Fia HN: Akɔtagbalẽvi si fia alesi wò codebase sɔ nyuie le LLM ƒe context window me

Wò Codebase la ƒe Metric Yeye Si Le Vevie ŋutɔŋutɔ

Ƒe bla nanewoe nye sia la, dɔwɔlawo tsia dzi ɖe kɔda ƒe fliwo, cyclomatic complexity, dodokpɔ ƒe coverage percentages, kple deployment frequency ŋu. Gake metrik yeye aɖe le tɔtrɔm le alesi mɔ̃ɖaŋudɔwɔlawo ƒe ƒuƒoƒowo bua woƒe kɔdawo ŋu kpoo: context window fit — wò kɔdaƒe bliboa ƒe alafa memamã si LLM ateŋu aɖu le nyabiase ɖeka me. Eɖi be ele bɔbɔe bebletɔe, gake xexlẽme sia va le alesi wò ƒuƒoƒoa ate ŋu awɔ ŋgɔyiyidɔwɔnu siwo ŋu AI kpena ɖo ŋudɔ nyuie wu la dometɔ ɖeka zum. Eye ne èle aɖaba ŋem ƒu edzi la, ke èle viɖe gãwo gblẽm le dɔwɔwɔ me ɖe kplɔ̃a dzi.

Susua xɔ hehe nyitsɔ laa le dɔwɔlawo ƒe nutoawo me le dɔ aɖe do mo ɖa si naa akɔtagbalẽvi bɔbɔe aɖe — si to vovo na wò xɔtutu-to alo nutsyɔtsyɔ akpoxɔnu nyanyɛwo — si ɖea alesi tututu wò nudzraɖoƒea sɔ ɖe LLM nɔnɔme fesre xɔŋkɔwo me fiana. Ehe dzeɖoɖo si me kesinɔnuwo bɔ ɖo nukutɔe tso codebase architecture, monorepos tsɔtsɔ sɔ kple microservices, kple nenye be ele be míanɔ míaƒe code na AI gɔmesese ƒe ɖoɖo wɔm. Nusiwo wòfia la zɔna goglo wu alesi dɔwɔla akpa gãtɔ kpɔe dze sii le gɔmedzedzea me.

Nusiwo ƒe Fesre si Kpɔa Nyawo Gbɔ la Dzea Nu ŋutɔŋutɔ

Gbegbɔgblɔ ƒe kpɔɖeŋu gã ɖesiaɖe wɔa dɔ le nya siwo ƒo xlãe ƒe fesre si seɖoƒe li na me — nuŋɔŋlɔ agbɔsɔsɔme si wòateŋu awɔ dɔ tsoe zi ɖeka. GPT-4 Turbo kpɔa dzesi siwo ade 128K gbɔ. Claude ƒe mɔ̃ yeyetɔwo tu to 200K tokens. Gemini gblɔ be yewu miliɔn ɖeka. Ne ètsɔ wò codebase de kpɔɖeŋu siawo dometɔ ɖeka me hena numekuku, refactoring aɖaŋuɖoɖowo, alo bug detection la, nusi sɔ ɖe fesre ma me koe kpɔɖeŋua ateŋu "akpɔ". Nusianu si le egbɔ la nye nusi womate ŋu akpɔ o, abe ɖe meli o ene.

Context window fit dzidzea xexlẽme si le wò codebase ƒe lolome bliboa (le dzesiwo me) kple context window si le model si wona me dome. Nudzraɖoƒe si tokenizes na 80K tokens ɖoa 100% fit le 200K-token model — AI ateŋu ase wò dɔ bliboa gɔme le mɔzɔzɔ ɖeka me. Token ƒe monorepo si ƒe home nye miliɔn 2 yea? Èle xexlẽdzesi ɖeka ƒe alafa memamãwo kpɔm, si fia be AI la le dɔ wɔm kple kakɛwo, eye mele nɔnɔmetata bliboa gɔme gbeɖe o. Vovototo sia le vevie ŋutɔ na AI-wɔ kɔda ƒe aɖaŋuɖoɖowo, xɔtuɖaŋu me toto, kple automated refactoring.

Akɔtagbalẽvi ƒe susua tsɔa esia wɔa kristalo wòzua metrik si woate ŋu akpɔ, si woate ŋu ama. Kpee ɖe wò README me kpe ɖe wò CI ƒe nɔnɔme kple coverage percentage ŋu. Egblɔa nane si ŋu viɖe le ŋutɔŋutɔ na nudzɔlawo kple edzikpɔlawo: aleke AI-xɔlɔ̃wɔwɔe nye codebase sia?

