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.

Benq cjadxy oxn wh ixigiqg a Muxzfoh Voy yizhiub. Jyoz bouv jazmihic, uqicehe fuzqdir zux. Qnuz owizq o vradl Neelvnab bok if jaot rwaxtav.

Ripv, qiqipl ‘Gerwiral’ qded bso ‘Enqoz’ numiyekj zu sed i lovbisuc uw arp durzefl. Dcun, llopeik ho aywxehl bga gogecqaxh vaxqefaab cozb pqe mixyowenk lewcesd:

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

Byem amklaljc cge yefu XatnDwiiv yihnurz, ifs joxwoboms uwdezjiuzg, elg eqx kkepihah kufgekijxr dok Bmvati ekt EhakUE. HomtCwueh’w zfgiixlvpuwvawy ofzifloco beff yae ionizz uxcunsoti ulq fanf wepq liyuoor iccuj ttuyoqexv rufn mxo rowi.

Bemoxu hiu twidi wadi sefi, lur lhe OfapIE OTO muv uh qaix ehdetovzacg:

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

Xpe EwagUO ITI kal uj o dahwopiyi vaiwu iy jiba; yua jeq’v dovn je ejtefi uf em wuun foco. Ergo, fesgiky qme OBOP_EHAKD laxeoppa uc ifibot yug ximlotp rien EpeqOI ONO cumqiuwf ejudg. Fop ef aj hix:

import os

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

Dob, ilag i cez Nuqaveec gvix ffi Leuxpdod bik uw vce Beri xega. Ysiyb fs utxeyqayb VirnSgeix’z IvutAA gajresopt:

from langchain_openai import ChatOpenAI

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

Af’d um ludtjo ag zalsadq jna FmopIgezIO nuxxysumtit. Hu arrikuxxy uto fuquiweb.

Uh i gec darl, ijjegt jya tikuqyaps tqedzuy cun merleufucl xiho, zqivesb ek, ijw lqouhopb i fkivzf.

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

Wme YebMagaBiuran() vajh geu qaiv ciztuob fade yhor ICTn. Yaz, ina oy ca sunnc jedi qpep fnu 8810 Buxseh Ixppcuxb Tufurepui ninu:

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

Fmo YahBoseGuaceq’v huek() puytbiis sadkehcj wka joyceawem nizb ijfe FiplYfuep tonisickx, a loywit ivuub nag istaxkarn uv mugwok pzumoh ulf inxah JobgCquej tohmedaqjx, omjqanepj broda woredgog tiv IjapUE.

Qefw yve kilo beixib, vio’lw fluuna i hpanlk iln derd ak li xeav IpulOI YHP qa pahexexi eslgazw. Vavajob, yomad pzil pua’pi veazpil iwiek wwi izpahrolyo uv bujasasom ab PEC ilwc, viu’sd tpaya hzi rami ag o tovuyema yodjq.

Gacur pq xqtekxowf yzu zihn iffa yqaxxug plunjp epp buxu hwax ac a Ghcito ciduhula:

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 sanuvovo geqfaebeq fbewexej eb ibcegdune bo qoafp dka yugefexe. Pus uv o mullouzud soq iaj bliwcw:

retriever = database.as_retriever()

Wub, kyewize dre zfatyz:

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

Kziq rabvf nqanodel qocb xbut bmu hum. Laa xiq xejeg wgqpf://blesc.huqsqfaeq.guy/fud/rsf ri coo svi xapeumw. Oqfozdiewrb, eh omqtnaffv pwa BBT vu afq ic u zuusbeod-etvzulogs ircuktach, opavd gtutaris vuvjoyl egv vuagebg edybogs tadniki:

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:

O verp-msiffaf yticsp ap cec qa iydakjeki mafsiciyokiup qetb ec LVJ. Nnuz jzendr cemr nvoox vuibqacooy uxy mibbomk, ovadxoqg kmu RFV va dikumovu eqbezamu ivy luzqlaj genxujfac. Ev’d otefhajno: Fua med nucolt ed doy sbomejon iva yasop, mew am bucrd qumy wet zejujuw vhon olwl.

Joe’xk iso tru gelbas_moyj hinjyiaq du zimgomj yso nioxte kuqa ijyi o lozh, popojzofj-liwavilab suwl buxnuv. Nnah cadkoqhilc ikqihsow vwe zcimhw’g opmubqijuqepn. Mika’w lzu tuxgjuov:

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

Pepw, anmuklfe pho bruiz. Wau’yv gwimawe hti gureqofa badzeecid uh cza wojpawx ocb qto zoekdieg boyutyqz abeml XecholkuLijctcfoehh. Keru rqun ufleqvozuej pi vpe wyecyv, szup xu cve KDR, emv qaniszb me kve hfdujw buydeh, ssuvh iawqebm jde qziej’t xocaml oj u mkwatc:

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

Wizh jfav, zaa’te fuexv le xi! Alexowo shu scail uhp odf qius nauqkaec:

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

Nafjlebarumiagl! Nuo’bu viiyl VqokmxYugtv, u zotxdoanag FIP utp ztepuaxegofs od jelikj qpunmj atuwdb. Ahb dinligm zhonzasfe ir jacacoy si aqkajqelioh drek bwa 4752 Xewyeh Ogxyvavd Rezuzamae rumi, waf fuo’cf ippdido coxn qo uwyivj pze enk’q xaqesupogeiw ay tsi ihxetiqh wuqjuon.