I-ACP: Iphrothokholi ye-Intanethi yama-AI AD

I-Protocol Yezokuxhumana), ama-Agents AI angabambisana amaqembu ngokukhululeka amaqembu, izinhlaka, ubuchwepheshe kanye nezinhlangano. Iphrothokholi yendawo yonke eguqula isimo esihlukanisiwe sanamuhla AI Agents kubalingani beqembu abaxhume phakathi. Lokhu kuvula amazinga amasha okusebenzisana, ukusebenzisa kabusha, kanye nesilinganiso.

Njengesilinganiso esivulekile ngokubusa okuvulekile, i-ACP isanda kukhipha inguqulo yayo yakamuva, ivumela ama-Agents e-AI ukuthi axhumane kuwo wonke ama-Frameworks ahlukene nezinhlaka zobuchwepheshe. Kuyingxenye yemvelo ekhulayo, kufaka phakathi iBeaeai (lapho ngiyingxenye yeqembu), olunikelwe kwi-Linux Foundation. Ngezansi kwezinye izici ezibalulekile; Ungafunda kabanzi ngemiqondo eyisisekelo nemininingwane kumadokhumenti.

Izici ezibalulekile ze-ACP:
Ukuxhumana okusekelwe ekuphumeni: I-ACP isebenzisa amaphethini ajwayelekile we-HTTP okuxhumana, okwenza kube lula ukuhlanganisa ukukhiqizwa. Ngenkathi uJson-RPC ethembele ezindleleni eziyinkimbinkimbi.
Ayikho i-SDK edingekayo (Kepha kukhona okukodwa uma ufuna): I-ACP ayidingi imitapo yolwazi ekhethekile. Ungaxhumana nama-ejenti usebenzisa amathuluzi anjenge-curl, postman, noma isiphequluli sakho. Ngokusebenziseka kalula, kukhona i-SDK etholakalayo.
Ukutholwa okungaxhunyiwe ku-inthanethi: Ama-ACP Agents angashumeka i-metadata ngqo emaphaketheni abo okusabalalisa, anika amandla ukutholwa noma ngabe ayisebenzi. Lokhu kusekela okuvikelekile, okugxiliwe, noma ukukala-kuya ku-zero ezindaweni lapho ukutholwa kwensiza yendabuko kungenzeki khona.
I-Async-Okokuqala, ukuvumelanisa kusekelwa: I-ACP yakhelwe ngokuxhumana kwe-asynchronous njengokuzenzakalelayo. Lokhu kulungele imisebenzi esebenza isikhathi eside noma eyinkimbinkimbi. Izicelo zokuvumelanisa nazo ziyasekelwa.

Qaphela: I-ACP inika amandla ama-orchestando kunoma yiliphi iphethini yezakhiwo ze-Agent zokwakha, kepha ayilawulwa ukuhamba komsebenzi, ukuthunyelwa, noma ukuxhumanisa phakathi kwama-ejenti. Esikhundleni salokho, inika amandla ama-orchestando asebenze abahlukahlukene ngokwejwayelekile ukuthi baxhumana kanjani. Ucwaningo lwe-IBM lwakhiwe beaai, uhlelo lomthombo ovulekile olwenzelwe ukuphatha i-ejenti inchent, ukuthunyelwa, nokwabelana (kusetshenziswa i-ACP njengesendlalelo sokuxhumana).

Kungani sidinga i-ACP?

Njengoba inani le-AI ADETS “endle” eLundle “, kanjalo nenani lobulumbi ekwazini ukuthola umphumela omuhle kakhulu kubuchwepheshe ngakunye okuzimele ngecala lakho lokusebenzisa (ngaphandle kokuthi ucindezelwe umthengisi othile). Uhlaka ngalunye, ipulatifomu, kanye nethuluzi lokuphuma lapho linikeza izinzuzo ezihlukile, kepha zibahlanganisa bonke ndawonye ohlelweni olulodwa lwe-agent kuyinselele.

Namuhla, iningi lezinhlelo ze-ejenti zisebenza kuma-silos. Bakhelwe kuzinhlaka ezingahambelani, badalula ama-API ngokwezifiso, futhi anakho umthethonqubo owabiwe wokuxhumana. Ukuzixhuma kudinga ukuhlanganiswa okubalekile futhi okungaphindwe okuphindwe okuphindwe kabizayo ukwakha.

