Esedlule bekulokhu kungumuntu ophelele (noma i-joyride) yokwenziwa ngokushesha kobuchwepheshe be-AI. Eminyakeni engamashumi amabili nanhlanu ngibale ukuthi ngibe ngunjiniyela wesoftware, angikwazi ukukhumbula ukuguquka kwe-tectonic kobukhulu obufanayo, okuvele kuchazwe ngendlela yesoftware.
Kungaba ukufushane ukukholelwa ukuthi le Revolution iyeka ukumane ukhiqize ikhodi, noma kunjalo. Nginama-agents ai okuvulekile kanye ne-ecosystem evulela ukuhlanganiswa okusha, izisekelo zendlela esiqapha ngayo, siqonde kanjani, futhi sazise kahle isoftware. Amathuluzi asidla kahle ezweni le-Human-Centric, akha imiqondo enjengezexwayiso zezandla, ama-datagrids, namadeshibhodi, ayatholakala angafanele futhi angasebenzi. Ukuhlola ukusebenza kwesicelo (APM) Platforms futhi, ikakhulukazi, indlela abeka ngayo izingodo, ama-metric, kanye nokulandela umkhondo, kuzodingeka ukuthi avume ukuthi imibuthalo yomuntu etholakele, iqembu selivele lidluliselwe kulowo msebenzi e-AI.
Ama-ejenti ahlakaniphile aqala ukuhlanganiswa kwi-SDLC (Isoftware Development Lifeccycle), ukuhlaziya ngokuzimele, ukuxilonga, kanye nokwenza ngcono izinhlelo ngesikhathi sangempela. Leli paradigm elisafufusa lidinga ukuthatha okusha ngenkinga yakudala. Ukuthola idatha ebonakalayo okufanele kufakwe ekwenzeni ama-ejenti namaqembu akhiqize kakhudlwana, kufanele ahlelwe ngemishini, hhayi kubantu. Ubuchwepheshe obunye bamuva owenza lokhu kungenzeka futhi okuye ngathola kahle i-buzz kamuva nje, I-Model Grotector Protocol (MCP).
Mcps ngamafuphi
Ekuqaleni kwethulwa ngu -ObubizinIsimo somongo wemodeli (i-MCP) simele iTier Communication Tier phakathi kwama-Agents ai nezinye izinhlelo, ezivumela ama-ejenti ukuthi afinyelele imithombo yedatha eyengeziwe futhi enze izenzo njengoba zibona kufanelekile. Okubaluleke kakhulu, ama-MCP avulekele ama-agent amasha we-ejenti akhethe ngobuhlakani ukwenza okungaphezu kwesilinganiso sawo ngokushesha futhi ngaleyo ndlela akhulise uhla lwamacala okusetshenziswa kungabhekana namacala.
Ubuchwepheshe abusha, kepha i-ecosystem iyi. Engqondweni yami, kufana nokuvela kokuthuthukiswa kohlelo lokusebenza lweselula ukuthi ube nesitolo sohlelo lokusebenza. Akulona ngengozi ukuthi njengamanje kuthola ukukhula kwezilinganiso zeCambria, njengokumane kokuba ne-ecosystem ecebile futhi esezingeni eliphakeme ivule imakethe ngamathuba amasha. Ukukhuluma kabanzi kakhulu, ama-MCPs amelela imodeli ye-agent-centric yokwakha imikhiqizo emisha engaguqula indlela uhlelo lokusebenza lakhiwa khona futhi indlela aletha ngayo inani lokugcina abasebenzisi.
Ukulinganiselwa kwemodeli ye-centirm yabantu
Izicelo eziningi zesoftware zakhiwa abantu njengabasebenzisi bawo abaphambili. Ngokuvamile, umthengisi unquma ukutshala imali ekwakheni izici ezithile zomkhiqizo, okukholelwa ukuthi kuzoba umdlalo omuhle nezidingo nezidingo zabasebenzisi bokugcina. Abasebenzisi bese bezama ukusebenzisa lelo elinikezwe izici zokuzama ukufeza izidingo zabo ezithile.
Kunemikhawulo emithathu esemqoka kule ndlela, eseba ngaphezulu kwesithiyo njengoba amaqembu amukela ama-Agents ai ukuhambisa izinqubo zawo:
- Isikhombimsebenzisi esilungisiwe– Abaphathi bomkhiqizo kufanele balindele futhi basebenzise icala lokusebenzisa ukudala indawo efanelekile kuhlelo lokusebenza. Isethi ye-UI noma i-API ihleliwe futhi ayikwazi ukuzivumelanisa ngokwayo nesidingo ngasinye esiyingqayizivele. Ngenxa yalokho, abasebenzisi bangathola ukuthi ezinye izici azisizi ngokuphelele kwizidingo ezithile. Kwezinye izikhathi, ngisho nenhlanganisela yezici, umsebenzisi akakwazi ukuthola konke abakudingayo.
