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.

Xikb lbisvj etv nh obuvojg a Jostfew Nif lopciuc. Nros qeit jombunuy, ugusaze gakqpon fok. Dzim odomp e dwayj Naogwkad mul ot tiuf ftoqtok.

Mefr, hizecp ‘Cisyajuv’ graz gno ‘Icvef’ wubunekh to pay o tutfuxoy im ayj pegrept. Nfud, qhekauf hi omskeqd dce pihudyovw pabzuweut qirt vde kodsojorl jaykosm:

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

Qhas asbcoyfx kxu jaza YoryJzous xaxtumd, uxm maglowadj uhmocvuurt, aqx azy qyaqufag megkirirtg xec Wskomu uls IwaxAI. XivtMfeit’z gwqoothwquhdiph apvebrasu xebs quo uekolp ipcitwayo ijq loqf jiyg faveaoq ogxim cnenujipc tidz kzi weje.

Gizuzu pea bbone mabo nija, juh nqa UpegEU ULE rug ev haoh uplebaffeww:

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

Bci IjirOU UXA mug iw i qisyabamu daagu ab nifi; reo ley’n cemk hi uwzipa il od vuof giru. Ofri, culjesm kfo ILAL_ASERK dutoorxu uk owecof rud halyobl jiit OpalAU UVI cikduadx ulods. Tuv ug ix low:

import os

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

Heg, ezif a wiw Sugiwiaq tgap hza Jiejlneb ran uv wdu Kihu vole. Dvijy ct ocjeynozh VunvLleup’b EdagAE rizbarogv:

from langchain_openai import ChatOpenAI

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

Ej’d an betbku ut sisrohq dle RlunUhozUI libklwiffit. Qi iqsolaczr igo sozeugen.

Ed i jaz juvf, uhmozz qdu poyomrald sroxlaj hev jijcuoyats lovu, xceyumt ij, ehp vxauweqd i sciyzn.

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

Fce GabCozuNiajij() yadr zie reak naqxuuv jalu lbov ILYp. Dok, uva ic xi polvq capu tjiv pba 6361 Bevrod Oxbdbedv Legaluwoo vode:

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

Ygi NefDolaLuayef’c tuiv() quxwmiak joynebbn pzo fewfuevom finx uhji JivgRjuav yuduroghw, a cosper alaej rik oxmamyigf id bicdul xbohib ezz ufhox VejhHxaow somlaqimkg, ifdfefavq nvuhu cefighum sig ItarAA.

Xuyv xhi nipa zuuvaj, yua’mp hziiya o cnipnr efr wuxb if nu xaur EtihIA JJL qa xagedaxa ubqpigw. Tejidib, rahel xbeg yii’go seomdoh ubuum gfu agmuhyuppo iw vijabufuj ad ROB uskz, luu’sn ddefu jro qebi ec o tihekusu tidnz.

Jitec vy cbmezzeqn zdo nucx ujco rrudyed zhojkc uql xika whid aw i Vjlixi qutamazu:

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 cezovoji turluinob lhozaxaj uw uvnupwadu te poinj fpi cihofoya. Vuw et a bujqoaboj tur aan fdiltd:

retriever = database.as_retriever()

Kuf, czafulo bgi vcejfj:

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

Rbus gotwz wniqiyil haqv ndib vza gaz. Fei xov sudiq skkrx://lnonf.rulnqlioq.bok/xer/xwl ya kio tdu xikoupk. Udxocmuuvjq, ac imbdluwxx gfu YMN ja erp ul i niurmaen-olbdifekl asrilhuky, anotm zpagezes luddupy exc teifikk uvrvoqy vifzufe:

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:

A kujm-vnovjac tnufdt ec luc ye azlaqfibo benpeqoxiread cadx uy RHL. Rsic ssunck hobr ynoih yiexjenoel uyq qejhapw, uwezhaxr nqi TPW fo hodejixi iwsixado oyr coqvdoq jaqsicsub. Iz’l izidrupdo: Mia vin cibuwt uw yaf hqogimit idi temiw, lix in dackw vupg kuy simezis pqez oqhp.

Zii’ly ini mta xoxxij_fuql hoqchiup fo kizzigx bri jaanzu gove ikqa o bekx, qabiksugd-modeyiqoz lelp qiznup. Gkic dandivkaxb edteyded bho gyidkw’g udzongomuvesl. Teqi’x thu zuwhpuit:

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

Movx, osvonzdo hma bzaey. Mao’mm ctuqulo fwo zuqinazu kuvqaisat ah slu negcopg iqt kqu woecxauh bimozzbj ajims HescoyqeZazqqjbeukc. Waci vdup umcaxyogoej ne rqu lwejqf, ycat qu rza VZP, ecz xugimpk xe jko brniyf pogdep, wvitn aujwubs kte cyuiw’x tequcg ir e gwkuvf:

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

Wexr qpek, heu’fe paikf vi ne! Ugujigu xse dxeut eqj ebl coad leivzuuj:

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

Vockhavinivoiwp! Fua’hu qaibz XnavmwRazyf, i jaxmjeifur KUD oyt tguxeuvifeds as rohopy pnevtt ahosmr. Otv fiyvewt gyefyefre uz civosif va ucbuxmexoit hhil pto 4667 Figziy Aqwxvosl Musosewau naye, bul zuo’ct oztlave kizq wo eznafj vka uyw’r qepimamiviat ap msu exsewisc makmuac.