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

Capstone Project: Retriever

Explore how to configure a retriever that fetches relevant data chunks from a vector store and rewrite user queries to improve clarity and effectiveness. This lesson guides you through implementing a Python retriever.py file that manages content fetching and vector store session storage, and a rewrite_query.py file that refines user queries using LLMs for better chatbot interactions.

We'll cover the following...

We continue this capstone project by configuring the retriever to fetch the relevant chunks from the vector store based on the user’s query and refining the user query to enhance its clarity and make it more understandable by the LLM.

Step 8: Setting up the retriever.py file

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

Python
import os
import requests
import streamlit as st
from tempfile import NamedTemporaryFile
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import SKLearnVectorStore
# Suppress warnings
import warnings
warnings.filterwarnings("ignore", category=UserWarning)
warnings.filterwarnings("ignore", category=FutureWarning)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
os.environ["TOKENIZERS_PARALLELISM"] = "false"
##############################################################################################################
################################### Function for retrieving the vector store ###################################
##############################################################################################################
def build_vector_store(content):
    if content:
        # If the vector store is not already present in the session state
        if not st.session_state.vector_store:
            with st.spinner(text=":red[Please wait while we fetch the information...]"):
                ################################# Fetch the embedding file ##################################
                embedding = HuggingFaceEmbeddings()
                embedding_file = 'https://raw.githubusercontent.com/Samuelchazy/Educative.io/8f0e764c1b69e2d61f4e44e3084c0695d85cd6e8/persistence/user_manuel.json'
                # Download the embedding file from the URL and save it temporarily
                with NamedTemporaryFile(delete=False, suffix=".json") as tmp_file:
                    response = requests.get(embedding_file)
                    tmp_file.write(response.content)
                    tmp_file_path = tmp_file.name
                vector_store = SKLearnVectorStore(embedding=embedding,
                                                    persist_path=tmp_file_path,
                                                    serializer='json')
                ######################### Save the vector store to the session state ########################
                st.session_state.vector_store = vector_store
                return vector_store
        else:
            # Load the vector store from the cache
            return st.session_state.vector_store
    else:
        st.error('No content was found...')
##############################################################################################################
###################### Function for retrieving the relevant chunks from the vector store #####################
##############################################################################################################
def retrieve_chunks_from_vector_store(vector_store, re_written_query):
    ########################### Perform a similarity search with relevance scores ############################
    with st.spinner(text=":red[Please wait while we fetch the relevant information...]"):
        relevant_documents = vector_store.similarity_search_with_score(query=re_written_query, k=5)
        return relevant_documents
##############################################################################################################
################################### Function for retrieving the chat history #################################
##############################################################################################################
def retrieve_history():
    ############################## Go through all the chat messages in the history ###########################
    for message in st.session_state.messages:
        with st.container(border=True):
            with st.chat_message(message['role']):
                st.markdown(message['content'])

  • Line 1: We import the os module for ...