I-Orchestrating AI Agents nge-LangChain ne-DSPy Yokwakhiwa Kwesofthiwe Ethuthukisiwe | B4N1

I-Orchestrating AI Agents nge-LangChain ne-DSPy Yokwakhiwa Kwesofthiwe Ethuthukisiwe

8 months ago · b4n1 Read time: 3 min

Ukuhlela ama-AI Agents nge-LangChain ne-DSPy ye-Enhanced Software Architecture

Isingeniso

Ezweni lanamuhla ledijithali elisheshayo, amasistimu esofthiwe aya ngokuya eba yinkimbinkimbi, ngokuhlanganiswa kwamakhono e-Artificial Intelligence (AI) kanye Nokufunda Ngomshini (ML). Ngenxa yalokhu, ukuklama amasistimu angakala futhi aqinile angahlela ngempumelelo ama-ejenti e-AI sekuyinkinga enkulu konjiniyela besoftware. Lesi sihloko sihlola ukuhlangana kwe-AI, ubunjiniyela besofthiwe, namathuluzi omculo wokuklama izakhiwo zesofthiwe ezisebenzayo, ngokugxila ekusebenziseni ama-agent e-AI kusetshenziswa i-LangChain ne-DSPy.

Isidingo se-AI Orchestration

Amasistimu e-software endabuko avame ukuklanywa nge-architecture ye-monolithic, lapho zonke izingxenye zihlanganiswa ngokuqinile futhi zisebenza ndawonye ngendlela yomugqa. Kodwa-ke, ngokufika kwe-AI ne-ML, izinhlelo zesofthiwe ziya ngokuya zisakazwa futhi zihlukaniswa, nama-ejenti amaningi asebenza ndawonye ukuze afinyelele umgomo ofanayo. Lokhu kuholele esidingweni sokucula kwe-AI, okubandakanya ukuklama amasistimu angaphatha ngempumelelo futhi axhumanise ukusebenzisana phakathi kwama-agent amaningi e-AI.

LangChain: Uhlaka Lokwakha Ama-AI Agents

I-LangChain iwuhlaka lwePython lokwakha ama-agent e-AI angaxhumana nomunye kanye nezinhlelo zangaphandle. Ihlinzeka ngesethi yamathuluzi nemitapo yolwazi evumela abathuthukisi ukuthi bakhe ama-ejenti angakwazi ukucabanga, afunde, futhi azivumelanise nezimo ezishintshayo. I-LangChain yakhelwe phezu kolimi lwezinhlelo ze-Rust futhi inikeza i-API yezinga eliphezulu yokwakha ama-agent e-AI.

I-DSPy: Uhlaka Lwe-Orchestration

I-DSPy iwuhlaka lwePython lokwakha amasistimu asabalalisiwe angahlela ama-agent amaningi e-AI. Ihlinzeka ngesethi yamathuluzi nemitapo yolwazi enika amandla onjiniyela ukuthi badizayine futhi basebenzise amasistimu angakala futhi aqinile angaphatha ukusebenzisana okuyinkimbinkimbi phakathi kwama-ejenti amaningi. I-DSPy yakhelwe phezu kwesistimu yemiyalezo ye-Apache Kafka futhi inikeza i-API yezinga eliphezulu yokwakha amasistimu asabalalisiwe.

Ukusebenzisa ama-AI Agents nge-LangChain ne-DSPy

Ukuze usebenzise ama-agent e-AI usebenzisa i-LangChain ne-DSPy, sidinga ukulandela lezi zinyathelo:

  1. Chaza Isakhiwo Somenzeli: Chaza i-architecture ye-ejenti ye-AI, okuhlanganisa okokufaka, okuphumayo, nokusebenzisana nabanye abasebenzeli.
  2. Sebenzisa I-Agent Logic: Sebenzisa ingqondo ye-ejenti ye-AI usebenzisa i-LangChain, okuhlanganisa nokucabanga, ukufunda, nezindlela zokuzivumelanisa nezimo.
  3. Hlela Ukusebenzisana Kwe-ejenti: Sebenzisa i-DSPy ukuze uhlele ukusebenzisana phakathi kwama-ejenti amaningi e-AI, okuhlanganisa izimiso zokuxhumana nezindlela zokushintshisana ngedatha.
  4. Khipha Uhlelo: Sebenzisa isistimu usebenzisa inkundla yefu noma uhlaka lokuqukatha, olufana ne-Docker.

Isibonelo Sekhodi: Ukusebenzisa I-ejenti Elula ye-AI nge-LangChain

Nasi isibonelo sekhodi esilula esibonisa indlela yokusebenzisa i-ejenti ye-AI usebenzisa i-LangChain:

ngenisa i-langchain
kusuka ku-langchain.chains ngenisa i-LLMChain


i-ejenti = langchain.Agent(
    input_schema={"umbhalo": str},
    output_schema={"umbhalo": str},
    iketango=LLMChain(
        imodeli="davinci:instructix",
        ubuningi_ubude=2048,
        izinga lokushisa=0.7,
    ),
)

# Sebenzisa i-logic ye-ejenti
def agent_logic(input_text):
    okuphumayo_umbhalo = i-agent.run(input_text)
    buyisela okuphumayo_umbhalo

# Hlola i-ejenti
input_text = "Iyini incazelo yokuphila?"
okuphumayo_umbhalo = i-ejenti_logic(input_text)
phrinta(output_text)

Lesi sibonelo sekhodi sibonisa indlela yokuchaza i-ejenti ye-AI kusetshenziswa i-LangChain, okuhlanganisa ama-schema okokufaka nokukhiphayo, kanye nengqondo ye-ejenti. I-ejenti isetshenziswa kusetshenziswa iketango Lemodeli Yolimi Olukhulu (LLM), okuwuhlobo lwe-ejenti ye-AI engasebenzisa ingqondo futhi ikhiqize umbhalo.

