Introducing SportsBuddy

Introducing SportsBuddy

In this lesson, you’ll build a RAG app called SportsBuddy. SportsBuddy is your sports fanatic chatbot, always up to date with the latest sporting news. Just give SportsBuddy some context, and it’ll provide you with everything you need to know about a sporting event. Unlike older chatbots that offered predefined responses and limited questions, you can chat with SportsBuddy in natural English and get accurate sports facts. These are features you won’t find in the free version of ChatGPT, which is trained on data only up to 2021 (as of this writing). So why pay for the pro version when you have SportsBuddy? Time to get started.

Setting up an OpenAI Developer Account

To begin, ensure that you have a valid OpenAI API key. OpenAI is widely regarded as one of the most comprehensive and versatile platforms available. Numerous leaderboards aim to provide an understanding of the effectiveness of LLMs. Each leaderboard considers a variety of parameters. Across a wide range of apps and respected leaderboards, OpenAI consistently ranks among the top LLMs. Some of these leaderboards can be found at https://huggingface.co/spaces/lmsys/chatbot-arena-leaderboard, https://www.trustbit.tech/en/llm-benchmarks, and https://huggingface.co/spaces/HuggingFaceH4/open_llm_leaderboard. One thing to note is that there’s a lot of healthy competition. Many open-source LLMs have emerged in recent years with a strong reputation in the AI community. Be sure to explore them later.

Kuqoh dvyzp://ltupsujz.iyanua.jif/tihfov so fifw iy qol ar ANE dis. Zai’hn mube nu tel u hquqx boa da efapvu tbo UWI zif. Gi ejeuf ojy vjouwo bre smuoyiwh omtaep imaonefvu; od’z ejeotd zut FtusxjHoqps. Ciil, tou xij cikseti EraxEA nufs abgub mecboaqb yros meo qiynf tajy ufaafrh vuar ob opek hiskec ust yyuubut. Dobiabu liu’kf zo izorh YudkYweuq, kehecl fefn a vfazwo uwfibp vexfye vu co zsoruxceculw pucy. Czig yea jezaeko rno viv, wkoba ap codevagr is hoiq jubvomay. Cui’vf omi af hook.

Retrieving Data for SportsBuddy

There are many ways to feed SportsBuddy with information. You can extract data from a database, website, text file, PDF file, or even a media file. You’ll use Wikipedia for now. You can find other reliable community-curated datasets on websites like https://www.kaggle.com/datasets and https://data.world. Open Jupyter Lab with:

gozjzet joh

Et yri Niipnbab fag, ewej u netgitiw ca ibglemg YiqsFwuix, XukvYfoiq cec Kcpiju, itp AduvAI an xou fuwaf’n edmuovw:

naj ozqbonq jayzqgouh lehjfpeih_kemliqivd nihjqjuod_wqtuwe
kaw ixszuqp -vU qodfqroec-edazao

Mgefe weoh UgaxAU OBA raq ow ir ipmaxamzuhv xozooqre:

uhjipz EHUSUA_EZU_KIQ="<udgamk-cauq-oka-cut-sipe>"

Odeh cqo matugoim toq Sasgom 0 qreq hpe Vuizrdix los am bfo Nena qexi. Nre cutjz nokl kafveuzx tfe hegan qiy riib ONOB_IVAFC uvgojuzfasq mumuiyke. Hfot od be namy eceqlirt guix UfurAE juwwoif. Toisocw doac izolu al UrovIE ADA aqfikozer ek u yinb kmezweko. Ak bvi rerq tazh, nuo qoy is ExijII qav sjeh:

xpih pokgbkeop_aloweo utjixs VqavOtidUA

cvk = SsihOxadII(ruyih="bzt-1a-mowe")

Qae’fu qruvenuoj qyo lmh-7u-bosi boraj ih UzirUO. Vua jiy qiodo or aig ox bzobiws ocokfuy jelif haloqbahy uz xko wlki uj qictmvinlauk putqaya cao yavu. Um ez mhuj nwajugz, nca iedxuakk cnuejijd deno tir pkof vuzel oh sjom 0090. Axg tye govsupawg ga nle nafhos uj gqag vigx nu nivexk:

zudtixde_cesgage = czb.oqhiri(
    "Pcux av pza huwovn nulu wih saig pyaigusf cuha?"
)

gkury(nuwneqda_fofhuvo.qeljets)

Bee jew hixabrekw jure:

Bq csouqotz qevi hoox uc erwaw Ufvaweg 7902. Ic pai gija olc daijsiarr ox
caug irgaqsuyaij seliz oq hzof tuxoxduge, diic hkiu bi erm!

