** Gina Tsarin Software na Warkar da Kai tare da Multi-Agent AI Architectures da LangChain ***

** Gabatarwa ***

A cikin yanayin haɓaka software mai sauri na yau, gina tsarin software na warkar da kai ya zama muhimmin buƙatu don tabbatar da dogaro, haɓakawa, da juriya na hadadden tsarin software. Gine-gine na al'ada na al'ada ba su isa su cika buƙatun ci gaban software na zamani ba, inda tsarin dole ne su iya daidaita da buƙatu masu canzawa, magance gazawar da ba zato ba, da murmurewa daga kurakurai cikin sauri. Wannan shi ne inda gine-ginen AI masu yawa da LangChain suka shiga cikin wasa, suna ba da mafita mai ƙarfi don gina tsarin software na warkarwa wanda zai iya koyo, daidaitawa, da amsa yanayin canzawa.

** Menene Multi-Agent AI Architectures?**

Multi-agent AI gine-ginen wani nau'i ne na tsarin AI da aka rarraba wanda ya ƙunshi wakilai masu cin gashin kansu da yawa waɗanda ke hulɗa da juna da kuma yanayin su don cimma manufa ɗaya. Kowane wakili yana da nasa ikon yanke shawara, ilimi, da albarkatu, kuma za su iya sadarwa tare da juna don raba bayanai, daidaita ayyuka, da kuma daidaita yanayin canjin yanayi.

A cikin mahallin tsarin software, ana iya amfani da gine-ginen AI masu yawa don gina tsarin warkar da kai wanda zai iya ganowa da amsa kurakurai, gazawa, da sauran abubuwan da ba su da kyau a cikin ainihin lokaci. Ta hanyar rarraba tsarin yanke shawara a tsakanin wakilai da yawa, waɗannan tsarin zasu iya koyo daga gwaninta, daidaitawa ga yanayin canzawa, da murmurewa daga kurakurai da sauri.

** Menene LangChain?**

LangChain babban ɗakin karatu ne na Python mai buɗewa don gina tsarin AI mai wakilai da yawa. Yana ba da tsari mai sauƙi, sassauƙa, da ma'auni don gina hadaddun tsarin AI wanda zai iya koyo, daidaitawa, da amsa yanayin canzawa. An tsara LangChain don yin aiki tare da nau'i-nau'i na AI, ciki har da TensorFlow, PyTorch, da Keras, kuma ana iya amfani da su don gina tsarin da ke hade da nau'o'in bayanan bayanai, ciki har da bayanan bayanai, APIs, da na'urorin IoT.

** Gina Tsarin Software na Warkar da Kai tare da LangChain ***

Don gina tsarin software na warkar da kai tare da LangChain, muna buƙatar tsara tsarin gine-ginen AI mai yawa wanda zai iya ganowa da amsa kurakurai, gazawa, da sauran abubuwan da ba su da kyau a cikin ainihin lokaci. Anan ga babban bayyani na gine-ginen:

Agent Design: Muna buƙatar tsara tsarin wakilai masu cin gashin kansu waɗanda za su iya hulɗa da juna da muhallinsu don cimma manufa ɗaya. Kowane wakili ya kamata ya kasance yana da ikon yanke shawara, ilimi, da albarkatunsa.
** Hotunan Ilimi ***: Muna buƙatar ƙirƙirar jadawali na ilimi wanda ke wakiltar alaƙa tsakanin wakilai, yanayi, da manufofin tsarin. Ana iya amfani da wannan jadawali don yin tunani game da yanayin tsarin da kuma yanke shawara.
** Yanke Shawara ***: Muna buƙatar aiwatar da tsarin yanke shawara wanda zai ba da damar wakilai su yi tunani game da yanayin tsarin da yanke shawara dangane da jadawalin ilimin.
** Aiki Kisa ***: Muna buƙatar aiwatar da tsarin aiwatar da aikin da ke ba da damar wakilai su aiwatar da shawararsu da yin hulɗa tare da yanayi.