- Umthwalo wokuqonda – Inqubo yokuxhumana nedatha yesicelo ukuthola imininingwane izidingo zomsebenzisi idinga umzamo wezandla, izinsizakusebenza, futhi kwesinye isikhathi izinsiza. Ukuthatha ama-APM njengesibonelo, ukuqonda imbangela yenkinga yokusebenza futhi ukulungisa kungathatha uphenyo, njengoba kungenzeka ukuthi umagazini ngamunye uhlukile. Ukuntuleka kwe-automation nokuthembela ezinqubweni zokuzithandela ngokuzithandela kuchaza ukuthi idatha ayisetshenziswa nhlobo.
- Isilinganiso Esinqunyelwe – Umkhiqizo ngamunye uvame ukubamba ingxenye yesithombe edingekayo ukuxazulula imfuneko ethile. Isibonelo, i-APM ingaba nedatha yokulandela, kepha akukho ukufinyelela kwiKhodi, umlando we-GitHub, amathrekhi weJira, idatha yengqalasizinda, noma amathikithi ekhasimende. Ishiywe umsebenzisi ukuze asebenze ngokusebenzisa imithombo eminingi ukuze afike empandeni yenkinga ngayinye.
I-Agent-Centric MCPS – uhlelo lokusebenza oluhlanekezelwe
Ngokuqala kwe-MCPS, abathuthukisi be-software manje banokukhethwa kokwamukela imodeli ehlukile yokuthola isoftware. Esikhundleni sokugxila ecaleni elithile lokusetshenziswa, uzama ukukhahlela izinzipho ezifanele ze-UI zamaphethini wokusebenzisa anezinhlamvu ezinzima, izinhlelo zokusebenza zingaguqula zibe yisisetshenziswa sezinqubo eziqhutshwa yi-AI. Lokhu kuchaza ukuguquka kusuka ekusekeleni okumbalwa kokuxhumana okuchazwe ngaphambilini ukusekela okuningi -phambililwayo Sebenzisa amacala. Esikhundleni sokutshala imali esicini esithile, uhlelo lokusebenza manje lungakhetha ukubolekisa ubuchwepheshe bawo besizinda ku-Agel e-AI umbhalo na- isenzo Lokho kungasetshenziswa ngokusobala noma nini lapho kufaneleka, noma ngabe ngokungaqondile.
Njengoba lesi sikali semodeli, umenzeli angahlanganisa umthungo idatha nezenzo ezivela kuzicelo ezahlukahlukene kanye nezizinda, njenge-github, Jira, amapulatifomu wokubonwa, amathuluzi wekhodi uqobo. Umenzeli angabe eseshintsha inqubo yokuhlaziya uqobo njengengxenye ye ukuhlanganisa Imininingwane, isuse izinyathelo ezinancwajana kanye nesidingo sobuchwepheshe obukhethekile.
Ukuqashelwa akuyona uhlelo lokusebenza lwewebhu; Ubuchwepheshe bedatha
Ake sibheke isibonelo esisebenzayo esingabonisa ukuthi imodeli ye-Agent-Centric evula ngayo izindlela ezintsha ze-neural kwinqubo yobunjiniyela.
Wonke unjiniyela uyazi Ukubuyekezwa Kwekhodi dinga umzamo omningi; Ukwenza izinto zibe zimbi kakhulu, isibuyekezo kwesinye isikhathi sicishiwe umongo sangeminye imisebenzi yazo, siphinde sikhiphe umkhiqizo weqembu. Ebusweni, lokhu kungabonakala sengathi ithuba lokubukwa kwezicelo zikhanye. Ngemuva kwakho konke, ikhodi esetshenziswa ngaphansi kokubuyekezwa isivele iqoqe imininingwane enengqondo esebenza ekuhlolweni nasekukhiqizweni kwangaphambi kokukhiqiza. Ngokweqiniso, lolu lwazi lungasiza ekuqondiseni okwengeziwe ngezinguquko, ukuthi zinomthelela yini, nokuthi ngabe zishintshile kanjani indlela yokuziphatha kohlelo. Ngokuyalela, izindleko eziphakeme zokwenza umqondo wayo yonke leyo datha kuzo zonke izinhlelo zokusebenza eziningi kanye nokusakazwa kwedatha, kwenza eduze kwe-Isher.