Imiyalo/Amazwibela asetshenziswa kakhulu angu-5-10 we-LangChain ne-DSPy

Nansi imiyalo/ amazwibela aphezulu angu-5-10 e-LangChain ne-DSPy:

  1. langchain.Agent(): Idala isenzakalo esisha somenzeli we-AI.
i-ejenti = langchain.Agent(
    input_schema={"umbhalo": str},
    output_schema={"umbhalo": str},
    iketango=LLMChain(
        imodeli="davinci:instructix",
        ubuningi_ubude=2048,
        izinga lokushisa=0.7,
    ),
)
  1. i-agent.run(): Isebenzisa i-ejenti ye-AI ngokufaka okunikeziwe.
okuphumayo_umbhalo = i-agent.run(input_text)
  1. DSPy.create_cluster(): Idala iqoqo elisha lamanodi esistimu esabalalisiwe.
iqoqo = DSPy.create_cluster(
    inombolo_amanodi=5,
    node_type = "isisebenzi",
)
  1. I-DSPy.create_agent(): Idala isenzakalo esisha somenzeli we-AI wesistimu esabalalisiwe.
i-ejenti = DSPy.create_agent(
    iqoqo=iqoqo,
    uhlobo_lwe-ejenti="LLM",
)
  1. I-DSPy.deploy_agent(): Ithumela i-ejenti ye-AI kusistimu esabalalisiwe.
DSPy.deploy_ejenti(
    i-ejenti=i-ejenti,
    iqoqo=iqoqo,
)
  1. DSPy.get_agent_status(): Ibuyisa isimo somenzeli we-AI.
isimo = DSPy.get_agent_status(
    i-ejenti=i-ejenti,
    iqoqo=iqoqo,
)
  1. I-DSPy.update_agent(): Ibuyekeza i-ejenti ye-AI ngedatha entsha noma amapharamitha.
DSPy.update_ejenti(
    i-ejenti=i-ejenti,
    iqoqo=iqoqo,
    new_data={"text": "umbhalo wokufakwayo omusha"},
)
  1. I-DSPy.delete_agent(): Isusa i-ejenti ye-AI ohlelweni olusabalalisiwe.
DSPy.delete_ejenti(
    i-ejenti=i-ejenti,
    iqoqo=iqoqo,
)
  1. I-DSPy.create_chain(): Idala uchungechunge olusha lwabasebenzeli be-AI besistimu esabalalisiwe.
iketango = DSPy.create_chain(
    ama-ejenti=[i-ejenti1, i-ejenti2, i-ejenti3],
)
  1. I-DSPy.deploy_chain(): Ithumela uchungechunge lwama-ejenti e-AI ohlelweni olusabalalisiwe.
I-DSPy.deploy_chain(
    iketango=ketango,
    iqoqo=iqoqo,
)

Ukubona

Nakhu ukubonwa kwesakhiwo se-LangChain ne-DSPy kusetshenziswa i-Chart.js:

{
  "type": "pieChart",
  "idatha": {
    "labels": ["LangChain", "DSPy"],
    "idathasethi": [
      {
        "label": "AI Agents",
        "idatha": [50, 50],
        "backgroundColor": ["#4CAF50", "#FF9800"]
      }
    ]
  },
  "izinketho": {
    "isihloko": {
      "Display": iqiniso,
      "text": "LangChain kanye ne-DSPy Architecture"
    },
    "i-legend": {
      "isibonisi": iqiniso
    }
  }
}

Lokhu kuboniswa kubonisa ukusatshalaliswa kwama-ejenti e-AI phakathi kwe-LangChain ne-DSPy, nama-ejenti angu-50% asebenza ku-LangChain futhi angu-50% asebenza ku-DSPy.

Isiphetho

I-LangChain ne-DSPy ihlinzeka ngohlaka olunamandla lokwakha nokuhlela ama-ejenti e-AI, olwenza abathuthukisi bakwazi ukuklama amasistimu angalawuleki futhi aqinile angaphatha ngempumelelo ukusebenzisana okuyinkimbinkimbi phakathi kwama-ejenti amaningi. Ngokulandela izinyathelo ezichazwe kulesi sihloko, abathuthukisi bangasebenzisa ama-agent e-AI basebenzisa i-LangChain ne-DSPy, futhi bawathumele ohlelweni olusabalalisiwe usebenzisa i-DSPy. Izibonelo zekhodi nokuboniswa okuhlinzekiwe kubonisa amandla e-LangChain ne-DSPy, futhi imiyalo/ amazwibela aphezulu angu-5-10 anikeza inkomba esheshayo yabathuthukisi.

Have Questions?

Let's Talk

Contact me on LinkedIn