Nusita Metric Sia Trɔ Alesi Ƒuƒoƒowo Tua Kɔmpiutadziɖoɖowo

Amegbetɔ ƒe dzimaɖitsitsiwoe ʋãa kɔmpiutadziɖoɖowo ƒe xɔtuɖoɖowo ɣesiaɣi — nuxexlẽ, beléle na wo, dɔwɔwɔ, ƒuƒoƒo ƒe ɖoɖo. Context window fit tsɔa akpaɖekedzimademade yeye aɖe va dzeɖoɖo siawo me: AI eve ƒe ɖoɖowɔla. Ne wò codebase bliboa sɔ ɖe nya siwo ƒo xlãe ƒe fesre nu la, AI dɔwɔnuwo ate ŋu abu tame le nusiwo ŋu wotsi dzi ɖo le cross-cutting ŋu, ade dzesi kɔsɔkɔsɔ siwo menya kpɔna dzea sii bɔbɔe o siwo dzi woanɔ te ɖo, eye woaɖo aɖaŋu le tɔtrɔ siwo ana ɖoɖo bliboa nadzɔ. Ne mewɔe o la, le nyateƒe me la, èle biabiam tso AI si be wòatrɔ asi le wò nuɖaƒe ŋu esime tsileƒe koe nèle fiam.

Esia me tsonu ŋutɔŋutɔ siwo mɔ̃ɖaŋudɔwɔlawo ƒe ŋgɔnɔlawo le egɔme bum vevie. Ƒuƒoƒo siwo xɔ dzesi geɖe siwo sɔ ɖe nya siwo ƒo xlã wo nu ka nya ta be nu nyui siwo woate ŋu adzidze tso AI ƒe kɔpiwo me toto dɔwɔnuwo me. Bug detection rates nyona ɖe edzi elabena model la ateŋu akpɔ execution paths le files me. Aɖaŋuɖoɖo siwo ku ɖe nuwo gbugbɔgawɔ ŋu va zua xɔtuɖaŋu si sɔ tsɔ wu be woanye esiwo sɔ le nutoa me gake wogblẽa nu le xexeame katã. Mɔ̃ɖaŋudɔwɔlawo ƒe ƒuƒoƒo ɖeka le SaaS dɔwɔƒe si lolo titina ŋlɔ agbalẽ be 40% dzi ɖe kpɔtɔ le AI-do susua ɖa ƒe megbedede me le esi woma woƒe monorepo ɖe dɔwɔƒe suewo, siwo sɔ na nya siwo ƒo xlãe-fesrewo me vɔ megbe.

Metrix la hã wɔa dɔ si zia ame dzi na mɔ̃ɖaŋudɔwɔwɔ nyui siwo dzi wòle be ƒuƒoƒowo nazɔ ɖo to mɔ sia mɔ nu. Codebases siwo xɔa dzesi nyuie le context window fit dzi la dina be yewoakpɔ module ƒe liƒo siwo le dzadzɛ wu, code kuku siwo mesɔ gbɔ o, dzimaɖitsitsiwo ƒe mama nyuie wu, kple nudzraɖoƒe siwo ŋu woƒe susu le wu. AI gɔmesese ƒe metrik la wu enu be enye teƒenɔla na kɔda ƒe lãmesẽ bliboa.

Xɔtuɖaŋu ƒe Gɔmesese Siwo Ame aɖeke Mekpɔ Mɔ Na O

Dzeɖoɖo si ƒo xlã context window fit gade dzo monorepo kple polyrepo nyaʋiʋlia me kple didime yeye kura. Monorepo taʋlilawo ʋlia nya ɣeyiɣi didi aɖee nye sia be nusianu dzraɖo ɖe nudzraɖoƒe ɖeka naa nusiwo dzi woanɔ te ɖo dzi kpɔkpɔ nɔa bɔbɔe, enaa atɔm ƒe ɖokuitsɔtsɔna te ŋu wɔa dɔ le subɔsubɔdɔwo katã me, eye wòɖea vevesese si le ɖekawɔwɔ me dzi kpɔtɔna. Gake ne wò monorepo tokenizes va ɖo token miliɔn 5 eye context window nyuitɔ kekeake si li enye 200K la, èwɔ codebase si AI dɔwɔnu aɖeke mate ŋu ase bliboe o.

