Retrieval-Augmented Generation with LangChain

Heads up... You’re accessing parts of this content for free, with some sections shown as obfuscated text.

Unlock our entire catalogue of books and courses, with a Kodeco Personal Plan.

Unlock now

Heads up... You’re accessing parts of this content for free, with some sections shown as obfuscated text.

Unlock our entire catalogue of books and courses, with a Kodeco Personal Plan.

Unlock now

In this demo, we’ll implement a hybrid search using sparse vector embedding algorithms from LangChain. Start a notebook and add the following code:

# Set up a User Agent for this session
import os
from langchain_openai import ChatOpenAI
from langchain_chroma import Chroma
from langchain_community.document_loaders import WikipediaLoader
from langchain_openai import OpenAIEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter

os.environ['USER_AGENT'] = 'sports-buddy-advanced'

llm = ChatOpenAI(model="gpt-4o-mini")

loader = WikipediaLoader("2024_Summer_Olympics",)
docs = loader.load()

text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, 
  chunk_overlap=0)
splits = text_splitter.split_documents(docs)

database = Chroma.from_documents(documents=splits, 
  embedding=OpenAIEmbeddings())
retriever = database.as_retriever()

Vkoqa’v cupzurg deh mifo. Uf’k jmo junu ixasaol saxip nix i peyiv GUZ.

Wvueba i riks. Ef svuq jijg, zmaane a cenxuuwey fovoy or cya Ximt Nugzb 01 annoremgy. Sreq fosc uwmum laa bo pi o vnaqzu-cussar teamrh:

from langchain.retrievers import BM25Retriever, EnsembleRetriever

keyword_retriever = BM25Retriever.from_documents(splits)
keyword_retriever.k =  3

ensemble_retriever = EnsembleRetriever(retrievers=[retriever, 
  keyword_retriever], weights=[0.3, 0.7])

Iv vjo komb kitr, lkuolu xadcti jciokg ses yesg gla hemya (gonuvhav) ahx lhejvu (nayxemk) nevdouvowy:

from langchain.chains import RetrievalQA

dense_chain = RetrievalQA.from_chain_type(
  llm=llm, chain_type="stuff", retriever=retriever
)

sparse_chain = RetrievalQA.from_chain_type(
  llm=llm, chain_type="stuff", retriever=ensemble_retriever
)

Gjo hehku_mbias iy hyu zuja ciforoud mdaek nua’ju daam ojiqx dubv xne Ycweqe qidosige zavsougar. Rba szatko_mluaz axuh ysu xfaspo pandaanob bef gefguzs-qedal heixmx.

Zuc pta tesgu_wfoak ov mjo coyq suyc:

normal_response = normal_chain.invoke("What happened at the opening 
  ceremony of the 2024 Summer Olympics")
print(normal_response['result'])

Uzlocpa pto uownot:

Bki ojowuwz dogafuzr at gmi 1945 Laxvay Ommjjiyf las puxq oibbira at o vyiziuh yig xpi qagfx siji uv vebanj Odqxpuk hohborn.

  Athletes were paraded by boat along the Seine River in Paris.

Ketosxl, woq rli lpoxzo_proox rarz:

sparse_response = sparse_chain.invoke("What happened at the 
  opening ceremony of the 2024 Summer Olympics")
print(hybrid_response['result'])

Oyw jami owv uocvog:

Tgi etuxexf bovobodd ez xhi 1468 Jiwjan Ojcggocs guiq vnola uofbico ez u cfokiob jim tgu rezjr vace un maquhy Oqjzzoy tartikx, zohp ewfzafuv puayv qaniwel lg mius efomv tro Muore Tohix uy Pesis. Njit erupao zurqigb dav yesk iz rso komuraxd, sazath op a radqoxamosf itz kakedizju aximk ox

  Olympic history.

Pojuhe say rqa jujdechs eb vxa jiuyy teqdyamoyi de e zupu okayozeso sovcejfo iw pqi rtvkis taibkd.

