E hana i kahi luna LangChain | B4N1

E hana i kahi luna LangChain

8 months ago · b4n1 Tiempo de lectura: 7 min

Hoʻohana ʻana i nā Agents AI no ka hoʻouna ʻana i nā lako polokalamu: Hoʻohui ʻana i ka LangChain a me DSPy i nā Pipelines DevOps

Hoʻolauna

Ke ulu wikiwiki nei ka ʻāina hoʻomohala polokalamu, me ka hoʻonui nui ʻana o Artificial Intelligence (AI) a me Machine Learning (ML) i nā ʻano like ʻole o ke ola ola hoʻomohala. ʻO kekahi o nā wahi ʻoi loa o ka noi no AI ʻo ia i ka hoʻokaʻawale ʻana i nā kaʻina hana lako polokalamu, me ka telemetry a me ka hoʻohui mau ʻana / hoʻomau mau ʻia (CI/CD). Ma kēia ʻatikala, e ʻimi mākou i ke kuleana o nā mea hana AI i ka hoʻokaʻawale ʻana i nā kaʻina hana lako polokalamu a komo i ka hoʻokō pono o nā mea hana orchestration e like me LangChain a me DSPy.

ʻO ke kuleana o nā Agents AI i ka hoʻouna lako polokalamu

ʻO nā ʻelele AI nā polokalamu lako polokalamu hiki ke hana i nā hana me ke kūʻokoʻa, e hoʻoholo ana i nā ʻikepili a me nā lula. I ka pōʻaiapili o ka hāʻawi ʻana i nā polokalamu, hiki ke hoʻohana ʻia nā mea hana AI e hoʻokaʻawale i nā kaʻina hana like ʻole, e like me:

  1. Telemetry: Hiki i nā ʻelele AI ke hōʻiliʻili a kālailai i ka ʻikepili mai nā kumu like ʻole, e like me nā moʻolelo noiʻi, nā ana hana, a me nā manaʻo manaʻo o ka mea hoʻohana, e hāʻawi i nā ʻike i ka hana a me ka hana o ka noi.
  2. Hoʻohui Hoʻomau: Hiki i nā ʻelele AI ke hoʻokaʻawale i ke kūkulu ʻana, hoʻāʻo, a me ka waiho ʻana o nā hoʻololi code, me ka hōʻoia ʻana i ke ʻano o ka noi ma kahi kūlana hiki ke hoʻoili ʻia.
  3. Hoʻohana mau ʻia: Hiki i nā ʻelele AI ke hoʻokaʻawale i nā hoʻololi i ka hana ʻana, e hōʻemi ana i ka manawa a me ka hoʻoikaika ʻana e loaʻa i nā hiʻohiʻona hou a me nā hoʻoponopono i nā mea hoʻohana.
  4. Monitoring and Alerting: Hiki i nā ʻelele AI ke nānā i ka hana o ka polokalamu a hoʻomaopopo i nā mea hoʻomohala a me nā mea hoʻohana i nā pilikia ma mua o ka lilo ʻana i mea koʻikoʻi.

LangChain: He Kūlana no ke kūkulu ʻana i nā ʻĀpana AI

ʻO LangChain kahi hoʻolālā Python no ke kūkulu ʻana i nā mea hana AI hiki ke launa pū me nā kumu ʻikepili like ʻole a hana i nā hana me ka autonomously. Hāʻawi ʻo LangChain i kahi API maʻalahi a intuitive no nā mea hana hale hiki ke:

  1. E launa pū me nā kumu ʻikepili: Hāʻawi ʻo LangChain i kahi ʻano o nā kumu ʻikepili, me nā ʻikepili, nā API, a me nā ʻōnaehana faila, hiki i nā ʻelele ke launa pū me.
  2. Hana i nā hana: Hāʻawi ʻo LangChain i nā ʻano hana e hiki ai i nā ʻelele ke hana, me ka hoʻoponopono kikokikona, ʻike kiʻi, a me ka hana ʻōlelo kūlohelohe.
  3. E aʻo mai ka ʻikepili: Hāʻawi ʻo LangChain i kahi ʻano algorithm aʻo mīkini e hiki ai i nā ʻelele ke hoʻohana e aʻo mai ka ʻikepili.

DSPy: He Kiʻi no ke kūkulu ʻana i nā mea hoʻolaha AI i hoʻolaha ʻia

ʻO DSPy kahi hoʻolālā Python no ke kūkulu ʻana i nā mea hoʻolaha AI hiki ke hoʻohālikelike i nā pono o nā noi nui. Hāʻawi ʻo DSPy i kahi ʻano o nā hiʻohiʻona, me:

  1. Ka helu helu hoʻolaha ʻia: ʻAe ʻo DSPy i nā ʻelele e holo ma luna o nā mīkini he nui, i kūpono ia no nā noi nui.
  2. Data parallelism: ʻAe ʻo DSPy i nā ʻelele e hoʻoponopono i ka ʻikepili ma ke ʻano like, no laila kūpono ia no nā noi e koi ana i ka hoʻopili helu kiʻekiʻe.
  3. Scalability: ʻAe ʻo DSPy i nā ʻelele e hoʻokō i nā pono o nā noi nui.