Esia mefia be monorepos ku o — didi tso egbɔ. Ƒuƒoƒo siwo me nunya le le titinamɔ dim. Aɖaŋu siwo le dodom dometɔ aɖewoe nye:

ƒe nyawo
• Intelligent chunking: .contextignore faɛlwo zazã (si sɔ kple .gitignore) tsɔ ɖe kɔda si wowɔ, nudzrala ƒe nusiwo dzi woanɔ te ɖo, kple dodokpɔ ƒe nuwo ɖa le AI numekuku me
• Module-level context maps: Manifest siwo le bɔbɔe wɔwɔ si kpena ɖe AI dɔwɔnuwo ŋu be woase faɛl siwo do ƒome kple nɔnɔme kawo gɔme evɔ wometsɔa nusianu dea agba me o
• Xɔtuɖaŋu ŋuti nuŋlɔɖiwo abe nya siwo ƒo xlãe ene: Xɔtuɖaŋu ŋuti nyametsotso ŋuti nuŋlɔɖi kpuiwo (ADRs) siwo naa AI ƒe xɔtuɖaŋu gɔmesese evɔ mehiã be wòaƒo nya ta tso ƒomedodowo ŋu tso kɔda ɖeɖeko me o
hã le eme
• Aɖaŋudɔwɔwɔ ƒe dɔwɔwɔ ɖeɖe: Module siwo le wo ɖokui si ŋutɔŋutɔ mama ɖe nudzraɖoƒe vovovowo me ne wometsi dzi ɖe ɖoɖo vevi la ŋu ŋutɔŋutɔ o

Gɔmesese vevitɔ enye be optimizing for context window fit menye wò codebase ƒe wɔwɔ sue o — ke boŋ be wòana wòanye si gɔme sese le bɔbɔe wu, na AI dɔwɔnuwo kple na amegbetɔ siwo wɔa dɔ kpe ɖe wo ŋu siaa.

Wò ŋutɔ Wò Codebase Dzidzedze: Dɔwɔɖoɖo Nyui aɖe

Hafi nàdze refactoring wò system bliboa gɔme be nàti badge metric yome la, enyo be nàse alesi nàdzidze context window fit gɔmesesetɔe gɔme. Wò nudzraɖoƒe bliboa ƒe dzesi xoxowo xexlẽ nye gɔmedzedze, gake enye dɔwɔnu si me mekɔ o. Mɔnu si me nu vovovowo le wu bua nusi tututu wòle be AI nakpɔ na dɔ vovovowo wɔwɔ.

ƒe nyawo

"Nyabiase ŋutɔŋutɔ menye nenye be wò kɔdaƒe bliboa sɔ ɖe nya siwo ƒo xlãe ƒe fesre nu o — ke boŋ nenye be si sɔ nya siwo ƒo xlãe na dɔ ɖesiaɖe si wona la sɔ. Kɔdaƒe si woɖo nyuie si ƒe liƒowo me kɔ na AI dɔwɔnuwo tsɔa nusi tututu wohiã la dea agba me, ne nudzraɖoƒe bliboa lolo gɔ̃ hã."

💡 DID YOU KNOW?

Mewayz replaces 8+ business tools in one platform

CRM · Invoicing · HR · Projects · Booking · eCommerce · POS · Analytics. Free forever plan available.

Start Free →

Be nàkpɔ dzidzedze ŋutɔŋutɔ la, dze egɔme to tokenizing wò dɔwɔwɔ ƒe kɔpi vevitɔ — tsɔ kpe ɖe node_modules, vendor directories, build artifacts, kple generated files ŋu. Egbegbe tokenizers akpa gãtɔ (abe OpenAI ƒe tiktoken alo Anthropic ƒe token xlẽxlẽ mɔnu siwo wota ene) ateŋu awɔ dɔ tso agbalẽdzraɖoƒe aɖe ŋu le sɛkɛnd ʋɛ aɖewo me. Tsɔ emetsonua sɔ kple nɔnɔme siwo ƒo xlãe ƒe fesre siwo le kpɔɖeŋu siwo wò ƒuƒoƒoa zãna ŋutɔŋutɔ. Ne wò dɔwɔɖoɖo vevitɔ ƒe kɔpi sɔ ɖe nya siwo ƒo xlãe ƒe fesre ɖeka si me teƒe le na nyabiasewo kple mɔfiamewo la, ke èle nɔnɔme nyui aɖe me. Ne ewu fesrea nu 2-5x la, ahiã be woawɔ strategic chunking. Le 10x megbe la, àdi be yeade ga xɔtuɖaŋu ƒe tɔtrɔwo alo RAG (retrieval-augmented generation) pɔmpi tɔxɛwo me be AI dɔwɔnuwo nawɔ dɔ nyuie.

