Namuhla i-Google ikhiphe i-ejenti yayo yomthombo ovulekile kuphrothokholi yomenzeli, eqanjwe ngomcabango A2A noma i-ejenti ku-ejenti. Njengoba sesivele sibona umfutho omningi ngemuva kwe-Anthropic I-MCP (I-Model Context Protocol), i-Google ikufunile lokho I-A2A ihambisana ne-MCP . Baze basebenzise i-emoji yenhliziyo ukuze baqonde iphuzu.

Angiqiniseki kangako, ngakho-ke nginqume ukubheka ngokujulile futhi ngihlole ukuthi kuzoba yini indawo ye-A2A endaweni yonke ye-agency. Ngakho-ke sizofaka ukuthi i-A2A isebenza kanjani empilweni yangempela, kanye nokuqhathanisa ne-MCP.

Idrayivu yokuhlola i-A2A

Ukusebenzisa i-A2A kufana ngokumangazayo ne-MCP. Ungasebenzisa ama-ejenti/amaseva e-A2A ambalwa, bese iklayenti le-A2A lingakwazi ukuxhuma kuwo wonke. Izindaba ezinhle ukuthi ngokuvamile awudingi ukusebenzisa ama-ejenti e-A2A kanye neklayenti le-A2A.

Abenzeli be-A2A abasebenzayo (amaseva)

Ngihlanganise wonke ama-ejenti ayisibonelo endaweni yangakini

1. Umenzeli we-Google ADK engakulethela imibiko yezindleko
2. Umenzeli we-CrewAI lokho kungathola ukukhiqiza isithombe
3. I-ejenti ye-LangGraph engathola izinga lokushintshisana lakwamanye amazwe lakamuva

Indlela iseva ye-A2A eyenza ngayo umhlaba wazi amandla ayo ingokuthi “Ikhadi Lomenzeli” ngefomethi ye-JSON. Njengesibonelo ikhadi le-ejenti le-google ADK libukeka kanje:

 { "name": "Reimbursement Agent", "description": "This agent handles the reimbursement process for the employees given the amount and purpose of the reimbursement.", "url": "http://localhost:10002/", "version": "1.0.0", "capabilities": { "streaming": true, "pushNotifications": false, "stateTransitionHistory": false }, "defaultInputModes": [ "text", "text/plain" ], "defaultOutputModes": [ "text", "text/plain" ], "skills": [ { "id": "process_reimbursement", "name": "Process Reimbursement Tool", "description": "Helps with the reimbursement process for users given the amount and purpose of the reimbursement.", "tags": [ "reimbursement" ], "examples": [ "Can you reimburse me $20 for my lunch with the clients?" ] } ] }

Yethula uhlelo lokusebenza lwedemo yeKlayenti ye-A2A

Asiqhubeke neklayenti. Imiyalo yokuthola uhlelo lokusebenza lwewebhu yedemo ilapha. https://github.com/google/A2A/tree/main/demo

Uma uhlelo lokusebenza lwewebhu selusebenza, ungalufinyelela esipheqululini sakho. Iklayenti libukeka lifana ne-Gemini AI Studio ene-signature ye-Google Material design.

URL: localhost:12000

Okokuqala, sidinga ukungeza wonke ama-ejenti kuklayenti ngokucacisa i-URL yawo eyisisekelo. Njengoba kimina ngasebenzisa bonke abasebenzeli endaweni, i-URL yabo eyisisekelo bekuyi:

• I-Google ADK
  • indawo yendawo: 10002
• crewI
  • Indawo yokuhlala: 10001
• I-LangGraph
  • Indawo: 10000

Inothi eseceleni: ngaphakathi kwephrothokholi, i-URL yokugcina ibukeka kanjena:

https://localhost:10002/.well-known/agent.json

Manje ungabona wonke ama-ejenti amathathu axhunyiwe:

Ungabona umlando wengxoxo lapha

Lonke uhlu lomcimbi

Futhi lonke uhlu lwemisebenzi

Izilungiselelo ziyisisekelo

Hlola i-ejenti ye-Google ADK ukuze uthole isimangalo sezindleko

Test LangGraph for forex rate

Hlola i-ejenti ye-CrewAI yokwenza isithombe

Ukuhlolwa kwe-combo kwama-ejenti amaningi

Ngifuna ukubona ukuthi iklayenti le-A2A lingakwazi yini ukusebenzisa ama-ejenti amaningi ukuze lifinyelele umgomo owodwa . Ngakho-ke ngivivinye ukuthi ingakwazi yini ukuhlanganisa i-ejenti yesimangalo sezindleko ne-ejenti yesilinganiso se-forex. Futhi kwasebenza.