Ekugelezeni kwe-ejenti-ke, noma kunjalo, lapho unjiniyela ebuza i-A ejenti ye-AI ukusiza ukubukeza ikhodi entsha, leyo nqubo iba nokuzimela ngokuphelele. Ngemuva, i-ejenti izofaka ama-orchestte izinyathelo zokuphenya kuzo zonke izinhlelo zokusebenza eziningi kanye nama-MCPs, kufaka phakathi amathuluzi wokuphambuka, ukubuyisa ukuqonda okusebenzayo mayelana nezinguquko zekhodi. Umenzeli angafinyelela idatha efanelekile yesikhathi sokusebenza (isib. Iphinde ixhumanisa i-diff nge-runtime efanelekile yesikhathi, i-Flags regnction regream noma ukusebenzisana okuhlulekile, futhi ikhomba izehlakalo zakamuva ezingahle zilandele ikhodi elungisiwe.
Kulesi simo, unjiniyela akadingi ukugcwala ngamathuluzi ahlukene noma amathebhu noma achithe isikhathi azama ukuxhuma amachashazi- omenzeli akulethela konke ngemuva kwezigcawu, ukukhomba kanye nokulungiswa okungenzeka. Njengoba impendulo ngokwayo ikhiqizwa ngamandla: kungaqala ngesifinyezo esifushane sombhalo, sandise etafuleni elibonisa amamethrikhi ngokuhamba kwesikhathi, kufaka phakathi isixhumanisi esiya ku-file elithintekile, futhi sishumeke ishadi elibona umugqa wesikhathi ngaphambi nangemva kokukhululwa.
Ngenkathi ukuchitheka komsebenzi okungenhla kukhiqizwe nge-ejenti, amanye amaklayenti we-AI azovumela umsebenzisi ukuthi athumele ukugeleza komsebenzi okufunwa ngokungeza imithetho yememori ye-ejenti. Isibonelo, leli yifayela lememori engile nto esebenzisa isikhombisi ukuze ngiqinisekise ukuthi konke ukubuyekeza ikhodi kuvuselelwa kuvuselelwa kuhlola imvelo yokuhlola futhi uhlole ukusetshenziswa okusekelwe ekukhiqizeni.
Ukufa ngamacala okusebenzisa inkulungwane
Isimo sokubukeza ikhodi singenye nje -ngakanani -phambililwayo Sebenzisa amacala akhombisa ukuthi i-AI angayisebenzisa kanjani ngokuthula idatha efanele ye-MCP ukusiza umsebenzisi afeze izinhloso zawo. Okubaluleke kakhulu, umsebenzisi akadingi ukuqaphela izinhlelo zokusebenza ezazisetshenziswa ngokuzimela yi-ejenti. Ngokombono womsebenzisi, badinga nje ukuchaza isidingo sabo.
Amacala okusetshenziswa aqhaphesayo angathuthukisa umkhiqizo womsebenzisi ngaphesheya kwebhodi ngedatha engeke yenziwe ifinyeleleke ngenye indlela. Nazi ezinye izibonelo ezimbalwa lapho idatha yokuqapha ingenza umehluko omkhulu, ngaphandle kwanoma ngubani okufanele avakashele ikhasi lewebhu elilodwa le-APM:
- Isizukulwane Sokuhlola Kususelwa ekusetshenzisweni kwangempela
- Ukukhetha izindawo ezifanele ukuze ukunqatshelwa kususelwa kuzinkinga zekhodi ezithinta ukusebenza kakhulu
- Ukuvimbela ukuphula izinguquko Lapho ikhodi isahlolwa
- Ukuthola Ikhodi engasetshenziswanga
Imikhiqizo idinga ukuguquka
Ukwenza Ukuphawula Kuwusizo ku-ejenti, noma kunjalo, kuhileleke kancane kunokushaya nge-adaptha ye-MCP ku-APM. Ngempela, amathuluzi amaningi esizukulwane, ekuphuthumeni ukusekela ubuchwepheshe obusha bathathe leyo ndlela, bangabheki ukuthi abasebenzisi be-AI nabo banokulinganiselwa kwabo.
