Backend.AI Client for Javascript (node.js / ES6+)

Requirements

This client SDK runs on CommonJs (with node.js) / ES6-compatible Javascript runtimes with async/await supports such as NodeJS 7+ and modern web browsers released since 2017.

This client library supports Backend.AI API v3/4/5/6.

Install

$ npm install backend.ai-client

You can also use yarn.

$ yarn install backend.ai-client

Build

Package preparation

NPM:

$ npm i

Yarn:

$ yarn install

ES6+ library

$ make es6

Node.js library

$ make node 

Usage

TypeScript:

import * as ai from 'backend.ai-client-node';

let config = ai.backend.ClientConfig.createFromEnv();
let client = new ai.backend.Client(config);

CommonJS-style:

const ai = require('backend.ai-client-node');

let config = ai.backend.ClientConfig.createFromEnv();
let client = new ai.backend.Client(config);

ES6+:

import './backend.ai-client-es6.js';

let config = new ai.backend.ClientConfig(
  '[ADD_ACCESS_KEY_HERE]',
  '[ADD_SECRET_KEY_HERE]',
  '[ENDPOINT_HERE]'
);
let client = new ai.backend.Client(
  config,
  `Backend.AI ES6 App.`,
);

When creating ClientConfig object, you can manually pass accessKey, secretKey, and optional endpoint arguments. The environment variables are:

• BACKEND_ACCESS_KEY
• BACKEND_SECRET_KEY
• BACKEND_ENDPOINT (optional, defaults to https://api.backend.ai)

All API functions return a promise that resolves into a parsed object when success according to server-provided Content-Type and rejects with an object with type and message attributes if failed.

client.createIfNotExists('python:latest', 'my-session-id')
.then(response => {
  console.log(`my session is created: ${response.kernelId}`);
}).catch(err => {
  switch (err.type) {
  case ai.backend.Client.ERR_SERVER:
    console.log(`session creation failed: ${err.message}`);
    break;
  default:
    console.log(`request/response failed: ${err.message}`);
  }
});

The result objects returned with success has different formats API by API. Please check out our official documentation.

When using backend.ai, you can use with SESSION mode or API mode.
If you want to use backend.ai with SESSION mode you need to input user_id, password, api_endpoint and SESSION to specify the mode.
If you want to use backend.ai with API mode you need to input ACCESS_KEY and SECRET_KEY instead of user_id and password. Also you don't need to input the mode because API mode is default value.

async Login() {
  config = new ai.backend.ClientConfig(
    '[ADD_USER_ID_HERE]',
    '[ADD_PASSWORD_HERE]',
    '[ENDPOINT_HERE]',
    '[CONNECTION_MODE_HERE]'
  );
  client = new ai.backend.Client(
    config,
    `Backend.AI Console.`,
  );
  let isLogon = await client.check_login();
  if (isLogon === false) {
    client.login().then(reponse => {
      if (reponse === false) {
        if (user_id != '' && password != '') {
          console.log(`Login information mismatch. Please check your login information.`);
        }
      } else if (reponse.fail_reason) {
        if (user_id != '' && password != '') {
          console.log(`Login failed: ${response.fail_reason}`);
        }
      } else {
        console.log(`Login succeeded.`);
      }
    }).catch(err => {
        console.log(`Login failed: ${err.message}`);
    });
  } else {
    console.log(`Login already succeeded.`)
  }
}

err.type is one of the following values:

• ai.backend.Client.ERR_SERVER: The server responded with failure. In this case, err.message includes HTTP status and additional error information returned by the API server.
• ai.backend.Client.ERR_RESPONSE: An error occurred while reading the response. err.message includes an exception value passed from your Javascript runtime.
• ai.backend.Client.ERR_REQUEST: An error occurred while sending the request. err.message includes an exception value passed from your Javascript runtime.
• ai.backend.Client.ERR_ABORT: An error occurred while request aborted by user.
• ai.backend.Client.ERR_TIMEOUT: An error occurred while no response returned during the timeout period.
• ai.backend.Client.ERR_UNKNOWN: An error occurred while unknown error occurs. In this case, err.message includes HTTP status and additional error information returned by the API server.