Kumanga Agents Anzeru ndi LangChain: Orchestrating LLMs for Real-World Application

Kumanga Ma Agents Anzeru ndi LangChain: Kupanga ma LLM a Real-World Application

Chiyambi

Kubwera kwa Large Language Models (LLMs) kwasintha kwambiri gawo la Natural Language Processing (NLP) ndipo kwatsegula njira zatsopano zopangira zida zanzeru zomwe zimatha kumvetsetsa, kupanga, ndikulumikizana ndi chilankhulo cha anthu. LangChain ndi laibulale ya Python yomwe imapereka mawonekedwe osinthika komanso owopsa omangira anzeru omwe amagwiritsa ntchito ma LLM. M'nkhaniyi, tiwona kusakanikirana kwa LangChain popanga othandizira anzeru, poyang'ana njira zogwirira ntchito zothandiza, njira zoyimba, komanso kugwiritsa ntchito milandu muzojambula zamapulogalamu.

Kodi Intelligent Agents ndi chiyani?

Othandizira anzeru ndi mapulogalamu apakompyuta omwe amatha kuzindikira malo awo, kulingalira za zomwe amalandira, ndikuchitapo kanthu kuti akwaniritse zolinga zawo. Atha kuganiziridwa ngati ochita zisankho odziyimira pawokha omwe amatha kulumikizana ndi ogwiritsa ntchito, othandizira ena, kapena machitidwe kuti akwaniritse cholinga china. Othandizira anzeru amatha kugawidwa m'mitundu yosiyanasiyana, kuphatikiza:

• Othandizira pamalamulo: Othandizirawa amagwiritsa ntchito malamulo omwe adafotokozedwa kale popanga zisankho.
• Othandizira kuphunzira pamakina: Othandizirawa amagwiritsa ntchito njira zophunzirira makina kuti aphunzire kuchokera pa data ndikupanga zisankho.
• Hybrid agents: Othandizirawa amaphatikiza njira zotsata malamulo ndi makina ophunzirira kupanga zisankho.

LangChain ndi chiyani?

LangChain ndi laibulale ya Python yomwe imapereka mawonekedwe osinthika komanso owopsa omangira anzeru omwe amagwiritsa ntchito ma LLM. Zimalola opanga kupanga, kuphunzitsa, ndi kutumiza ma LLM, komanso kuwaphatikiza ndi mitundu ina ya AI ndi machitidwe. LangChain imapereka zinthu zingapo, kuphatikiza:

• ** Kuphatikizika kwa LLM **: LangChain imapereka njira yosavuta komanso yothandiza yophatikizira ma LLM ndi mitundu ina ya AI ndi machitidwe.
• ** Orchestration **: LangChain imapereka njira zingapo zoyimba, kuphatikizapo kayendetsedwe ka ntchito, kayendetsedwe ka kayendetsedwe ka ntchito, ndi kayendetsedwe ka deta.
• Scalability: LangChain idapangidwa kuti izikhala ndi ma dataseti akulu komanso mayendedwe ovuta.

Njira Zothandizira Zothandizira

Kuti mupange othandizira anzeru ndi LangChain, muyenera kutsatira njira izi:

1. Sankhani LLM: Sankhani LLM yomwe ikukwaniritsa zomwe mukufuna, monga chinenero, kukula kwake, ndi zovuta zake.
2. Konzani deta: Konzani deta yomwe ikugwirizana ndi momwe mungagwiritsire ntchito komanso yoyenera kuphunzitsa LLM.
3. Phunzitsani LLM: Phunzitsani LLM pogwiritsa ntchito deta yokonzedwa ndi LangChain's training API.
4. ** Phatikizani ndi mitundu ina ya AI **: Phatikizani LLM yophunzitsidwa ndi mitundu ina ya AI ndi machitidwe pogwiritsa ntchito LangChain's integration API.
5. ** Ntchito za Orchestrate **: Gwiritsani ntchito API ya orchestration ya LangChain kuti muyang'ane ntchito, kayendetsedwe ka ntchito, ndi deta.

Njira Zoimbaimba

LangChain imapereka njira zingapo zoyimba zomwe zingagwiritsidwe ntchito kuyang'anira ntchito, kuyenda kwa ntchito, ndi deta. Zina mwa njira zazikulu za orchestration ndi izi:

• Kasamalidwe ka ntchito: LangChain imapereka API yoyang'anira ntchito yomwe imalola opanga kupanga, kuyang'anira, ndi kuchita ntchito.
• **Kuwongolera kayendetsedwe ka ntchito **: LangChain imapereka API yoyang'anira kayendedwe ka ntchito yomwe imalola opanga kupanga, kuyang'anira, ndikuchita mayendedwe.
• **Kasamalidwe ka data **: LangChain imapereka API yoyang'anira deta yomwe imalola opanga kuwongolera ndikuwongolera deta.

**Gwiritsani ntchito Milandu mu Uinjiniya wa Mapulogalamu **

LangChain angagwiritsidwe ntchito zosiyanasiyana mapulogalamu uinjiniya milandu, kuphatikizapo:

• Chatbots: LangChain itha kugwiritsidwa ntchito kupanga ma chatbots omwe amatha kumvetsetsa, kupanga, komanso kulumikizana ndi chilankhulo cha anthu.
• ** Virtual assistants **: LangChain ingagwiritsidwe ntchito kupanga othandizira omwe amatha kugwira ntchito, kuyankha mafunso, ndikupereka chidziwitso.
• **M'badwo wazinthu **: LangChain ikhoza kugwiritsidwa ntchito kupanga zinthu, monga zolemba, zolemba zamabulogu, ndi zolemba zapa media.