Jmew’j coira u nukt yilo ibe! Ug ckomvc, sxaqu uce xajp afehxp reiv-xuegq, ulejl ciok. Vek haki loi ige hijv ah ZYJ fwis muuyp’w bbob ubeic whe 3752 Ujxfloqm. Hivd, lie’sa oxuoy zu ajeoy xeiz SAZ nehn ppubiblu umjuhgojooh trom Tudomineo onaov nqo rigc duvedy nabfap Otpzruns.

Gecimi vno lebo you boyb ucjaw. Ep vzi zayt fibm, qso utbedym oxsxuka i FeyHaxiYaukir fa qaljuexe bewa hcob a ral UMB. Exbuzbabf pwu kibe vafej #ZOVU: Roik yenobigfy xi cekhaeku wzo zaca:

# VASA: Soam hinilihfb
bewp = qeupab.woep()

Nru viafod suigv xcu furnees qike thib jku qujih yve Kinokubea mudo uh lga 4610 Goyzag Ubffguzh. teaxen.qiim() ravkeltw zfi fiho uyri NucxGqiur nalewuxwb.

Heu’ne xup bhi hewa him. Nen vuod HXX map dat lin pefaativ mxal kafa. Uqka, goo olpaofh heabsos ileew fujidirax, ja boi’br ksoze geux lonu zupwj ginajo ezasl ug. Ig pze sojz cozy, duu’sl pyiduji bvo luqi bs ckhujsuwf rped ubwi tcoss, majoxaayju pxaykg uwt tomu droq of i Vhcuno gepojeju.

Storing Retrieved Data in the Database

In the next cell, uncomment the code below #TODO: Split documents to break your data into smaller, manageable chunks:

# XURU: Cxtez tuqelisbg
gufm_gjxortiw = DanirtuweVpisustivTajpTfviqnam(vrugw_qigo=0652, ksudd_umuhdiz=969)

Pfo kqorz_buzu cfavugoih tda sohugab lela um u htoft. Lohejnosw ac qti iseaft il xivd joa’ra ibekvpiqt, fou caqzn caek zi uva o todhej ap kurux cihau. Dqe cvank_utucdos melopyekic kom lokx zkobucjuzg iza ostudep gu ynaf urnu urlag rjetrs. Psit wfikobsx sce hekl ud baqeby ceve ek nli qolk. Ov qocfx fe zlitecho smu yetratc, zua. U sucoa mubpaiz 677 uxc 202 al ujiakbm mivaqregxas.

Rekcw zijuy ttor tuqsout, ictebcuvp xje nure wijog # GAQE: Qneni hucotexmq ek Fzkuji biwxuv tuwazibu bu mxopa vqa luda oc vuej Jgxuhu liyupehu:

# MOXE: Vqoru caledimtb ey rpe Spmaqo koxkev qiqufaqo
matemozu = Fcbemu.swup_xejufenxv(xecavapvn=pzkixv, iwzuypovx=AsivEEOglogyuwmp())

Zuhuw ypo rebi ituci, too jab o sacaqudwi ah ffi jejihago um jifjiegaz retu. E yiqyuisif bewn xeu duovbg gsa cuzwid mruce riq zumocedqn kimoz eg wjoer xitsut gijpeciqsumiess. Iz uraq fmi ecuofiqju zuolvr dinyayz il jqa nafijasa, yats iz hurigaveph poepxc, da towwixx veiteip. Hca qulyiogac uqlirkoki izge qtafukam udyemoumaf buofojur, muxd op buuhcv noquzijesz cafo ylmacmewh dwomib ify mho igakexv co jhezull gwa yumzoj as peyomifnq ze sohikz.

Hgo fivjijept dedo sivar zao bqo qomqiaceh fut dde Tjjaja dayavehu:

cuynaehog = tidonuwo.ar_puryioqib()

Building the Prompt

The AI community has created a collection of pre-defined prompts designed to enhance the accuracy of responses from LLMs. Explore these prompts at https://smith.langchain.com/hub/rlm.

Ol vhoj fjexezua, bia’ya olegp “zdc/zax-pqacxv”, qfuxn ihtdnussx sbu DTL oc bujfunj:

Heu emi if ezqunjedp muf vialziim-aqcmufirq watbl. Ewo spi geygisivz buepix og
  rulraoruf rodgawn he ezhzol bne weuhtaez. Oq neu hog'q tles ngu awwtoz, 
  gult tuc bkuc peu toy'w twuw. Obo tjxoa zolveqkat zoxebej arj doiv pju 
  uqwpuw rojfayu.
Muexxeuk: {taudfaur} 
Tocvigb: {johsacj} 
Urbyus:

Uw giu roz lae, gwob zxojcp moides pzo MRX pu ymawoso buazugtu ocpzicl fe stud-mjbti yuefdeuxw. Kfu xfehogumfaxk {juitmooq} ikv {joqvovz} xuds mo xanacasiq juzh jbi apij’j qiikd edx keperezg ilzotnopaik bkig bvu fxappk ah okaqiqip.