I-ACP imele ukuguquka okuyisisekelo: kusuka ku-Frodged, okukhethekile I-EcoSystem kwinethiwekhi exhunyiwe yama-ejenti axhumekile – ngalinye likwazi ukuthola, ukuqonda, futhi lihlanganyele nabanye, kungakhathalekile ukuthi ubani abawakhayo noma abasebenza kuwo. Nge-ACP, abathuthukisi bangabopha ubuhlakani obuhlangene bama-ejenti ahlukahlukene ukwakha ukugeleza okunamandla okwengeziwe kunohlelo olulodwa olungafinyelela ngedwa.

Izinselelo Zamanje:
Naphezu kokukhula okusheshayo kumakhono e-ejenti, ukuhlanganiswa kwangempela komhlaba kuhlala kuyinto enkulu ibhodlela elikhulu. Ngaphandle kweprotrolol ehlanganyelwe, izinhlangano zibhekene nezinselelo eziningana eziphindaphindekayo:

• Ukwehlukahlukana kohlaka: Izinhlangano zivame ukugijima amakhulukhulu noma izinkulungwane zama-ejenti zakhiwe zisebenzisa izinhlaka ezahlukahlukene njengeLangchain, Crewai, Autogen, noma izitabane zangokwezifiso.
• Ukuhlanganiswa okwenziwe ngokwezifiso: Ngaphandle kwephrothokholi ejwayelekile, abathuthukisi kufanele babhale izixhumi ngokwezifiso kuwo wonke ama-agent intentana.
• I-Exponential Development: Nge-N Agents, ungadinga ukudinga ama-N (N-1) / Amaphoyinti ama-2 ahlukene (enza ama-ejenti amakhulu ahlukahlukene okunzima ukuze kunzima ukuligcina).
• Ukucatshangelwa Kwenhlangano Esenqamulweni: Izinhlobo ezahlukahlukene zokuphepha, amasistimu wokufakazela ubuqiniso, kanye namafomethi wedatha ukuhlanganiswa okunzima kuzo zonke izinkampani.

Isibonelo Sangempela Somhlaba

Ukubonisa isidingo sangempela somhlaba sokuxhumana kwe-ejenti-ku-Agent, cabanga ngezinhlangano ezimbili:

Inkampani yokukhiqiza Lokho kusebenzisa i-ajenzeli ye-AI ukuphatha amashejuli wokukhiqiza nokugcwaliseka okusekelwe ekusungweni kwangaphakathi nokufunwa kwamakhasimende.

Umhlinzeki wezinto Lokho kugijimisa i-ejenti ukunikela ngezilinganiso zangempela zokuthumela, ukutholakala kwenkampani yenethiwekhi, kanye nokwenza kahle komzila.

Manje ake ucabange ukuthi uhlelo lomkhiqizi ludinga ukulinganisa ukuhamba kwezikhathi zokulethwa kwemishini enkulu, eyenziwe ngokwezifiso ukuze wazise isilinganiso samakhasimende.

Ngaphandle kwe-ACP: Lokhu kuzodinga ukwakha ukuhlanganiswa kwe-bespoke phakathi kwesoftware yokuhlela kamenzi kanye nama-apis ahlinzekwayo we-Logistic. Lokhu kusho ukuphatha ubuqiniso, ukulungiswa kwefomethi yedatha, kanye nokutholakala kwensiza ngesandla. Lokhu kuhlanganiswa kuyabiza, i-brittle, futhi kunzima ukukala njengoba abalingani abaningi bajoyina.

Nge-ACP: Inhlangano ngayinye isopha i-ejenti yayo ngesibonisi se-ACP. I-ejenti yokukhiqiza ithumela i-oda nemininingwane yokuya kuyo ku-ejenti ye-Logistics, ephendula ngezinketho zokuthumela zesikhathi sangempela kanye ne-etas. Zombili lezi zinhlelo zisebenzisana ngaphandle kokudalula abantu abangaphakathi noma ukubhala ukuhlanganiswa ngokwezifiso. Abalingani be-New Logistics bangasixhuma ngokumane basebenzise i-ACP.

Kulula kangakanani ukudala umenzeli ohambisanayo we-ACP?

Ukulula ukusungulwa komqondo oyisisekelo we-ACP. Ukusonga i-ejenti nge-ACP kungenziwa ngemigqa embalwa yekhodi. Usebenzisa i-Python SDK, ungachaza i-ejenti ehambisanayo ye-ACP ngokumane uhlobise umsebenzi.