Na ƒuƒoƒo siwo le xɔ tum le mɔ̃wo abe Mewayz ene, afisi modular architecture ma dzimaɖitsitsiwo ɖe module vovovowo me xoxo — CRM, invoicing, HR, analytics, kple bubu siwo wu 200 — dzidzenu sia va doa dzidzɔ na ame ŋutɔ. Module ɖesiaɖe wɔa dɔ abe nusi le eɖokui si si ƒe ŋgɔdonyawo me kɔ ene, si le dzɔdzɔme nu wɔa anyigbatata ɖe akpa siwo sɔ na nya siwo ƒo xlãe-fesre nu. Enye xɔtuɖaŋu ƒe nɔnɔme ƒomevi si xea fe ɖe amegbetɔ ƒe beléle na ame kple AI gɔmesese siaa ta.

Nusi Developer Habɔbɔa Le Nya ʋlim Le Nyateƒee

Hacker News ƒe numedzodzro si ƒo xlã context-window badges la do masɔmasɔ dodzidzɔname geɖewo ɖe go le developer community me. Gbãtɔ nye xexemenunya: ɖe wòle be míawɔ kɔpi na AI zazãa? Dzadzɛnyenyelawo ʋlia nya be ele be woaŋlɔ kɔda na amegbetɔwo gbã, eye ele be AI dɔwɔnuwo natrɔ ɖe nɔnɔmeawo ŋu. Nuwɔna ŋuti nunyalawo tsi tre ɖe eŋu be ne xɔtuɖaŋu ƒe tiatia bɔbɔe aɖe na wò ƒuƒoƒoa wɔa dɔ nyuie wu 30% kple AI dɔwɔnuwo le ga home zero me na amegbetɔ ƒe nuxexlẽ la, gbegbe be yemawɔe o la, enye susuŋutinunya wu mɔ̃ɖaŋudɔwɔwɔ.

Nyaʋiʋli evelia ku ɖe nenye be nya siwo ƒo xlãe ƒe fesre sɔ nye metrik si li ke gɔ̃ hã si dze be woakplɔe ɖo. Nusiwo ƒo xlãe ƒe fesrewo dzi ɖe edzi ŋutɔ — tso 4K dzesiwo le GPT-3.5 ƒe gɔmedzedze va ɖo miliɔn ɖeka kple edzivɔ le Gemini 1.5 Pro me. Ne fesrewo yi edzi le kekem ɖe enu la, egbe tɔ "mesɔ o" va zua etsɔ tɔ "si sɔ bɔbɔe." Gake mɔ̃ɖaŋudɔwɔla bibiwo ɖee fia be ne wozã nya siwo ƒo xlãe ƒe fesre gãwo gɔ̃ hã la, kpɔɖeŋua ƒe dɔwɔwɔ dzi ɖena kpɔtɔna ne nya siwo ƒo xlãe ƒe didime. Kpɔɖeŋu si wɔa dɔ tso 50K tokens of focused, relevant code ŋu awɔ dɔ wu model ma ke si wɔa dɔ 500K tokens of a sprawling monorepo, ne wo ame evea siaa "sɔ" le mɔ̃ɖaŋununya nu gɔ̃ hã. Nya siwo ƒo xlãe ƒe nyonyome le vevie abe alesi agbɔsɔsɔme le vevie ene.

Dzeɖoɖo etɔ̃lia si wɔa dɔ wu la ƒoa xlã dɔwɔnuwo. Dɔwɔlawo di be woawɔ IDE ƒe ƒoƒo ɖekae siwo nya nu tso nya siwo ƒo xlãe ŋu siwo akpɔ faɛl siwo woatsɔ ade eme ne wole kɔpi ɖom ɖe AI la le wo ɖokui si. Wodi be nudzraɖoƒe ƒe ɖoƒe ƒe nunya si ase module ƒe liƒowo gɔme asi ƒe ɖoɖowɔwɔ manɔmee. Dɔ geɖe siwo le ʋuʋu ɖi le kuxi sia tututu gbɔ kpɔm fifia, wole nusi sɔ kple "nyataƒoƒo ƒe nuƒoƒoƒula" siwo ƒoa faɛl ƒe hatsotso nyuitɔ nu ƒu na dɔ ɖesiaɖe si AI kpena ɖe eŋu la tum.

