Kushandisa Yakachengeteka uye Scalable AI-Driven Software Architecture ine Orchestration Zvishandiso | B4N1

Kushandisa Yakachengeteka uye Scalable AI-Driven Software Architecture ine Orchestration Zvishandiso

3 months ago · b4n1 Tiempo de lectura: 4 min

Nhanganyaya

Kubatanidzwa kweArtificial Intelligence (AI) mumapurogiramu esoftware kwashandura nzira yatinosvika nayo kugadzirisa matambudziko, kuita sarudzo, uye otomatiki. Nekudaro, sezvo maapplication anofambiswa neAI achiwedzera kuoma, kudiwa kwekuchengetedza uye scalable zvivakwa zvave kunetsa. Muchinyorwa chino, tichaongorora mharadzano yeAI, software engineering, uye DevOps, nekusimbisa kwakananga pakushandisa maturusi e orchestration seDSPy neLangChain kubatanidza vamiririri veAI mumapaipi eDevOps.

Matambudziko eAI-Inotyairwa Software Architecture

AI-inotyairwa software maapplication anounza akasiyana matambudziko kune software mainjiniya uye DevOps zvikwata. Mamwe ematambudziko makuru anosanganisira:

  • ** Scalability **: AI modhi inoda akakosha computational zviwanikwa, izvo zvinogona kutungamira kune scalability nyaya munzvimbo dzekugadzira.
  • ** Chengetedzo **: AI modhi inogona kuve panjodzi yekurwiswa, senge data chepfu uye modhi inversion kurwiswa.
  • ** Kubatanidzwa **: AI mhando dzinoda kubatanidzwa nehurongwa huripo hwepurogiramu, iyo inogona kuva basa rakaoma.
  • ** Kutarisisa uye Kugadzirisa **: AI modhi inoda kuenderera kutariswa uye kugadzirisa kuti ive nechokwadi chekuti inoramba iri yakarurama uye yakavimbika.

Orchestration Zvishandiso: DSPy uye LangChain

Maturusi eOrchestration akaita seDSPy neLangChain anopa mhinduro kumatambudziko aya nekugonesa kubatanidzwa kweAI agents mumapaipi eDevOps. Zvishandiso izvi zvinopa chimiro chekuvaka, kutumira, uye kutonga maAI modhi, kuve nechokwadi chekuchengetedza uye scalable zvivakwa zveAI-inotyairwa zvikumbiro.

DSPy

DSPy ndeye yakavhurika-sosi orchestration chishandiso chakavakirwa pamusoro pePython. Inopa iri nyore uye intuitive API yekuvaka, kutumira, uye kutonga AI modhi. DSPy inotsigira zvakasiyana-siyana zveAI masimusi, kusanganisira TensorFlow, PyTorch, uye scikit-dzidza.

LangChain

LangChain imwe yakavhurika-sosi orchestration chishandiso chinopa chimiro chekuvaka, kuendesa, uye kutonga maAI modhi. LangChain yakavakirwa pamusoro peRust uye inopa yakakwirira-inoshanda API yekubatanidza maAI maajenti mumapaipi eDevOps.

Kushandisa Yakachengeteka uye Scalable AI-Inotyairwa Software Architecture ine DSPy

Muchikamu chino, tichaongorora muenzaniso wakapfava wekushandisa yakachengeteka uye scalable AI-inotyairwa software architecture neDSPy.

Muenzaniso: Kuvaka Nyore AI Model neDSPy

Ngatitarisei muenzaniso wakapfava wekuvaka muchina wekudzidza modhi uchishandisa DSPy. Tichashandisa raibhurari ye-scikit-kudzidza kuvaka mutsara wekudzoreredza modhi.

import dSpy se dsp
kubva sklearn.linear_model pinza LinearRegression
kubva sklearn.datasets import load_diabetes

# Rodha dataset yechirwere cheshuga
dataset = load_diabetes()

# Kamura dhatabheti mukudzidziswa uye seti yekuyedza
X_train, X_test, y_train, y_test = dsp.split_dataset(dataset.data, dataset.target, test_size=0.2)

# Vaka mutsara wekudzoreredza modhi
modhi = LinearRegression()

# Dzidzisa modhi pane data rekudzidzisa
model.fit(X_train, y_train)

# Ongorora modhi pane yekuyedza data
kururama = model.score(X_test, y_test)
dhinda(f"Kurongeka kwemuenzaniso: {kururama:.2f}")

Iyi kodhi yekodhi inoratidza nzira yekuvaka yakapusa muchina kudzidza modhi uchishandisa DSPy. Isu tinorodha dataset yechirwere cheshuga, tinoipatsanura kuita kudzidziswa uye seti yekuyedza, kuvaka mutsara wekudzoreredza modhi, tinoidzidzisa padhata rekudzidziswa, uye kuongorora huchokwadi hwayo pane data rekuyedza.

Kushandisa Yakachengeteka uye Scalable AI-Inotyairwa Software Architecture neLangChain

Muchikamu chino, tichaongorora muenzaniso wakapfava wekushandisa yakachengeteka uye scalable AI-inotyairwa software architecture neLangChain.