Ngenkathi ihlakaniphile futhi inamandla, ama-ejenti awakwazi ukufaka esikhundleni noma yiluphi uhlelo lokusebenza olusebenzisana nanoma iyiphi idatha, ngesidingo. Ekuwohlokeni kwabo kwamanje, okungenani, baboshelwe ngosayizi wedathasethi bese bema isikhashana ukusebenzisa ama-algorithms ayinkimbinkimbi noma izibalo eziphakeme kakhulu. Uma ithuluzi lokuphawula lingumhlinzeki wedatha osebenzayo ku-ejenti, kufanele lilungiselele idatha kusengaphambili esikhundleni salokhu kulinganiselwa. Ukukhuluma kabanzi, lokhu kuchaza indima yemikhiqizo eneminyaka yobudala ye-AI – Ukuhlinzeka iziqhingi ze-Nontrivial Domain Ubuchwepheshe okufanele busetshenziswe kwinqubo eqhutshwa yi-AI.
Kukhona okuthunyelwe okuningi ngesihloko ngendlela engcono kakhulu yokulungiselela idatha ukuze isetshenziswe yi-generative AI Agents, futhi ngifake ezinye izixhumanisi ekugcineni kwalokhu okuthunyelwe. Kodwa-ke, singachaza ezinye zezidingo zokukhishwa okuhle kwe-MCP emithanjeni ebanzi:
- -Bukekayo (amabhizinisi aguqukayo aguqukayo, athayishiwe)
- Kufakwe phambili (Kuhlanganiswe, kunikezelwe, umaki)
- Okuqukethwe (ehlukaniswe yiseshini, impilo, noma inhloso)
- -Xhuma (izinkomba ngawo wonke amakhodi wekhodi, izingodo, ukwenza namathikithi)
Esikhundleni sokusebenzisa i-surfacing raw telemetry an I-MCP kufanele yondle ukulandisa kwedatha okuhambisanayo ku-ejenti, ukuhlaziya ngemuva. I-ejenti akuyona nje umbono wedeshibhodi ukuze anikezwe. Ngasikhathi sinye, kufanele futhi senze -qondeneIdatha eluhlaza etholakala ngokufunwa ukuvumela ukuthi kuqhubeke uphenyo, ukusekela izenzo zokucabanga ezizimele ze-Agent.
Ukunikezwa okulula kwedatha eluhlaza kungaba nengcaphun nge-ejenti ukukhomba inkinga ekhombisa ukulandelwa kwezigidi ezingama-5% kuphela, noma zenze isinqumo sokuthi lelo phethini liyathandeka.
Ukuvala igebe, imikhiqizo eminingi cishe izovela yaba 'presessors' ye-AI, ilethe izinqubo ze-ML ezinikezele kanye nokuhlaziywa kwezibalo okuphezulu kanye nobuchwepheshe besizinda.
Farewell to apms
Ekugcineni, ama-apms awawona amathuluzi wefa – amele a I-Legacy Mindsetlokho kancane kodwa nakanjani kuthathelwa indawo. Kungathatha isikhathi esithe xaxa somkhakha ukuba afinyelele, kepha ekugcineni kuzothinta imikhiqizo eminingi esiyisebenzisa njengamanje, ikakhulukazi embonini yesoftware, okuwumjaho ukuze wamukele i-AI ekhiqizayo.
Njengoba i-Ai iba namandla ngokwengeziwe kwisoftware ethuthukayo, ngeke isanqunyelwe ekusebenzisaneni kwabantu. Ukubonisana kwe-AI okujwayelekile kuzosetshenziswa njengengxenye yenqubo ye-CI, futhi kwezinye izimo, noma zigijime unomphela njengezinqubo zangemuva ngokuqhubekayo zibheka idatha nokwenza izinto. Ngalokho engqondweni, amathuluzi amaningi azoqhamuka nemiphumela yawo ye-Agen-Centric ehlelekile futhi kwesinye isikhathi athathe indawo yawo eqondile – noma engcupheni yokushiywa ngaphandle kwamakhasimende abo i-AI SLDC Stack entsha.
Izixhumanisi nezinsizakusebenza
- Ophuzi: Ukwenza okujwayelekile kubalulekile – ukuvumelana kwe-schema nokuxhumanisa okuhlobene nokuthuthukisa ukucabanga komthombo.
- UHarrison Clarke : Ukuhlupheka kwangaphambili kufanele kushaye indawo emnandi – ecebile ngokwanele ukuze uthole ukuthobeka, okuhlelekile ngokunemba.
- DigiLitocean : Ukuhlanganiswa ngemingcele ye-semantic (amaseshini omsebenzisi, ukugeleza) kuvula ukukhishwa okungcono nokucabanga okusekelwe izindaba.
Ufuna ukuxhuma?Ungangifinyelela ku-Twitter at @Doppleware noma nge-LinkedIn.
Landela okwami I-MCPUkuhlaziywa kwekhodi enamandla usebenzisa ukubonwa ku-