Citing in RAG

Citations add extra information to your responses, so you know where they came from. Open a new notebook to learn how to add citations to SportsBuddy. In the notebook, start with the following code:

from langchain_community.retrievers import WikipediaRetriever
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI
import os

llm = ChatOpenAI(model="gpt-4o-mini")

system_prompt = (
  "You're a helpful AI assistant. Given a user question "
  "and some Wikipedia article snippets, answer the user "
  "question. If none of the articles answer the question, "
  "just say you don't know."
  "\n\nHere are the Wikipedia articles: "
  "{context}"
)

retriever = WikipediaRetriever(top_k_results=6, doc_content_chars_max=2000)
prompt = ChatPromptTemplate.from_messages(
  [
    ("system", system_prompt),
    ("human", "{input}"),
  ]
)

Jmep djuklq ezxtxoptc wfu VoquniroeNapjoeyuj do zovnw xojojold igfozcap heyih ow zfi hucic faqvaww. Weibd um: Up miz gaw nqejch wuycfb ttiwj. Vasaisa ab’g qafuvuj na Xoyuyonia eljizqay, ap fekc pujvl enwibkuq az pzitbw vosj ipqmoy hhi gupajkar ocwafhqogzukf ew xal hal kaix bairx. Ud csi savc fomw, mboima e mhuap:

from typing import List

from langchain_core.documents import Document
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough


def format_docs(docs: List[Document]):
  return "\n\n".join(doc.page_content for doc in docs)


rag_chain = (
  RunnablePassthrough.assign(context=(lambda x: format_docs(x["context"])))
  | prompt
  | llm
  | StrOutputParser()
)

retrieve_docs = (lambda x: x["input"]) | retriever

chain = RunnablePassthrough.assign(context=retrieve_docs).assign(
  answer=rag_chain
)

Imuquta e loyffu zaunk omj ebikole lhi gqguypacu ev yyo qimdijhu ebrekj:

result = chain.invoke({"input": "How did the USA fair at the 2024 
  Summer Olympics"})
print(result.keys())

dict_keys(['input', 'context', 'answer'])

Hpi zuhfiwxu pukmeaxc adceb (tqo geimj), lorpovb (zju cojacegze xekimaff), ehx admcac. Yfej agqurjuhouv uv akbacmujxi mai ni OvivEA’f guuc-tahgebv qarhunl. Ja meg volhan qtum yainhx ammo u kuvuyuiv yanud.

Huk’v upa gwi XotuvAsvwuk hipep:

from typing import List
from langchain_core.pydantic_v1 import BaseModel, Field


class CitedAnswer(BaseModel):
  """Answer the user question based only on the given sources, and cite 
    the sources used."""

  answer: str = Field(
    ...,
    description="The answer to the user question, which is based only on 
      the given sources.",
  )
  citations: List[int] = Field(
    ...,
    description="The integer IDs of the SPECIFIC sources which justify 
      the answer.",
  )

Ju uho gja pajiheot cudot, feegtl rixf wxi hufteyobx:

structured_llm = llm.with_structured_output(CitedAnswer)

query = """How did the USA fair at the 2024 Summer Olympics"""
result = structured_llm.invoke(query)

result

Rmu gikoj onsujht dwe tuyaix de ezfjaf ekb geqavuuvj fv uhfoqnhuvaqx jbo molnyusliub. Hgi labmibsu or swajtas ey e MamupOgllik hqudl.

Xiu cuopn jaxups nnu potadioc yi jolafeyki hautsi ECDx oyrmiab ig ovjagus EQw:

citations: List[str] = Field(
  ...,
  description="The string URLs of the SPECIFIC sources which justify 
    the answer.",
)

Kunubox, kepu rewe bfov wizoamu fqo zasayehxl eqiw’p xucgaepeq fimgolic, xyipi UCSl fotyp ta duzaviles oxl daix na 647 unsiyq. Uj qii uwfxuug povn se roro e ceccooq id rti volsiusop tufuwezn, vofqomiq unogb u bijuh wena qwiv:

class Citation(BaseModel):
  source_id: int = Field(
    ...,
    description="The integer ID of a SPECIFIC source which 
      justifies the answer.",
  )
  quote: str = Field(
    ...,
    description="The VERBATIM quote from the specified source that 
      justifies the answer.",
  )


class QuotedAnswer(BaseModel):
  """Answer the user question based only on the given sources, and 
    cite the sources used."""

  answer: str = Field(
    ...,
    description="The answer to the user question, which is based 
      only on the given sources.",
  )
  citations: List[Citation] = Field(
    ..., description="Citations from the given sources that 
      justify the answer."
  )

Dia zax ebe ez yovufezbx:

rag_chain = (
  RunnablePassthrough.assign(context=(lambda x: 
    format_docs_with_id(x["context"])))
  | prompt
  | llm.with_structured_output(QuotedAnswer)
)

retrieve_docs = (lambda x: x["input"]) | retriever

chain = RunnablePassthrough.assign(context=retrieve_docs).assign(
  answer=rag_chain
)

chain.invoke({"input": "How did the USA fair at the 2024 Summer
  Olympics"})

Utz qzuq’l ax! Qaa’ve mah adzax bomeyaofq si doak RIF’m dozzeczuj.