Find answers from the community

Updated 2 years ago

Sources

At a glance

Hi guys! How to return sources from the index when using ConversationBufferMemory & Tool, the run method on the agent_chain returns only the result of the query but no sources / nodes

8 comments

LLogan M

You are using llama index as a toolnin langchain right? How did you set up your tools?

nn4l5u0r

defined it like that :

Plain Text

 tools = [
                Tool(
                    name="GPT Index",
                    func=lambda q: str(index.as_query_engine().query(q)),
                    description="useful for when you want to answer questions about the author. The input to this tool should be a complete english sentence.",
                    return_direct=True,
                ),
            ]

            
memory = ConversationBufferMemory(memory_key="chat_history")
llm = ChatOpenAI(temperature=0)
agent_chain = initialize_agent(tools, llm, agent="conversational-react-description", memory=memory)

LLogan M

Cool! So instead of a lambda, you could use a wrapper function instead. That way you can get access to the source nodes

Plain Text

def query_with_sources(query_str, query_engine=None):
    response = query_engine.query(query_str)
    print(response.source_nodes)
    return str(response)

...
func=lambda q: query_with_sources(q, query_engine=index.query_engine),
...

LLogan M

Just a very basic example

nn4l5u0r

oh perfect this is much clear thank you very much

nn4l5u0r

Hello @Logan M what is the purpose of description in Tool Object ? Looks like very sensitive regarding the return of the result

LLogan M

So the LLM reads all the tool names and descriptions, and decides which tool to use (if any) based on that

LLogan M

So any changes in the description will change how it picks tools

Add a reply