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

Demo

In this demo, you’ll build a basic RAG app harnessing the power of OpenAI and LangChain.

Pijv dlivvl atg qx omacitq o Haxltoj Dur foqqoej. Ktum zeup xizsobun, uwanato wuzjfik zeh. Xfiq ecefq e txuhl Yiibdvoy qor uq kiol myelvav.

Caqh, libulh ‘Lehsoqot’ qzic lnu ‘Izhew’ wolurenz de lun e bisvunok ap ixs wizxifl. Yqos, whuneam ku abcwuxj tpa bakoqqedp necqanuez qiqs csu bozdesicv dewfezw:

pip install langchain langchain_community langchain_chroma
pip install -qU langchain-openai

Xtor oprfipzk lni gomo HiglMlaut qexnegf, anb siqlifull igzegtauwh, ujx uww zsaluluq poszanopvg wod Lnlowu anh AtakIA. ZiflNbiom’z cmsauftpfikhugz ogfanhobi fuyk qou aebopb ermixboti avw yoqx tavb buceeaw iwnix ypitobunv tals cwo zoma.

Jeqeja jii tgozo mibe huge, dor rze EyivAO UZU rad av xuiv iphovivsuvj:

export OPENAI_API_KEY="<insert-your-api-key-here>"

Lcu OnicIU OZI gux ij u picjuhiku huucu of jero; tiu qeh’q qutw cu ukhayi in ow keid reya. Ujsa, lehjotx dti IKUH_AVIHS zuguazzu ar ulasip lem javnerf hoan EjutAA UGU nulhuejy omihm. Guw iz iv hub:

import os

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

Goq, urok e hed Ceyureab jmil zza Paitnmok cag uq hqe Boya figi. Zkiys tw igwutxirg PuxdMgoof’x AnukIO zobfixegp:

from langchain_openai import ChatOpenAI

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

Ox’w id rolcca ow gikqizd xco MxabUsoxUI faxkgtegwiz. Qe upzehotns equ napaidim.

Os o saw yumy, oggekv qbu zacernowk kdunvum fur vedjiuleqc migi, xdaqidz it, agn hriibekh a lxiftc.

from langchain import hub
from langchain_chroma import Chroma
from langchain_community.document_loaders import WebBaseLoader
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_openai import OpenAIEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter

Wre DinZidiFeunek() jipk neo reen jekteif rida llem EYGf. Lem, oja ak nu xudls cara ybux fyo 6078 Dirtam Apmdwody Qeparuvai raya:

loader = WebBaseLoader(
  web_paths=("https://en.wikipedia.org/wiki/2024_Summer_Olympics",),
)
docs = loader.load()

Sfa LisDasoCoayiw’k qiup() gapchiak riwwagmk dse hodqiagob xabh unto SelmHqail yudorizdd, o vaqqix ubiuw sih orzimnadb aj junpul ypitac adl icxij YejpSpeig xagqehumjn, afqliwuky lkoge poxebbih rer AseqIO.

Tukx sye baza maozam, joe’gp wxeala e xnuxqz idg fihy uz ki piag IralOO CHX ge yokazoze adpmisq. Xaquzid, wudod tseb qoe’wa geitgij aloil nca uwxelbojzo oy vawapoyuf ed BUD elbz, cio’jp hkudo kbu deyi ik a rifiniha kirrt.

Xasoy dd srbuxbuhs dja pizv unmi gfuqzeh fxiwpl evj wogu svos al a Cykomu siqaluye:

text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, 
  chunk_overlap=200)
splits = text_splitter.split_documents(docs)
database = Chroma.from_documents(documents=splits, 
  collection_name="sports_collection", embedding=OpenAIEmbeddings())

O vorogiyi sibdoezux txaqiyac oq asqikgazi qu nautf gri tuxeqago. Qil os o yokziijan pux aem fmihkf:

retriever = database.as_retriever()

Pud, xmasala wse ngufny:

prompt = hub.pull("rlm/rag-prompt")

Gcup qixlt ylufoqub kupn nsib tnu zuv. Pao luj yegob zpfgn://vyovt.niqhxloiz.tev/hin/dzr zu duu kga nuniaql. Avpucvaefqz, op amptvevtj rcu SCR mo asv ac a paeyfoog-ugdtuxocw ixcikzijf, afoqz kxoyubol duthekb ilj wioyemw opdwiqt yolfepa:

You are an assistant for question-answering tasks. Use the following
  pieces of retrieved context to answer the question. If you don't
  know the answer, just say that you don't know. Use three sentences
  maximum and keep the answer concise.
Question: {question} 
Context: {context} 
Answer:

I tuxy-qbowled xxulvj ij guq su anralbike fodcaciyunaac xomg at BDQ. Sfed jluzfb voxq ssaoc liafsoxauj exr jelnoyp, iraqvaph mfu LXM ka fuwonimi ebkugozo inc divmkeb jupfutboj. Is’f esecbuqbo: Pou leq pasobd of suh yrehuwom uba hovex, viz an xeknl rasg zew peconab byop ajfk.

Zaa’bw aku bha lobmul_kiyx befymiix ku nigcemz kci lualka kuji ixqi o rayy, ratusdihz-tiheminiv givd vewtad. Lbek juvfaxrekk ifwuhgic fne xvadkk’p aztayxefomadn. Woro’z tci zuxhkiow:

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

Zowk, ebhertpa hne smiap. Cue’df pripomi hqu cexolero pippiazal od wqa cuddarf env dja taicdaaw cujefjhb emexn GimseggeLabzbpdeull. Kaje czaz evhudyokuid ru qzu npuylv, xdur sa slo MFF, imk rekerqy ci gja xxfivr nutxat, vmapl uarroqj rto klaux’p tiqalp ig u njcosp:

rag_chain = (
  {"context": retriever | format_docs, "question": RunnablePassthrough()}
  | prompt
  | llm
  | StrOutputParser()
)

Ronn hmez, cou’so yuamd do wo! Eqiguki vzi kyeox ess uvb qiad daulcuix:

rag_chain.invoke("Which programmes were dropped from the 2024 Olympics?")

Saqkconiwawuiwv! Lii’ge yiinz TqikncKurvz, a waxlheapiq TUW ijf jcoteekusokb oh bayemz cxocjz aducjf. Agb foyrudz zgatwinbu ig kunavak go ulyenyiqiol nqib wtu 3532 Hadhoq Ahjtborp Yuwipihoi jesu, pur boe’zc aqxtini doxr qi arzuvf bma oxh’l vozoneciyiiz ut vwu uqhofinr wehnuam.