Handling follow-up questions in LangChain - Mini Project: Build & Apply

Handling Follow-Up Questions with Langchain

📖 Scenario: You are building a simple chatbot using Langchain that can remember the context of previous questions and handle follow-up questions smoothly.

🎯 Goal: Create a Langchain chatbot that stores conversation history and uses it to answer follow-up questions correctly.

📋 What You'll Learn

Create a list to store conversation history

Set a maximum number of previous messages to keep

Use Langchain's ConversationalRetrievalChain to process questions with context

Add the final call to run the chatbot with a sample question

💡 Why This Matters

🌍 Real World

Chatbots and virtual assistants need to remember previous messages to answer follow-up questions naturally.

💼 Career

Understanding how to manage conversation history is key for building intelligent chatbots in customer support, education, and personal assistant applications.

1

Create conversation history list

Create a list called chat_history and initialize it as an empty list to store previous messages.

LangChain

# Create a list called chat_history to store previous messages
# Your code here

Need a hint?

Think of chat_history as a notebook where you write down all previous messages.

2

Set maximum history length

Create a variable called max_history and set it to 5 to limit how many previous messages to keep.

LangChain

chat_history = []
# Set max_history to 5 to limit stored messages
# Your code here

Need a hint?

This helps the chatbot remember only the last few messages, like a short-term memory.

3

Create ConversationalRetrievalChain with history

Create a variable called qa_chain and assign it to ConversationalRetrievalChain.from_llm(llm, retriever). Use chat_history as the conversation history input when calling qa_chain.run() later.

LangChain

chat_history = []
max_history = 5
# Create qa_chain using ConversationalRetrievalChain.from_llm with llm and retriever
# Your code here

Need a hint?

This chain helps the chatbot answer questions using previous conversation context.

4

Run the chatbot with follow-up question handling

Call qa_chain.run() with the question "What is Langchain?" and pass chat_history as the chat_history argument. Append the question and answer to chat_history. Keep only the last max_history messages in chat_history.

LangChain

from langchain.chains import ConversationalRetrievalChain

chat_history = []
max_history = 5
qa_chain = ConversationalRetrievalChain.from_llm(llm, retriever)
# Run qa_chain with question and chat_history, update chat_history to keep last max_history messages
# Your code here

Need a hint?

Think of chat_history as a queue that keeps only the last few messages.