Search⌘ K
AI Features
Home
Courses
Build AI Chatbots with Open-Source LLMs, LangChain, and Streamlit

Capstone Project: Functions

Explore how to configure essential functions for your LLM-driven Toyota chatbot. This lesson guides you through session state management, loading and processing user manuals, handling vector stores, and displaying dynamic content with Streamlit to build a functional chatbot interface.

We'll cover the following...

We continue this capstone project by setting up the main functions of our chatbots, defining the session states that will persist in memory throughout the user interaction session, and finally, setting up the function that will load and process the user manual and display the Toyota images.

Step 5: Setting up chatbot.py file

We’ll now walk through the step-by-step configuration of the chatbot.py file:

Python
import streamlit as st
from src.load_file import load_file
from src.rewrite_query import rewrite_user_query
from src.retriever import build_vector_store, retrieve_chunks_from_vector_store, retrieve_history
from src.generator import generate_answer
# Load the user manual file
user_manual_content = load_file()
# If load is successful
if user_manual_content:
    try:
        ###################################### Display the title ######################################
        st.title(":rainbow[TOYOTA HIGHLANDER INTERACTIVE BOT]")
        st.write('')
        ####################### Set a clear conversation button on the side menu ######################
        clear_conversation = st.sidebar.button(label='Clear conversation',
                                               key='clear_conversation',
                                               use_container_width=True)
        if clear_conversation:
            st.session_state.messages = []
        ############################### Display the chatbot input space ###############################
        user_input = st.chat_input(
            'Ask me a question about the Toyota Highlander...',
            max_chars=1500,
            key='user_input')
        ######################### Build the vector store to store the vectors #########################
        vector_store = build_vector_store(user_manual_content)
        ######################### Check and display any previous chat history #########################
        history = retrieve_history()
        if user_input:
            ############################ Append the user input to the chat ############################
            st.session_state.messages.append({"role": "user", "content": user_input})
            #################### Rewrite the user query for better input to the LLM ###################
            re_written_query = rewrite_user_query(user_input)
            ###################### Retrieve the relevant chunks from the database #####################
            relevant_chunks = retrieve_chunks_from_vector_store(vector_store, re_written_query)
            ######################### Generate a final answer using the LLM ##########################
            answer = generate_answer(re_written_query, relevant_chunks)
            ##################### Display the answer with the relevant information ####################
            col_left, col_right = st.columns(2)
            with col_left:
                with st.expander(label='Re-written user query', expanded=False):
                    st.write(re_written_query)
            with col_right:
                with st.expander(label='Retrieved relevant text from the car user manual', expanded=False):
                    st.write(relevant_chunks)
    # Handle the exception if the user manual is not loaded
    except Exception as e:
        print(e)
        st.error("Sorry, an error occurred.")

  • Line 1: We import Streamlit for building the user interface.

  • Line 2: We import a function to load the Toyota User Manual.

  • Line 3: We import a function to rewrite user queries.

  • Line 4: We import functions for building the vector store, retrieving the relevant chunks, and fetching the chat history.

  • Line 5: We import a function to generate an answer using the LLM.

  • Line 8: We load the Toyota User Manual content using the load_file function.

  • Line 11: We check if the user manual is not empty.

  • Line 12: We set ...