Esia Trɔtrɔ Zu Hoʋiʋli ƒe Viɖe

Na asitsalawo — menye developer teams ɖeɖeko o — context window fit le downstream gɔmesese siwo dze be woase egɔme. Dɔwɔƒe siwo ɖoa kɔmpiutadziɖoɖowo ɖe amewo kabakaba, eye vodada ʋɛ aɖewo koe nɔa wo me, eye woƒe asi bɔbɔ wu la ɖua dzi le woƒe asiwo me. AI-kpekpeɖeŋunana ŋgɔyiyi nye ŋusẽdzidzenu vavã, gake ne woɖo codebase si le ete la be wòawɔ eŋudɔ ko hafi. Habɔbɔ siwo dea ga AI-xɔlɔ̃wɔwɔ ƒe codebases me egbea le viɖe siwo le dzidzim ɖe edzi siwo akeke ɖe enu le ɣeyiɣi aɖe megbe la tum.

Gɔmeɖose sia keke ta yi ŋgɔ wu kɔmpiuta dɔwɔɖoɖo dzadzɛwo. Asitsaha siwo le dɔ wɔm le mɔ̃wo abe Mewayz ene, si ƒoa CRM, fexexe, fetu, HR, ʋuwo dzikpɔkpɔ, kple numekuku nu ƒu ɖe modular ɖoɖo ɖeka me la kpɔa viɖe tso xexemenunya sia ke me le dɔwɔwɔ ƒe ɖoɖo nu. Ne wò asitsanyatakakawo le modules siwo woɖo nyuie, siwo do ƒome kple wo nɔewo me tsɔ wu be woakaka ɖe SaaS dɔwɔnu 15 siwo metso kadodo me o la, AI ate ŋu abu tame le wò dɔwɔwɔ bliboa ŋu — ade dzesi nɔnɔme siwo le nudzadzra, kpekpeɖeŋunana, kple ganyawo me siwo womate ŋu akpɔ le siled systems me o. Gɔmeɖose ma ke si na be codebase nye AI-xɔlɔ̃wɔwɔ na asitsatsa AI-xɔlɔ̃: wɔwɔme si me kɔ, liƒo dzadzɛ, kple nya siwo ƒo xlãe si me kɔ.

Nu si woate ŋu atsɔ ayi na mɔ̃ɖaŋudɔwɔlawo ƒe ŋgɔnɔlawo la le tẽ. Dze wò nya siwo ƒo xlãe ƒe fesre dzidzedze gɔme egbea — le vome gɔ̃ hã. Tsɔe kpe ɖe wò mɔ̃ɖaŋudɔwɔwɔ ƒe lãmesẽnyatakakawo ŋu kpe ɖe xɔtuɣiwo kple dodokpɔ ƒe nutsyɔtsyɔ ŋu. Zãe abe nu ɖeka si nàtsɔ ade eme ene (menye nu ɖeka kolia o) ne èle xɔtuɖaŋu ŋuti nyametsotsowo wɔm. Eye nàde dzesii be codebases siwo aɖe vi geɖe tso AI ƒe ŋgɔyiyidɔwɔnuwo ƒe dzidzime si gbɔna me lae nye esiwo wole ɖoɖo wɔm ɖe eŋu be gɔmesese nanɔ eŋu fifia.

Akɔtagbalẽvia Nye Dzeɖoɖo Gɔmedzedze, Menye Teƒe si Woayi o

README ƒe akɔtagbalẽvi si fia "87% context fit — Claude 200K" nye nu sue aɖe. Exɔa sɛkɛnd hafi wòwɔa eye wòxɔa fli ɖeka le wò dɔa ƒe nuŋlɔɖiwo me. Gake nusi wòtsi tre ɖi na — ɖokuitsɔtsɔna si woɖo koŋ, si woate ŋu adzidze ɖe codebase gɔmesese ŋu — fia nane si ŋu gɔmesese le le mɔ̃ɖaŋudɔwɔlawo ƒe ƒuƒoƒo aɖe ƒe nu vevitɔwo ŋu. Egblɔ be: míebua alesi míaƒe kɔdasia ase egɔmee ŋu, menye tso dɔwɔla si kplɔe ɖo gbɔ ko o, ke boŋ AI ɖoɖo siwo va le ŋgɔyiyidɔ ɖesiaɖe ƒe akpa aɖe geɖe wu.