Muenzaniso: Kuvaka Nyore AI Model neLangChain

Ngatitarisei muenzaniso wakapfava wekuvaka muchina wekudzidza modhi uchishandisa LangChain. Isu tichashandisa TensorFlow raibhurari kuvaka neural network modhi.

shandisa langchain selc;
shandisa tensorflow::prelude::*;

// Tsanangura iri nyore neural network modhi
fn modhi (inopinza: &Tensor) -> Tensor {
    regai kubuda = input.matmul(&Tensor::new(1, 1));
    output
}

// Gadzira LangChain mumiririri
regai mumiririri = lc::Agent::new(modhi);

// Tsanangura dataset
rega dataset = lc::Dataset::new(vec![
    lc::Muenzaniso::itsva(vec![1.0, 2.0], vec![3.0]),
    lc::Muenzaniso::zvitsva(vec![4.0, 5.0], vec![6.0]),
]);

// Dzidzisa modhi pane dataset
agent.train(&dataset);

// Ongorora modhi pane dataset
regai kururama = agent.evaluate(&dataset);
println!("Kururama kwemuenzaniso: {}", kunyatsoita);

Iyi kodhi yekodhi inoratidza nzira yekuvaka yakapusa muchina kudzidza modhi uchishandisa LangChain. Isu tinotsanangura yakareruka neural network modhi, kugadzira LangChain agent, kutsanangura dhatabheti, kudzidzisa modhi pane dataset, uye kuongorora kwayo kurongeka padhata.

Yepamusoro 5-10 Yakanyanya Kushandiswa Mirairo / Zvimedu

Heano epamusoro 5-10 anonyanya kushandiswa mirairo/zvimedu zveDSPy uye LangChain:

DSPy

  • dsp.split_dataset(data, target, test_size=0.2): Inotsemura dhatabheti kuita dzidziso nemaseti ekuedzwa.
  • dsp.build_model(model_name, model_params): Inovaka modhi yemuchina wekudzidza vachishandisa zita remuenzanisi uye maparameter akataurwa.
  • dsp.train_model(modhi, data): Inodzidzisa modhi yemuchina yekudzidza pane data rataurwa.
  • dsp.evaluate_model(modhi, data): Inoongorora mashandiro emuchina wekudzidza modhi pane data rakatarwa.
  • dsp.deploy_model(modhi, server_url): Inotumira modhi yekudzidza muchina kune yakataurwa sevha URL.

LangChain

  • lc::Agent::itsva(modhi): Inogadzira mumiririri weLangChain vachishandisa muenzaniso wakatarwa.
  • lc::Dataset::mitsva(mienzaniso): Inogadzira dhatabheti reLangChain vachishandisa mienzaniso yakataurwa.
  • lc::chitima(mumiririri, dataset): Inodzidzisa mumiririri weLangChain pane yakataurwa dataset.
  • lc::evaluate(agent, dataset): Inoongorora mashandiro emumiririri weLangChain pane dataset yakataurwa.
  • lc::deploy(agent, server_url): Inotumira mumiririri weLangChain kune yakataurwa sevha URL.

Kuona

Heino ratidziro yakapusa yeAI-inotyairwa software architecture uchishandisa Chart.js:

const ctx = document.getElementById('myChart').getContext('2d');
const chati = Chati itsva (ctx, {
    mhando: 'bar',
    data: {
        zvinyorwa: ['Nguva Yekudzidzisa', 'Inference Time', 'Model Accuracy'],
        dataset: [{
            zita: 'DSPy',
            data: [10, 5, 90],
            backgroundRuvara: 'rgba(255, 99, 132, 0.2)',
            borderColor: 'rgba(255, 99, 132, 1)',
            borderWidth: 1
        }, {
            zita: 'LangChain',
            data: [8, 6, 92],
            backgroundRuvara: 'rgba(54, 162, 235, 0.2)',
            borderColor: 'rgba(54, 162, 235, 1)',
            borderWidth: 1
        }]
    },
    sarudzo: {
        zviyero: {
            y:{
                startAtZero: chokwadi
            }
        }
    }
});

Iyi kodhi dhizaini inogadzira yakapfava bar chart vachishandisa Chart.js kuona mashandiro eDSPy neLangChain.

Mhedziso

Muchinyorwa chino, takaongorora mharadzano yeAI, software engineering, uye DevOps, tiine kukoshesa kwekushandisa maturusi e orchestration seDSPy neLangChain kubatanidza vamiririri veAI mumapaipi eDevOps. Takaratidza maitiro ekushandisa yakachengeteka uye scalable AI-inotyairwa software architecture tichishandisa DSPy uye LangChain, uye takapa muenzaniso wakapfava wekuvaka muchina wokudzidza modhi uchishandisa mudziyo wega wega. Isu takapawo yepamusoro 5-10 inonyanya kushandiswa mirairo / zvimedu kune yega yega chishandiso uye nyore kuona kweAI-inotyairwa software architecture uchishandisa Chart.js.

¿Tienes preguntas?

Hablemos

Contáctame en LinkedIn