** Ma'aikatan AI tare da LangChain da DSPy don Inganta Gine-ginen Software ***

** Gabatarwa ***

A cikin shimfidar wuri mai sauri na dijital na yau, tsarin software yana ƙara rikiɗawa, tare da haɗakar fasahar Artificial Intelligence (AI) da damar Koyon Inji (ML). Sakamakon haka, ƙirƙira tsarin daidaitawa da juriya waɗanda za su iya tsara ma'aikatan AI yadda ya kamata ya zama abin damuwa ga injiniyoyin software. Wannan labarin yana bincika haɗin kai na AI, injiniyan software, da kayan aikin ƙididdiga don tsara ingantaccen tsarin gine-ginen software, tare da mai da hankali kan aiwatar da wakilai AI ta amfani da LangChain da DSPy.

** Bukatar AI Orchestration ***

Ana tsara tsarin software na gargajiya sau da yawa tare da gine-gine na monolithic, inda duk abubuwan da aka haɗa aka haɗa su tare kuma suna aiki tare cikin layin layi. Duk da haka, tare da zuwan AI da ML, tsarin software yana karuwa da rarrabawa, tare da wakilai da yawa suna aiki tare don cimma manufa ɗaya. Wannan ya haifar da buƙatar ƙungiyar AI, wanda ya haɗa da tsara tsarin da zai iya sarrafawa yadda ya kamata da daidaita hulɗar tsakanin ma'aikatan AI da yawa.

** LangChain: Tsarin Gina Wakilan AI ***

LangChain shine tsarin Python don gina wakilai na AI waɗanda zasu iya hulɗa da juna da kuma tsarin waje. Yana ba da saitin kayan aiki da ɗakunan karatu waɗanda ke ba masu haɓakawa damar ƙirƙirar wakilai waɗanda za su iya yin tunani, koyo, da kuma daidaita yanayin yanayi. An gina LangChain a saman harshen shirye-shiryen Rust kuma yana ba da babban matakin API don gina wakilai AI.

** DSPy: Tsari don Orchestration ***

DSPy shine tsarin Python don gina tsarin rarrabawa wanda zai iya tsara ma'aikatan AI da yawa. Yana ba da saiti na kayan aiki da ɗakunan karatu waɗanda ke ba masu haɓaka damar tsarawa da ƙaddamar da tsarin daidaitawa da juriya waɗanda zasu iya sarrafa hadaddun hulɗar tsakanin ma'aikata da yawa. An gina DSPy a saman tsarin saƙon Apache Kafka kuma yana ba da babban matakin API don gina tsarin rarrabawa.

** Aiwatar da Wakilan AI tare da LangChain da DSPy ***

Don aiwatar da wakilan AI ta amfani da LangChain da DSPy, muna buƙatar bin waɗannan matakan:

** Ƙayyade Tsarin Gine-ginen Agent ***: Ƙayyade tsarin gine-gine na wakili na AI, ciki har da abubuwan da aka samu, abubuwan da aka samo, da kuma hulɗa tare da wasu wakilai.
** Aiwatar da Ma'anar Agent ***: Aiwatar da dabaru na wakilin AI ta amfani da LangChain, gami da tunani, koyo, da hanyoyin daidaitawa.
** Orchestrate the Agent Interactions ***: Yi amfani da DSPy don tsara hulɗar tsakanin wakilan AI da yawa, gami da ka'idojin sadarwa da hanyoyin musayar bayanai.
** Sanya Tsarin ***: Sanya tsarin ta amfani da dandamalin girgije ko tsarin kwantena, kamar Docker.

** Misalin lamba: Aiwatar da Sauƙaƙan Wakilin AI tare da LangChain ***

Ga misali mai sauƙi na lamba wanda ke nuna yadda ake aiwatar da wakilin AI ta amfani da LangChain: `` Python shigo da langchain daga langchain.chains shigo da LLMChain

wakili = langchain.Agent( input_schema={"rubutu": str}, Output_schema={"rubutu": str}, sarkar=LLMChain( model="davinci:instructix", max_length=2048, zafin jiki = 0.7, ), )

Aiwatar da dabaru na wakili

def agent_logic(input_text): fitarwa_text = wakili.run(input_text) mayar da fitarwa_text

Gwada wakili

input_text = "Mene ne ma'anar rayuwa?" fitarwa_text = agent_logic(input_text) buga (fitarwa_text)

Wannan misalin lambar yana nuna yadda ake ayyana wakili na AI ta amfani da LangChain, gami da tsarin shigarwa da tsarin fitarwa, da dabaru na wakili. Ana aiwatar da wakilin ta amfani da Sarkar Babban Harshe (LLM), wanda nau'in wakili ne na AI wanda zai iya yin tunani da samar da rubutu.