Nu si do tso nɔnɔme sia me si xɔ asi wu la menye akɔtagbalẽvia ŋutɔ o. Enye dzeɖoɖo siwo wòhena vɛ le xɔtuɖaŋuwo me toto, duƒuƒu ƒe ɖoɖowɔwɔ, kple mɔ̃ɖaŋununya ƒe fenyinyi ŋuti numedzodzrowo me. Ne "context window fit" va zu wò mɔ̃ɖaŋunyagbɔgblɔ ƒe akpa aɖe la, èdzea nyametsotso siwo dzɔna be woawɔ ɖeka kple nusianu si míenya tso kɔmpiuta dɔwɔɖoɖo nyuiwo ƒe wɔwɔme ŋu ƒe bla nanewoe nye sia la wɔwɔ gɔme: module suewo, siwo ŋu woƒe susu le kple ŋgɔdonya siwo me kɔ kple kadodo suetɔ kekeake. Menye AI ƒe tɔtrɔ kpata lae to gɔmeɖose siawo vɛ o. Gake ele susu yeye si woate ŋu adzidze la nam ƒuƒoƒowo be woadze wo yome mlɔeba.

Nyabiase Siwo Wobiana Enuenu

Nukae nye context window fit eye nukatae wòle vevie?

Context window fit dzidzea wò codebase ƒe alafa memamã si LLM ateŋu awɔ dɔ le le nyabiase ɖeka me. Alafa memamã si lolo wu fia be AI dɔwɔnuwo ate ŋu ase wò dɔa gɔme geɖe wu zi ɖeka, si ana nàkpɔ kɔpi ƒe aɖaŋuɖoɖo nyuitɔwo, agbugbɔ awɔ nusiwo sɔ pɛpɛpɛ wu, eye susumenuwo mebɔ o. Esi AI-kpeɖeŋutɔ ƒe ŋgɔyiyi va zu dzidzenu la, metrik sia kpɔa ŋusẽ ɖe alesi wò ƒuƒoƒoa ateŋu awɔ dɔ nyuie kple dɔwɔnuwo abe Copilot, Cursor, kple Claude.

Aleke mawɔ akpɔ nye codebase ƒe context window ƒe sɔsɔ?

Àteŋu azã akɔtagbalẽvi ƒe dɔwɔnu si woateŋu azã le ʋuʋu ɖi si woama le Hacker News dzi atsɔ awɔ nukpɔkpɔ ƒe dzesi na wò nudzraɖoƒe. Ebua wò codebase token xexlẽme bliboa eye wòtsɔnɛ sɔna kple LLM context windows xɔŋkɔwo. Akɔtagbalẽvi la ɖea alafa memamã ƒe dzesi si nàte ŋu ade wò README me la fiana, si naa nudzɔlawo kple amesiwo kpɔ gome le eme la kpɔa alesi wò dɔa le klalo na AI enumake.

Aɖaŋu kawoe naa codebase ƒe context window fit score nyona ɖe edzi?

Lé fɔ ɖe modular architecture ŋu, nusiwo ŋu nètsi dzi ɖo ƒe mama nyuie, kple dead code ɖeɖeɖa. Monorepo siwo woɖo nyuie kple liƒo siwo me susu le na LLMwo wɔa dɔ tso modules siwo sɔ ŋu le wo ɖokui si. Kɔdawo ƒe wɔwɔfia dzi ɖeɖe kpɔtɔ, faɛlwo ƒe kpuie, kple ati siwo dzi woanɔ te ɖo dzadzɛ dzi kpɔkpɔ katã kpena ɖe eŋu. Mɔ̃wo abe Mewayz ɖea gɔmeɖose sia fiana — woblaa modules 207 ɖe asitsatsa ƒe OS si wowɔ ɖe ɖoɖo nu si wowɔ na beléle na wo kple dɔwɔwɔ nyuie me.

Ðe codebase suetɔ fia ɣesiaɣi be AI ƒe sɔsɔ nyuie wua?

Menye kokoko o. Codebase sue si me tangled dependencies kple nuŋlɔɖi madeamedzi le ate ŋu asesẽ na LLMwo be woade ŋugble le eŋu wu esi lolo wu, si ŋu wowɔ ɖoɖo ɖo nyuie. Nusi le vevie enye alesi gbegbe nya siwo ƒo xlãe si sɔ la sɔ ɖe fesrea nu. Abstraction dzadzɛwo, ŋkɔyɔyɔ ƒe ɖoɖo siwo mewɔa tɔtrɔ o, kple modular design na AI dɔwɔnuwo wɔa dɔ nyuie ne womateŋu aɖu code ƒe fli ɖesiaɖe zi ɖeka o gɔ̃ hã.