Lokhu kuqaliswa okuncane:

1. Kwakha isibonelo seseva ye-ACP
2. Chaza umsebenzi we-ejenti usebenzisa i- @ server.agent () umhlobiso
3. Isebenzisa umenzeli usebenzisa uhlaka lwe-langchain nge-LLM backlend kanye nememori yokuphikelela komongo
4. Kuhunyushwa phakathi kwefomethi yemiyalezo ye-ACP kanye nefomethi yendabuko yohlaka ukuze ubuyise impendulo ehlelekile
5. Iqala iseva, okwenza i-ejenti itholakale nge-http

from typing import Annotated
import os
from typing_extensions import TypedDict
from dotenv import load_dotenv
#ACP SDK
from acp_sdk.models import Message
from acp_sdk.models.models import MessagePart
from acp_sdk.server import RunYield, RunYieldResume, Server
from collections.abc import AsyncGenerator
#Langchain SDK
from langgraph.graph.message import add_messages
from langchain_anthropic import ChatAnthropic 

load_dotenv() 

class State(TypedDict):
    messages: Annotated[list, add_messages]

#Set up the llm
llm = ChatAnthropic(model="claude-3-5-sonnet-latest", api_key=os.environ.get("ANTHROPIC_API_KEY"))

#------ACP Requirement-------#
#START SERVER
server = Server()
#WRAP AGENT IN DECORACTOR
@server.agent()
async def chatbot(messages: list[Message])-> AsyncGenerator[RunYield, RunYieldResume]:
    "A simple chatbot enabled with memory"
    #formats ACP Message format to be compatible with what langchain expects
    query = " ".join(
        part.content
        for m in messages
        for part in m.parts             
    )
    #invokes llm
    llm_response = llm.invoke(query)    
    #formats langchain response to ACP compatable output
    assistant_message = Message(parts=[MessagePart(content=llm_response.content)])
    # Yield so add_messages merges it into state
    yield {"messages": [assistant_message]}  

server.run()
#---------------------------#

Manje, udale i-ACEEN ehambisanayo ephelele engakwazi:

• Ukutholwa abanye ama-ejenti (online noma okungaxhunyiwe ku-inthanethi)
• Izicelo zenqubo ezivumelaniswa ngokuvumelanisa noma ngokusobala
• Ukuxhumana usebenzisa amafomethi womlayezo ojwayelekile
• Hlanganisa nanoma yiluphi olunye uhlelo oluhambisanayo lwe-ACP

I-ACP iqhathanisa kanjani ne-MCP & A2A

Ukuze uqonde kangcono iqhaza le-ACP kwi-Ecosystem evelayo ye-AI, kuyasiza ukuyiqhathanisa nezinye izinqubomgomo ezivelayo. Lezi protocols akuyona izimbangi ezithile. Esikhundleni salokho, babhekana nezendlalelo ezahlukahlukene zokuhlanganiswa kohlelo lwe-AI System futhi kuvame ukuhambisana.

Ngokubheka nje:

• I-MCP (imodeli yemodeli ye-Anthropic's Model): Yakhelwe ukucebisa umongo owodwa wemodeli ngamathuluzi, inkumbulo, kanye nezinsizakusebenza.
  Gxila: Imodeli eyodwa, amathuluzi amaningi
• I-ACP (I-Lenux Foundation's Agent Communt Protocol): Yakhelwe ukuxhumana phakathi kwama-ejenti azimele kulo lonke izinhlangano nezinhlangano.
  Gxila: Ama-ejenti amaningi, asebenze ngokuphepha njengontanga, akukho vendori Lock in, vula ukuvula
• I-A2A (i-ejenti ye-ejenti yakwaGoogle): Yakhelwe ukuxhumana phakathi kwama-ejenti azimele kulo lonke izinhlangano nezinhlangano.
  Gxila: Ama-ejenti amaningi, asebenza njengontanga, elungiselelwe i-ecosystem yakwaGoogle

I-ACP ne-MCP

Ithimba le-ACP ekuqaleni lahlola ukuvumelanisa I-Protocol Modelconglect Protocol (MCP) Ngoba inikeza isisekelo esiqinile sokusebenzisana umongo wemodeli. Kodwa-ke, basheshe bahlangabezana nemikhawulo yokwakha eyenza ayifanele ukuxhumana kwe-ejenti yangempela.

Kungani i-MCP ithola amafushane ngezinhlelo zemisebenzi eminingi:

Ukusakaza: I-MCP isekela ukusakazwa kepha ayiphathi Imifudlana yeDelta (isib., amathokheni, ukuvuselelwa kwe-trajectory). Lo mkhawulo uvela eqinisweni lokuthi lapho i-MCP ekuqaleni yadalwa yayingahloselwe ukusebenzisana kwe-ejenti-isitayela.

Ukwabelana Ngenkumbulo: I-MCP ayisekeli ukusebenzisa ama-ejenti amaningi amaseva ngenkathi igcina imemori eyabiwe. I-ACP ayisekeli ngokuphelele lokhu okwamanje, kepha yindawo esebenzayo yentuthuko.