Umsebenzi wami “ kwakuwukufuna izindleko zikabhiya eJalimane ngenkathi ngisohambweni lwebhizinisi, ama-euro angu-5, ngo-April 4 2025 ”. Ingxoxo yadlula imizuliswano embalwa yokubuyela emuva naphambili, futhi ekugcineni yathola inani elifanele lamadola ase-US efomini lesicelo sezindleko.

Ukubhekwa Kokuqala kwe-A2A

Ngithanda ukuthi i-A2A iyimodeli ehlanzekile ye-Client-Server engasetshenziswa yomibili futhi isingathwe ukude. Iklayenti alithwele kanzima ngokucacisa kanye nokwethula ama-ejenti/amaseva.

Ukucushwa komenzeli kulula nje ngokucacisa i-URL eyisisekelo, futhi “Ikhadi Lomenzeli” linakekela ukushintshana komongo. Futhi ungangeza futhi ususe ama-ejenti ngemuva kokuthi iklayenti seliqalisiwe.

Kufomethi yedemo yamanje, kunzima ukuqonda ukuthi ama-ejenti axhumana kanjani kanye nokufeza imisebenzi eyinkimbinkimbi. Iklayenti libiza i-ejenti ngayinye ngokuhlukene ngemisebenzi eyahlukene, ngaleyo ndlela kufana nokushaya amathuluzi amaningi.

Qhathanisa i-A2A ne-MCP

Manje sengizamile i-A2A, sekuyisikhathi sokuyiqhathanisa ne-MCP engibhale ngayo ekuqaleni Lesi sihloko .

Nakuba kokubili i-A2A ne-MCP zihlose ukuthuthukisa ukuthuthukiswa kwesistimu ye-ejenti ye-AI, ngokombono zibhekana nezidingo ezihlukile. I-A2A isebenza ezingeni le-ejenti ne-ejenti, igxile ekusebenzelaneni phakathi kwezinhlangano ezizimele, kuyilapho i-MCP isebenza ezingeni le-LLM, igxile ekunothiseni umongo namandla emodeli yolimi ngalunye.

Futhi ukunikeza amazwibela okufana kwawo okuyinhloko nomehluko ngokuya ngemibhalo yawo yephrothokholi:

Iziphetho

Noma i-Google ikwenze kuzwakale sengathi i-A2A iyiphrothokholi ehambisana ne-MCP, ukuhlolwa kwami kokuqala kubonisa ukuthi zigqigqana ngokwedlulele ngenjongo nezici . Bobabili babhekana nezidingo zabathuthukisi bohlelo lokusebenza lwe-AI ukusebenzisa ama-ejenti amaningi namathuluzi ukufeza izinhloso eziyinkimbinkimbi. Njengamanje, bobabili abanayo indlela enhle yokubhalisa nokuthola amanye ama-ejenti namathuluzi ngaphandle kokucushwa ngesandla.

I-MCP iqale kusenesikhathi futhi isivele ithole ukwesekwa okukhulu okuvela emphakathini onjiniyela kanye namabhizinisi amakhulu. I-A2A isencane kakhulu, kodwa isivele inosekelo oluqinile lokuqala oluvela kumakhasimende amaningi ebhizinisi le-Google Cloud.

Ngikholwa ukuthi lezi yizindaba ezinhle konjiniyela, njengoba bazoba nokukhetha okuningi kumaphrothokholi avulekile najwayelekile womenzeli . Isikhathi kuphela esingasho ukuthi yikuphi okuzobusa, noma zingase zihlangane zibe izinga elilodwa.