Complete Code

Load pretty print library and ignore some annoying UserWarning message. These are optional steps.

from pprint import pprint
from warnings import filterwarnings
filterwarnings("ignore", category=UserWarning)

Remember to create reranker_sample.csv, according to the sample provided in Appendix in this chapter. And place it in the same directory of this code.

Use CSVLoader to load the CSV file and include only the specified columns in the page_content of each Document. Note that any columns not listed under metadata_columns will be included in page_content.

from langchain_community.document_loaders.csv_loader import CSVLoader
loader = CSVLoader(file_path="./reranker_sample.csv",
                   metadata_columns=["ticket_number", "date", "caller", "responder", "timestamp"])
data = loader.load()

Define embedding model

from langchain_huggingface.embeddings import HuggingFaceEmbeddings
embedding = HuggingFaceEmbeddings(
    model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
)

Embed the selected data into VectorStore and create the base retriever. Note that the top_k is set to 4 which is the default value. 4 entries mostly closed to searched string will be printed out.

from langchain_community.vectorstores import FAISS
retriever = FAISS.from_documents(data, embedding).as_retriever(
    search_kwargs={"k": 4})

query = "if i forget the password, how to resolve the problem?"

## the following snippet shows the result of similarity_search
# docs = retriever.invoke(query)
# pprint(docs)

Define and load LLM with HuggingFaceEndpoint

import os
HUGGINGFACEHUB_API_TOKEN = os.getenv("HUGGINGFACEHUB_API_TOKEN")

from langchain_huggingface import HuggingFaceEndpoint
repo_id = "mistralai/Mistral-7B-Instruct-v0.2"
llm = HuggingFaceEndpoint(
    repo_id = repo_id,
    temperature = 0.5,
    huggingfacehub_api_token=HUGGINGFACEHUB_API_TOKEN,
    max_new_tokens = 250,
)

Define Contextual Compression Retriever. The LLMChainExtractor uses an LLMChain to extract from each document only the statements that are relevant to the query.

from langchain.retrievers.document_compressors import LLMChainExtractor
from langchain.retrievers.contextual_compression import ContextualCompressionRetriever
compressor = LLMChainExtractor.from_llm(llm)
compression_retriever = ContextualCompressionRetriever(
    base_compressor=compressor, base_retriever=retriever
)
compressed_docs = compression_retriever.invoke(query)
pprint(compressed_docs)

Generate chain with compression_retriever

from langchain.chains.qa_with_sources.retrieval import RetrievalQAWithSourcesChain
chain = RetrievalQAWithSourcesChain.from_chain_type(
    llm=llm, 
    retriever=compression_retriever,
)
pprint(chain.invoke(query))

The output looks like this:

{'answer': ' To reset a forgotten password for an email client, follow the '
           'steps provided in the context: click the "Forgot Password" link on '
           'the login page, enter your email address, check your email for a '
           'password reset link, and click on the link to create a new '
           'password.\n',
 'question': 'if i forgot the password, how to resolve the problem?',
 'sources': './reranker_sample.csv'}