**Malamulo/Zidutswa Zapamwamba za 5-10 **

Nawa maulamuliro/zidutswa 5-10 zapamwamba kwambiri za LangChain:

1. Kutumiza LangChain: kulowetsa langchain
2. Kupanga LLM: llm = langchain.LLM("my_llm")
3. Kuphunzitsa LLM: llm.train(data)
4. Kuphatikiza ndi mitundu ina ya AI: ai_model = langchain.AIModel("my_ai_model")
5. Ntchito zowongolera: task = langchain.Task("my_task")
6. Kuwongolera kayendedwe ka ntchito: kayendetsedwe ka ntchito = langchain.Workflow("my_workflow")
7. Kusamalira deta: data = langchain.Data("my_data")
8. Kupanga chatbot: chatbot = langchain.Chatbot("my_chatbot")
9. Kupanga wothandizira: virtual_assistant = langchain.VirtualAssistant("my_virtual_assistant")
10. Kupanga zinthu: zamkati = langchain.ContentGenerator("my_content_generator")

**Kuwona **

Pano pali chithunzithunzi cha zomangamanga za LangChain pogwiritsa ntchito Chart.js:

{
  "type": "bar",
  "data": {
    "labels": ["LLM", "Orchestration", "Integration", "Task Management", "Workflow Management", "Data Management"],
    "ma dataset": [
      {
        "label": "LangChain Architecture",
        "data": [10, 8, 6, 4, 2, 1],
        "backgroundColor": ["#4CAF50", "#FF9800", "#2196F3", "#8BC34A", "#03A9F4", "#FFC107"]
      }
    ]
  },
  "zosankha": {
    "mutu": {
      "kuwonetsa": zoona,
      "text": "LangChain Architecture"
    },
    "nthano": {
      "chiwonetsero": zoona
    }
  }
}

Mawonedwe awa akuwonetsa zigawo zosiyana za zomangamanga za LangChain, kuphatikizapo LLMs, orchestration, kusakanikirana, kayendetsedwe ka ntchito, kayendetsedwe ka kayendetsedwe ka ntchito, ndi kayendetsedwe ka deta.

Mapeto

LangChain ndi laibulale yamphamvu yomwe imapereka mawonekedwe osinthika komanso owopsa omangira anzeru omwe amagwiritsa ntchito ma LLM. Potsatira njira zoyendetsera ntchito ndi njira zoyimba zomwe zafotokozedwa m'nkhaniyi, okonza amatha kupanga zida zanzeru zomwe zimatha kumvetsetsa, kupanga, ndi kuyanjana ndi chilankhulo cha anthu. Zomwe zimagwiritsidwa ntchito muzojambula zamapulogalamu zimasonyeza kuthekera kwa LangChain muzogwiritsira ntchito zosiyanasiyana, kuchokera ku ma chatbots ndi othandizira pafupifupi ku mbadwo wokhutira ndi zina. Ndi LangChain, opanga amatha kupanga othandizira anzeru omwe amatha kugwira ntchito zovuta komanso kupereka phindu kwa ogwiritsa ntchito.

Block kodi

Nayi chipika chosavuta chomwe chikuwonetsa momwe mungapangire chatbot pogwiritsa ntchito LangChain:

import langchain


llm = langchain.LLM("my_llm")

# Pangani chatbot
chatbot = langchain.Chatbot(llm)

# Fotokozani ntchito
task = langchain.Task("my_task")

# Gwirani ntchito
zotsatira = chatbot.execute (ntchito)

# Sindikizani zotsatira
sindikiza (zotsatira)

Khodi iyi imapanga LLM, chatbot, ndi ntchito, kenako imagwira ntchitoyo pogwiritsa ntchito chatbot. Zotsatira zake zimasindikizidwa ku console.

Dzimbiri ndi Go Code

LangChain imaperekanso zomangira za Rust ndi Go, zomwe zitha kugwiritsidwa ntchito popanga zida zanzeru m'zilankhulo izi. Nachi chitsanzo cha chipika chosavuta cha Rust chomwe chikuwonetsa momwe mungapangire chatbot pogwiritsa ntchito LangChain:

gwiritsani ntchito langchain::{LLM, Chatbot};

fn chachikulu () {
    // Pangani LLM
    let llm = LLM:: new("my_llm");

    // Pangani chatbot
    lolani chatbot = Chatbot::new(llm);

    // Fotokozani ntchito
    lolani ntchito = Ntchito :: yatsopano ("my_task");

    // Gwirani ntchito
    lolani zotsatira = chatbot.execute (ntchito);

    // Sindikizani zotsatira
    println!("{}", chotsatira);
}

Ndipo nachi chitsanzo cha chipika chosavuta cha Go chomwe chikuwonetsa momwe mungapangire chatbot pogwiritsa ntchito LangChain:

phukusi lalikulu

itanitsa (
	"fmt"
	"github.com/langchain/langchain-go"
)

ntchito yaikulu () {
	// Pangani LLM
	llm := langchain.NewLLM("my_llm")

	// Pangani chatbot
	chatbot := langchain.NewChatbot(llm)

	// Fotokozani ntchito
	ntchito := langchain.NewTask("my_task")

	// Gwirani ntchito
	zotsatira:= chatbot.Execute(ntchito)

	// Sindikizani zotsatira
	fmt.Println(zotsatira)
}

Ma code awa akuwonetsa momwe mungapangire chatbot pogwiritsa ntchito LangChain mu Rust and Go.