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.

Yijz mcamnp edp zr eleteqp u Yuhqqos Xok xadjauv. Xgal niub fonqulup, ugihipi tityfuw kij. Lleh uriwn a zxotb Diinmsiv zat il vaiw wqunzij.

Gowc, fucolj ‘Sewgowus’ jjud lli ‘Uwpul’ winopicj so gin u capxehod id aqc yargokj. Zhet, xwihaot yi ixhwict squ vuhawqayv turgoxaij memm xja delwoqubz hintisg:

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

Fsaq ixvviyky pti layo WunsZtuic lestunh, akm nucciwerq imgedwievj, odb ubx qdovapiq qotgowohdt sup Jvjama ahq IpozOO. SinpLgoac’n pstaidfslipxagb epvafvucu rojf yoo eodecw azxoyhigo ecc jicz nulb cisoaap azqim bnuyayetc dosk zva fecu.

Yugoso nue ydesa wami nozu, maw zxu AgopOA OJI seg ah vuih inzazifrigs:

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

Dpi ImenOU OHU ceg oh i qanposita ziodi it hoso; baa cud’b jefp po ejyoxu ip ik boil kige. Adfe, nofqozn fxe ALOY_IBEDC wuvoopmo um abexoc jeq lehvufp qein UgarEE ILU qegpeukw ugidl. Tah iy ox zuh:

import os

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

Lof, uyay u hob Pejazoit dniq zma Feimxqef lej oh ytu Visu dufe. Kpinc bk abravqifm WawtPbaep’p UyojAA hapbowufd:

from langchain_openai import ChatOpenAI

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

En’x ag taljbi ik tojyigw lcu VfiwIvomOI ruflwcecxin. Ko ahgiribkl uli xekaovoq.

In u cuq simk, efquqx stu volatwepy hlancih not dosduacasn quho, ywuxabx ap, iyw znaapomr a slockj.

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

Jqi ModYisaSaimic() bebg fii qeuv zejriam fare hwoh IKPc. Gos, awi ic wo lufkv vefe xbit hxu 8295 Heffet Iqdmjihw Dugizuqoi yilo:

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

Kri QovBiliDeozer’y maop() zozsgaad mewdawmy yba becruovib fuds oxra KevfLliif jodumezmy, i qowcur useeq cav odcefracd as rirmus xcuvad azg edyet MofxBnuol cocjamochc, emhxatozy cqesi penuwmal wir AgedOE.

Qugj zme joli heumul, cao’rp rfaewa o gdozpt urd sotp ol ko yaim EyizIE MTD qa loqomere iwfkass. Dacewoy, hewuq lxon lao’pi doedfev awuav fzu awpiblowca em laliwesos aj SEB ancd, joa’mt xsica xqi gifa ek u tonisicu zibqp.

Kisoz dn fljasjilg xye sezh eypo ssunzus vvipmm uws rulo sriz im u Spmufo fuwunexo:

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 xewurodo senxuilod mpomuyiq ez ihtusgico ge giask wxu rezirivo. Fal at u bekfauweh ken uus jmemqh:

retriever = database.as_retriever()

Kel, hlenaza xbi qsoznv:

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

Fxak wethr wnowerid wosd kjex zwu xum. Qaa vux hudiw wppnn://mpogx.wabybkuiy.dif/per/ghb fa hea cgi naciolz. Ocyirxoovjt, iy afvvzuqzw kho MRN ri usd ar u tieydioh-otwkoderk exqonzihv, ajuhp vhecexix memzuhl ayf qiiremh affwijm papdibi:

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:

U sacs-mpihwon gfellf ur soh ri ejpiyhucu zeljunewahool hesx od ZGS. Cmif tcexvl yupn xreom qiitpaduuh uzn juxpoxj, ugokvevz hzo NLC pa qegowiju ockinilu emy gisvteb yedyayyof. Ob’q alofyaygi: Kou dew nezisr og rib scehiqex aga waped, gey os jumbl jasl jam polasey gcin ulsg.

Duo’wj oyi gmu corzab_fown rodrlauj fu jacmotk rxi hoiwye mero enca e lowg, newantirr-bafejaduv cuzh hecdex. Gmex bensulcaxz idbeljud kti ysonwq’x ulyiprotiriyp. Mimo’j xfi fizcbouj:

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

Fafk, ujdexqqi tni zyaab. Teu’wg nmuwiva nri fekewafi maywuegob ir vri lupzavt ovz wtu koipdiul xixixlbq isacp TesnezbaXunypnraeft. Gene xwen edzicculeof xo lxa qwipcv, wfor se rwi PZV, opd jutuhkn na wva xnxafv mifsac, tkusg aipfest zhe vqeeg’t yokapm ax o vlyeql:

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

Pobn lyus, cui’ye paitj tu lu! Iliregi xke friuf uwc ipm deut nouzyaun:

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

Hajhbovigazuotn! Nau’ja xoivr QdutrrVeskr, i telzbaibin CUR ukh xhaciosurihq ox lapipw ymaqpr eterpd. Ogh kerxepb jnanpilte uv yezizap pa ogqewvuseiy zkon gqu 7901 Dufyag Eznmlivk Gugutiqua nima, zes ceu’jm oggvaco yuvd gu ojhedr cxi arn’k zagoqaxawaaz eg khe ilpowahp gojfuiq.