Create a conversational chatbot capable of asking questions in order to fulfil user's query

[swapnil3597]

How to create a conversational chatbot capable of asking questions in order to fulfil user's query. I will explain this using some example scenarios.

Scenario 1:

# Scenario 1:
USER: I want to book airline ticket from A to B
AI AGENT: Sure, I have booked the ticket for you. Here's you airline ticket from A to B: <link-to-ticket>

This can be implemented in the following way:

# Tools
from langchain.agents import Tool

tools = [
    Tool(
        name = "Ticket booking tool",
        func=book_ticket,
        verbose = True,
        description="useful only when user wants to book a ticket",
        return_direct=True
    )
]

# Defining book_ticket function
def book_ticket(q):
    # Get source and destination using named entity recognition model
    source, dest = NER(q)

    # Book to ticket
    link_to_ticket = ticket_booking_api(source, dest)
  
    return "Sure, I have booked the ticket for you. Here's you airline ticket from A to B: " + str(link_to_ticket)

Is there any better way to implement this? Especially the, named entity recognition part. Can I leverage the power of LLMs for this?

---

Scenario 2:

# Scenario 2:
USER: I want to book airline ticket
AI AGENT: Sure, Can you please provide me with the source and destination
USER: Yes, Please book ticket from A to B
AI AGENT: Sure, I have booked the ticket for you. Here's you airline ticket from A to B: <link-to-ticket>

Such a scenario is generally implemented using conversational flows in traditional chatbot systems. But, how can I leverage the LLMs to built such a chatbot?

Please help with this scenarios.

[swapnil3597]

Achieved above mentioned scenario. Please refer to notebook in following GitHub repo: repo-link

[oddrationale]

This is awesome!

You can also look at the Human as a tool section of the documentation, which may provide an alternative way of structuring the agent.