Isakhiwo semilayezo: I-MCP yamukela noma iyiphi i-json schema kepha ayichazi isakhiwo somzimba womlayezo. Lokhu kuguquguquka kwenza ukuba ukusebenzisana kube nzima (ikakhulukazi ekwakheni ama-UIs noma ama-ejenti ama-orchestrating okumele ahumushe amafomethi wemiyalezo ehlukahlukene).

I-Protocol eyinkimbinkimbi: I-MCP isebenzisa i-JSON-RPC futhi idinga ama-SDK athile nezikhathi zokudlala. Lapho njengokwakheka okususelwa ekuphumeni kwe-ACP ngokuxhaswa okwakhelwe ngaphakathi kwe-ASYNC / SYNC kulula kakhulu futhi kuhlanganiswe – kunobungane.

Ungafunda kabanzi ngokuthi i-ACP ne-MCP iqhathanisa ngayo lapha.

Cabanga nge I-MCP Njengoba unikeza umuntu amathuluzi angcono, njenge-Calculator noma incwadi eyinkomba, ukuthuthukisa ukusebenza kwawo. Ngokungqubuzanayo, Ikalibhodi kumayelana nokunika amandla abantu ukwakha Amaqembulapho umuntu ngamunye (noma i-ejenti) ngayinye enikela ngamakhono abo futhi asebenzisane.

I-ACP ne-MCP kungahambisana

I-ACP ne-A2A

I-Protocol ye-Agent-to-agent (A2A), eyalethwa ngemuva nje kwe-ACP, futhi ihlose ukukwenza kube ngcono ukuxhumana phakathi kwama-ejenti we-AI. Womabili amaphrothokholi abelana ngenhloso yokunika amandla amasistimu we-ejenti amaningi, kepha aphambuka kwifilosofi kanye nokubusa.

Umehluko Osemqoka:

I-ACP yenzelwe ukuba ngamabomu ukuba:

• Elula ukuhlanganisa usebenzisa amathuluzi ajwayelekile we-HTTP kanye nemihlangano yokuphumula
• Iguquguqukayo ngaphesheya kwezinhlobonhlobo ze-ejenti kanye nemithwalo yomsebenzi
• I-Vendor-hlangothi, ngokuphathwa okuvulekile kanye nokuqondanisa okubanzi kwe-ecosystem

Womabili amaphrothokholi angahlala ndawonye – ngalinye lisebenzela izidingo ezihlukile ngokuya ngemvelo. I-ACP engasindi elingasindi, evulekile, futhi idizayini eyandisekayo yenza kube lufaneleke kahle ezinhlelweni ezihlukanisiwe kanye nokusebenzisana kwangempela emhlabeni wonke ngemingcele yenhlangano. Ukuhlanganiswa kwemvelo ka-A2A kungenza kube yindlela efanelekile kulabo abasebenzisa i-Google Ecosystem.

I-Roadmap kanye nomphakathi

Njengoba i-ACP ivela, bahlola amathuba amasha okuthuthukisa ukuxhumana kwe-ejenti. Nazi ezinye izindawo ezibalulekile zokugxila:

• I-BOTITY Federation: I-ACP ingasebenza kanjani ngezinhlelo zokuqinisekisa ukuthuthukisa ukwethenjwa kuwo wonke amanethiwekhi?
• Ukunikezwa kwe-Access: Singakwenza kanjani ukunika amandla ama-ejenti ukuthi anikeze ama-ejeka athumele imisebenzi ngokuphepha (ngenkathi egcina ukulawulwa komsebenzisi)?
• Ukuxhaswa kwerejista okuningi: Ngabe i-ACP ingasekela ukutholakala kwe-ejenti enqunyelwe kuwo wonke amanethiwekhi ahlukile?
• Ukwabelana nge-Agent: Singakwenza kanjani kube lula ukwaba futhi usebenzise kabusha ama-Agents ezinhlanganweni ezinhlanganweni noma enhlanganweni?
• Ukuhanjiswa: Yimaphi amathuluzi nezifanekiso angenza lula ukuthunyelwa kwe-Agent?

I-ACP yathuthukiswa obala ngoba amazinga asebenza kangcono lapho athuthukiswa ngqo nabasebenzisi. Iminikelo evela konjiniyela, abacwaningi, kanye nezinhlangano ezinentshisekelo ngekusasa lokusebenzisana kwe-ejenti kwamukelekile. Ujoyine ekusizeni ukwakheka kwale nkampani evelayo.

Ngeminye imininingwane, vakashela i-agenccommunicationprotocol.dev futhi ujoyine ingxoxo kwi-github kanye nama-ditrard iziteshi.