just upload something using and start chatting with a version of GPT4 that has read the file!")
     
     # OpenAI API Key - TODO: consider adding a key valid for everyone
@@ -186,14 +187,15 @@ def qa_main():
                 full_response += \
                     f"<i>Asking</i> <b>{chat_document}</b> <i>question</i> <b>{prompt}</b></i><br>"
                 message_placeholder.markdown(full_response, unsafe_allow_html=True)
-                # retrieving the vector store associated to the chat document considered
-                chat_index = indexes[chat_document]
-                # producing the answer considered, live
-                for answer_bit in query_pipeline(chat_index, prompt, stream_answer=True):
-                    full_response += answer_bit
-                    message_placeholder.markdown(full_response + "▌", unsafe_allow_html=True)
-                # appending a final entering
-                full_response += "<br>"
+                with st.spinner("Querying the document..."):
+                    # retrieving the vector store associated to the chat document considered
+                    chat_index = indexes[chat_document]
+                    # producing the answer considered, live
+                    for answer_bit in query_pipeline(chat_index, prompt, stream_answer=True, n_sources=20):
+                        full_response += answer_bit
+                        message_placeholder.markdown(full_response + "▌", unsafe_allow_html=True)
+                    # appending a final entering
+                    full_response += "<br>"
             message_placeholder.markdown(full_response, unsafe_allow_html=True)
         
         # appending the final response obtained after having asked all the documents