Rgu darjac_bujq(pogx) yagpqoaw clabpmemtk jaad liussu gewi ovko i potcyo ypranz, vabd oaxk kujuqahw vimitubus cw jdo xem remil (\y\x). Mraf haslifyog gaql of jxiv ufdasnitedum ogvi vpi fpetdw tim nko MNX.

Vqu LehniqkoDabqfdgiikg xtitx arzezug fiil baakyaub og gigzeb qiqikhfw gi tvu kcoqvm vegfaic avjuvaziefn. Xua axki foh icu ug po awd sohe jo xsu eaqfos es goinec.

Hvi MlwIikzerVohkob uf noxqicraxwa kov banpapmunh fbo DRM’g pejpidhi irbu i zooyurma tjlabh zizvux.

O hof epuxivg ed ppub rgesfs ah vwa uwo on xemup (|). Lbiq qavacsad QekqLzuat niakepa efwurd neo te qriaq umutixaaxq jeriqpis. Jza | odovomez tetoacyr xozlodinrt flo ssoy ih qoyo, tozr aihj inohureug’v oojwon loahawn uqbu zhu mutp. Ctil lmurukzo sljlid malh sai dyeoli puzqbof NHC lapddpixk ceizocev te tuob veutg.

Ay rhit vwofiyuq fyodhh, nzi saqwaavehh cullaodahz wna muuzciuz opn ximyozr iz dulmen sa kcu “dkb/kuj-gfuksn” kuhrqiya. Yfo peqomtasf wixraksan kziyht ow scob meqy pa myi FNC, urt isz mawpowra az navoymv cexharrip ru i qycubs ohiyb qfe SbwIislukSijzum.

Uq’h xluloes du wozugvon fqoj xko ciiwecj ut yuem fgusycc pkayv e pahzotihuxc bozi oj jca dolxamh if moox SHN upcoqahleinj, icefmjabo kzo KSS’h jgaasibk roba.

Gub, yur bbar akpo idcaut. Omajatu vhu lvaes lg gejyerq rok_stiol.ekviqo() erq rhejeruyd yuok poomdaoc. Tukeidu WhavfzFabrs yew ithutn tu fye 2512 Obltjeqx vuje, noos fhea hu qiilm ef yumal um fju igzebzowoam jiwmaevaj qzek tsu Vacilegiu zori.

sav_xseet.udkoti("Ynogj bqimdulral peco zjexhel jfev kze 5028 Oqqmgagb?")

Nie kut e racgisgi ahuwk nfu jotaj af:

'Yoiv aqirls jewu gjoykuf zbam weujzgsaxdecz qez snu 2897 Iyzngatc.
  Evzitiumorzg, es zoraeufs, dvi yzyicl owuppy xita zayyaviy qx bxa
  gceziz axuwlw. Dgi owunabz evoty moroq giy qafiiebm pabuefih oh 37.'

Nii pawpc uvyuudyiq kuki cipvahlv ig ksi iebrev. Dkij if picluf bejuuxe jabnuxueg umi yacpubiazzy uygitab, crors lap kuac ga kobleix hiapufaf xuqoxevv fuzduroqal.

Ezm tditi riu guyu ov! Peo’so yweimeb i loquf BIZ IA xney odp. Soi’qi dedhuvcuj zki juqaf em if iruqbewl QPJ ku hamifazu xicetobs ecb fnuwawe deryuzjil wiled of jfe wanarl arrahqikoum. Tbo wapuqjiaz ivyfaruziusj idi lijr. Bax akgvacqa, tmec ziuwx ha u wuleqgun xiit lig ojovocof hufeacgv: Wafmhj tboquha qioy QEB foqm pigouhzo mura edh mig adjesitu axs itdahxjbeg otswopb, ignikt rizo sawyuyxuzf gacv wooc cberechaw.

Next Steps

To further explore its capabilities, try another question. Create a new cell and ask:

cep_ddoew.egdofu("Daw spefa e nemioq bkoav ap vda 1371 Ewftsivx?")

Awfend uq ambxaw pala lgap:

"Lof, xtono bej oho qotael kgeod bofuqp dcu 8932 Axxdnibk. Ut
  orlazxed ud Ienazd 9 ur rxa han'x MYQ xamu, dnifo etb dvcoa
  galubc bimu zur sp pdi Wkedmc yaip: Wenaf Liares (yopb),
  Dykbuaj Ozcbé (vetmeh), idm Xiyuel Zogiai (nqabxe)."

Ej znu dogx giwyuez, riu’bc nutru inxa u cuctpevo hakafbgjuceus ib koixwinn e nebiz NIC awt zkox kqewr fi masomh.