Ga misali mai sauƙi na tsarin software na warkar da kai wanda aka gina tare da LangChain: `` Python shigo da langchain daga langchain.Agents shigo da Agent daga langchain.llms shigo da LLM

jadawali = { 'wakilai': ['agent1', 'wakili2', 'wakili3'], 'yanayin': ['database','API','IoT na'urar'], 'manufa': ['gane kurakurai', 'amsa ga kasawa', 'murmurewa daga kurakurai'] }

Ƙayyade tsarin yanke shawara

def decision_making (wakili, ilimi_graph): # Dalili game da yanayin tsarin jiha = ilimi_graph ['agent] [wakili] # Yi shawara bisa ga jihar yanke shawara = 'amsa ga gazawa' mayar da shawara

Ƙayyade tsarin aiwatar da aikin

def action_execution (wakili, yanke shawara): # aiwatar da hukuncin idan yanke shawara == 'amsa ga gazawa': # Amsa ga gazawa buga ('Masanin gazawar...') dawo

Ƙirƙiri wakilin LangChain

wakili = wakili ( name='wakilin warkar da kai', yanke shawara = yanke shawara, action_execution=aiki_execution, ilimi_graph = jadawali )

Ƙirƙiri LangChain LLM

LLM = LLM suna = 'warkar da kai-llm', model='davinci' )

Ƙirƙiri tsarin LangChain

tsarin = langchain.System( wakili = wakili, lm=llm )

Gudanar da tsarin

tsarin.run()

Wannan misalin yana nuna tsarin software mai sauƙi wanda zai iya gano kurakurai, amsa gazawar, da dawo da kurakurai cikin sauri. Tsarin ya ƙunshi jerin wakilai masu cin gashin kansu waɗanda ke hulɗa da juna da muhallinsu don cimma manufa guda. Wakilai suna amfani da jadawali na ilimi don yin tunani game da yanayin tsarin kuma su yanke shawara bisa jadawali.

** Manyan Dokoki 5-10 da Akafi Amfani da su/Snippets**

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

1. `langchain.Agent()': Yana ƙirƙira sabon wakilin LangChain.
2. `langchain.LLM()`: Yana haifar da sabon LangChain LLM.
3. `langchain.System()`: Yana haifar da sabon tsarin LangChain.
4. `agent.decision_making()`: Yana gudanar da tsarin yanke shawara ga wakili.
5. `agent.action_execution()`: Yana gudanar da aikin aiwatar da aikin wakili.
6. `llm.predict()`: Yana yin tsinkaya ta amfani da LLM.
7. `system.run()`: Yana gudanar da tsarin LangChain.
8. `graph.add_node()`: Yana ƙara sabon kumburi zuwa jadawali na ilimi.
9. `graph.add_edge()`: Yana ƙara sabon gefen jadawali na ilimi.
10. `graph.remove_node()`: Yana cire kumburi daga jadawali na ilimi.

**Kallon gani**

Anan ga sauƙin gani na jadawali ilimin ta amfani da Chart.js:
Json
{
  "type": "graph",
  "data": {
    "nodes": [
      {"id": "agent1", "label": "Agent 1"},
      {"id": "agent2", "label": "Agent 2"},
      {"id": "agent3", "label": "Agent 3"},
      {"id": "database", "label": "Database"},
      {"id": "API", "label": "API"},
      {"id": "Na'urar IoT", "lakabi": "Na'urar IoT"}
    ],
    "baki": [
      {"source": "agent1", "target": "database"},
      {"source": "agent2", "manufa": "API"},
      {"source": "agent3", "manufa": "Na'urar IoT"},
      {"source": "agent1", "manufa": "agent2"},
      {"source": "agent2", "manufa": "agent3"}
    ]
  },
  "zaɓi": {
    "tsari": {
      "type": "circular"
    },
    "nodes": {
      "girma": 50
    },
    "baki": {
      "fadi": 2
    }
  }
}

Wannan hangen nesa yana nuna alaƙa tsakanin wakilai, muhalli, da manufofin tsarin.

Kammalawa

Gina tsarin software na warkar da kai tare da gine-ginen AI masu yawa da LangChain yana ba da mafita mai ƙarfi don tabbatar da aminci, haɓakawa, da juriya na tsarin software mai rikitarwa. Ta hanyar rarraba tsarin yanke shawara a tsakanin wakilai da yawa, waɗannan tsarin zasu iya koyo daga gwaninta, daidaitawa ga yanayin canzawa, da murmurewa daga kurakurai da sauri. Tare da LangChain, masu haɓakawa za su iya gina tsarin AI masu rikitarwa waɗanda ke haɗawa tare da kewayon hanyoyin bayanai, gami da bayanan bayanai, APIs, da na'urorin IoT. Ta bin lambar misali da amfani da manyan 5-10 mafi yawan umarnin da aka yi amfani da su / snippets, masu haɓakawa za su iya gina nasu tsarin software na warkarwa tare da LangChain.

Gina Tsarin Software na Warkar da Kai tare da Multi-Agent AI Architectures da LangChain