Hoʻohui i ka LangChain a me DSPy i nā DevOps Pipelines

No ka hoʻohui ʻana iā LangChain a me DSPy i nā pipelines DevOps, hiki iā mākou ke hoʻohana i kēia mau ʻanuʻu:

  1. E hoʻouka i ka LangChain a me DSPy: E hoʻouka i ka LangChain a me DSPy me ka hoʻohana ʻana i ka pip.
pip hoʻokomo i ka langchain dspy
  1. E hana i kahi luna LangChain: E hana i kahi luna LangChain hiki ke launa pū me nā kumu ʻikepili a hana i nā hana.
lawe mai i ka lanchain


luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)
  1. E hana i kahi ʻelele DSPy: E hana i kahi ʻelele DSPy hiki ke holo ma nā mīkini he nui a hana i ka ʻikepili i ka like.
lawe mai i ka dspy

# E hana i kahi luna DSPy
luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)
  1. Hoʻohui i nā mea hana LangChain a me DSPy: Hoʻohui i nā mea hana LangChain a me DSPy i kahi pipeline DevOps me ka hoʻohana ʻana i kahi mea hana e like me Jenkins a i ʻole GitLab CI/CD.
lawe mai i ka lanchain
lawe mai i ka dspy

# E hana i kahi ʻoihana LangChain
langchain_agent = langchain.Agent(
    inoa = "my_langchain_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)

# E hana i kahi luna DSPy
dspy_agent = dspy.Agent(
    inoa = "my_dspya_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)

# Hoʻohui i nā mea hana i kahi pipeline DevOps
pipeline = [
    langchain_agent,
    dspy_agent
]

** 5-10 kiʻekiʻe loa i hoʻohana ʻia i nā kauoha/nā ʻāpana**

Eia nā 5-10 kiʻekiʻe loa i hoʻohana ʻia i nā kauoha / snippets no LangChain a me DSPy:

  1. LangChain: E hana i kahi luna LangChain
lawe mai i ka lanchain

luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)
  1. DSPy: Hana i kahi luna DSPy
lawe mai i ka dspy

luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)
  1. LangChain: Hoʻopili me kahi kumu ʻikepili
lawe mai i ka lanchain

luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)

ʻikepili = agent.data_sources[0].get_data()
  1. DSPy: Hoʻopili me kahi kumu ʻikepili
lawe mai i ka dspy

luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)

ʻikepili = agent.data_sources[0].get_data()
  1. LangChain: Hana i kahi hana
lawe mai i ka lanchain

luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)

hopena = agent.tasks[0].perform_task(data)
  1. DSPy: Hana i kahi hana
lawe mai i ka dspy

luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)

hopena = agent.tasks[0].perform_task(data)
  1. LangChain: E aʻo mai ka ʻikepili
lawe mai i ka lanchain

luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)

agent.learn_from_data(data)
  1. DSPy: E aʻo mai ka ʻikepili
lawe mai i ka dspy

luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)

agent.learn_from_data(data)
  1. LangChain: Hoʻohui me kahi paipu DevOps
lawe mai i ka lanchain

luna = langchain.Agent(
    inoa = "my_agent",
    data_sources=[langchain.DataSource("database", "my_database")],
    hana=[langchain.Task("text_processing", "my_text_processing")]
)

pipeline = [
    luna
]
  1. DSPy: Hoʻohui me kahi paipu DevOps
lawe mai i ka dspy

luna = dspy.Agent(
    inoa = "my_agent",
    data_sources=[dspy.DataSource("database", "my_database")],
    hana=[dspy.Task("text_processing", "my_text_processing")]
)

pipeline = [
    luna
]

Hōʻikeʻike

Eia kahi hiʻohiʻona o nā mea hana LangChain a me DSPy i kahi pipeline DevOps:

{
  "type": "pā",
  "ʻikepili": {
    "labels": ["LangChain", "DSPy"],
    "mau ʻikepili": [
      {
        "label": "Agent",
        "ʻikepili": [10, 20],
        "backgroundColor": ["rgba(255, 99, 132, 0.2)", "rgba(54, 162, 235, 0.2)"],
        "BorderColor": ["rgba(255, 99, 132, 1)", "rgba(54, 162, 235, 1)"],
        "BorderWidth": 1
      }
    ]
  },
  "koho": {
    "poʻo inoa": {
      "hōʻike": ʻoiaʻiʻo,
      "text": "LangChain a me DSPy Agents ma DevOps Pipeline"
    },
    "nā unahi": {
      "yAxes": [
        {
          "kuʻi": {
            "beginAtZero": ʻoiaʻiʻo
          }
        }
      ]
    }
  }
}

Hōʻike kēia hiʻohiʻona i nā mea hana LangChain a me DSPy i kahi pipeline DevOps, me ka luna LangChain he 10 mau hana a me ka luna DSPy he 20 mau hana.

Hoʻoholo

Ma kēia ʻatikala, ua ʻimi mākou i ke kuleana o nā ʻelele AI i ka hoʻokaʻawale ʻana i nā kaʻina hana lako polokalamu a komo i loko o ka hoʻokō pono o nā mea hana orchestration e like me LangChain a me DSPy. Ua ʻike mākou pehea e hoʻohana ʻia ai ʻo LangChain a me DSPy e hana i nā mea hana AI e hiki ke launa pū me nā kumu ʻikepili, hana i nā hana, a aʻo mai ka ʻikepili. Ua ʻike pū mākou pehea e hiki ai ke hoʻohui ʻia ʻo LangChain a me DSPy i nā pipelines DevOps me ka hoʻohana ʻana i nā mea hana e like me Jenkins a i ʻole GitLab CI/CD. Me ka hoʻonui nui ʻana o AI a me ML i ka hoʻomohala ʻana i nā polokalamu, ua lilo ʻo LangChain a me DSPy i nā mea hana koʻikoʻi no nā mea hoʻomohala a me nā mea hoʻohana e ʻimi nei e hoʻokaʻawale i nā kaʻina hana hoʻopuka.

¿Tienes preguntas?

Hablemos

Contáctame en LinkedIn