I have a markdown file in which there are texts and table. Can you give me suggestions how can I ensure that my table is not chunked in the middle ? I want to ensure that the table is still intact
Is there a way to get metadata when I am using chat engine ?, say response.metadata , but I am getting empty list. Basically I want to get from which document / filename that response comes from
I am currently using agent with PandasQueryEngine as a tool for reading excel file. It can read information from whatever is available in the file. But is there a way so that it can also read a formula that is available in the excel file ?
Hi I am using ReAct agent and I put query_engine_tools. Can ReAct agent pick up more than one tools when processing a query ? is it by default ? something like LLMMultiSelector. Fyi, I am using PandasQueryEngine for query_engine_tools
I have implemented agent to query over PandasQueryEngine as tools. It works fine. However, how can I get metadata once it gives the results ? I need some metadata for example for citation purposes.
@kapa.ai I want to do pre-processing for uploaded pdf files. I used PyMuPDF4LLM but it is not multithreading safe. The reason I am using that library because it removes headers and footer and turn the documents into markdown from which I can do some pre-processing. Do you have any suggestion how to turn uploaded PDF files into markdown so that I can remove headers and footers while implementing multithreading ?
@kapa.ai I am currently using CitationQueryEngine, which provides a nice citation. However, it does not have a conversation history like what chat engine does. Can you give me some ideas / codes that allow me to use CitationQueryEngine but at the same time, it can maintain a context or preserve a conversation history so that it will have the ability of chat engine at the same time ?
@kapa.ai Here is my code. I am wondering if there is a way to put conversation history as parameter as chat_engine.chat(query, conversation_history) instead of chat_engine.chat(query). The whole idea is I don't have to reinstantiate chat engine every time there is a new conversation history
def create_chat_engine(self): # Create a chat memory buffer for conversation history memory = ChatMemoryBuffer.from_defaults(token_limit=default_configuration.TOKEN_LIMIT_MEMORY_BUFFER_CHAT_ENGINE)
# Set a conversation history into a memory as chat_history memory.set(chat_messages)
# Create a chat engine chat_engine = CondensePlusContextChatEngine.from_defaults( index.as_retriever(), memory=memory, llm=llm, context_prompt=( "You are a chatbot, able to have normal interactions as well as talk in a professional manner." "Here are the relevant documents for the context:\n" "{context_str}" "\nInstruction: Use the previous chat history and the context above to interact and help the user." ), verbose=False, )
return chat_engine
def use_chat_engine(self):
chat_engine = self.create_chat_engine() query = "Please tell me about LlamaIndex) chat_engine.chat(query)
@kapa.ai I am using this code in order to maintain a conversation history. However, since I use Flask, python garbage collector frees up all values of variables after sending response to frontend. It means that values of memory and self.chat_engine are always wiped out after sending the response to the frontend. Do you have any suggestions to maintain memory and self.chat_engine considering client & servers work including Flask ?
Create a chat memory buffer for conversation history
# Create a chat engine self.chat_engine = CondensePlusContextChatEngine.from_defaults( index.as_retriever(), memory=memory, llm=llm, context_prompt=( "You are a chatbot, able to have normal interactions, as well as talk in a professional manner" "about the attached document(s)." "Here are the relevant documents for the context:\n" "{context_str}" "\nInstruction: Use the previous chat history, or the context above, to interact and help the user." ), verbose=False, )
