GoogleMLEngineHelper.ts

import { google } from 'googleapis';
import { auth } from 'google-auth-library'

export default class getModel {
    modelName: string;
    projectName: string;

    constructor(attrs) {
        this.modelName = attrs.modelName
        this.projectName = attrs.projectName
    }
    // our data format was an array of strings, this needs to match your model
    predict(dataToPredict:Array<string>):Promise<Array<any>> {
        return new Promise((resolve,reject)=>{
            auth.getApplicationDefault((err, authClient)=>{
                if (err) {
                  reject(err);
                }
             
                // Create authenticated ml engine client
                const ml = google.ml({
                  version: 'v1',
                  auth: authClient
                });
    
                // Prediction
                const params = {
                    auth: authClient,
                    name: `projects/${this.projectName}/models/${this.modelName}`,
                    resource: {
                        instances: typeof dataToPredict === "object" ? dataToPredict : [ dataToPredict ]
                    }
                }
           
                ml.projects.predict(params, function (err2, result){
                  if (err2){
                    reject(err2)
                  }
                  resolve(result.data.predictions)
                });
             })
        })  
    }
}