** Manyan 5-10 Mafi Amfani da Dokokin/Snippets don LangChain da DSPy ***

Anan akwai manyan umarni 5-10 da aka fi amfani da su/snippets don LangChain da DSPy:

1. ** langchain.Agent ()**: Ƙirƙirar sabon misali AI wakili.
`` Python
wakili = langchain.Agent(
    input_schema={"rubutu": str},
    Output_schema={"rubutu": str},
    sarkar=LLMChain(
        model="davinci:instructix",
        max_length=2048,
        zafin jiki = 0.7,
    ),
)

agent.run(): Gudanar da wakilin AI tare da shigar da aka ba. `` Python fitarwa_text = wakili.run(input_text)

3. ** DSPy.create_cluster()**: Yana ƙirƙira sabon gungu na nodes don tsarin rarraba.
`` Python
tari = DSPy.create_cluster(
    lamba_nodes=5,
    node_type = "ma'aikaci",
)

** DSPy.create_agent () ***: Ƙirƙiri sabon misali na AI don tsarin rarraba. `` Python wakili = DSPy.create_agent ( tari = tari, wakili_type = "LLM", )

5. ** DSPy.deploy_agent () ***: Sanya wakilin AI zuwa tsarin rarraba.
`` Python
DSPy.deploy_agent(
    wakili = wakili,
    tari = tari,
)

** DSPy.get_agent_status()**: Yana dawo da matsayin wakilin AI. `` Python matsayi = DSPy.get_agent_status( wakili = wakili, tari = tari, )

7. ** DSPy.update_agent () ***: Yana sabunta wakilin AI tare da sababbin bayanai ko sigogi.
`` Python
DSPy.update_agent(
    wakili = wakili,
    tari = tari,
    new_data={"rubutu": "sabon rubutu"},
)

** DSPy.delete_agent () ***: Yana share wakilin AI daga tsarin rarraba. `` Python DSPy.delete_agent( wakili = wakili, tari = tari, )

9. ** DSPy.create_chain () ***: Ƙirƙirar sabon jerin wakilan AI don tsarin rarraba.
`` Python
sarkar = DSPy.create_chain(
    wakilai=[wakili1, wakili2, wakili3],
)

** DSPy.deploy_chain () ***: Ƙaddamar da jerin sunayen wakilan AI zuwa tsarin rarraba. `` Python DSPy.deploy_chain( sarkar = sarka, tari = tari, )

**Kallon gani**

Anan ga hangen nesa na LangChain da gine-ginen DSPy ta amfani da Chart.js:
Json
{
  "type": "pieChart",
  "data": {
    "labels": ["LangChain", "DSPy"],
    "datasets": [
      {
        "label": "AI Agents",
        "bayanai": [50, 50],
        "Bayan launi": ["#4CAF50", "#FF9800"]
      }
    ]
  },
  "zaɓi": {
    "take": {
      "nuna": gaskiya,
      "rubutu": "LangChain da DSPy Architecture"
    },
    "tatsuniya": {
      "nuni": gaskiya
    }
  }
}

Wannan hangen nesa yana nuna rarraba wakilan AI tsakanin LangChain da DSPy, tare da 50% na wakilan da ke gudana akan LangChain da 50% suna gudana akan DSPy.

Kammalawa

LangChain da DSPy suna ba da tsari mai ƙarfi don ginawa da tsara ma'aikatan AI, suna ba masu haɓaka damar tsara tsarin daidaitawa da juriya waɗanda zasu iya sarrafa hadaddun hulɗar tsakanin ma'aikata da yawa. Ta bin matakan da aka zayyana a cikin wannan labarin, masu haɓakawa za su iya aiwatar da wakilan AI ta amfani da LangChain da DSPy, da kuma tura su zuwa tsarin da aka rarraba ta amfani da DSPy. Misalai na lambar da hangen nesa da aka bayar suna nuna iyawar LangChain da DSPy, kuma manyan 5-10 mafi yawan umarnin da aka yi amfani da su / snippets suna ba da tunani mai sauri ga masu haɓakawa.

Gudanar da Wakilan AI tare da LangChain da DSPy don Inganta Gine-ginen Software

Aiwatar da dabaru na wakili

Gwada wakili

¿Tienes preguntas?