Page Contents

A Repository represents a specialized Service interface that provides strong-typed data access (for example, CRUD) operations of a domain model against the underlying database or service.

Repository diagram

A Repository can be defined and implemented by application developers. LoopBack ships a few predefined Repository interfaces for typical CRUD and KV operations. These Repository implementations leverage Model definition and DataSource configuration to fulfill the logic for data access.

interface Repository<T extends Model> {}

interface CustomerRepository extends Repository<Customer> {
  find(filter?: Filter<Customer>, options?: Options): Promise<Customer[]>;
  findByEmail(email: string): Promise<Customer>;
  // ...

See more examples at:


Juggler support has been enabled in loopback-next and can be imported from the @loopback/repository package. In order to do this, save @loopback/repository as a dependency in your application.

You can then install your favorite connector by saving it as part of your application dependencies.

Common Tasks

Repository Mixin

@loopback/repository provides a mixin for your Application that enables convenience methods that automatically bind repository classes for you. Repositories declared by components are also bound automatically.

Repositories are bound to repositories.${ClassName}. See example below for usage.

import {Application} from '@loopback/core';
import {RepositoryMixin} from '@loopback/repository';
import {AccountRepository, CategoryRepository} from './repositories';

// Using the Mixin
class MyApplication extends RepositoryMixin(Application) {}

const app = new MyApplication();
// AccountRepository will be bound to key `repositories.AccountRepository`
// CategoryRepository will be bound to key `repositories.CategoryRepository`

Configure datasources

DataSource is a named configuration of a connector. The configuration properties vary by connectors. For example, a datasource for MySQL needs to set the connector property to loopback-connector-mysql with settings as follows:

  "host": "localhost",
  "port": 3306,
  "user": "my-user",
  "password": "my-password",
  "database": "demo"

Connector is a provider that implements data access or api calls with a specific backend system, such as a database, a REST service, a SOAP Web Service, or a gRPC micro-service. It abstracts such interactions as a list of operations in the form of Node.js methods.

Typically, a connector translates LoopBack query and mutation requests into native api calls supported by the underlying Node.js driver for the given backend. For example, a connector for MySQL will map create method to SQL INSERT statement, which can be executed through MySQL driver for Node.js.

When a DataSource is instantiated, the configuration properties will be used to initialize the connector to connect to the backend system. You can define a DataSource using Juggler in your LoopBack 4 app as follows:


import {juggler} from '@loopback/repository';

// this is just an example, 'test' database doesn't actually exist
export const db = new juggler.DataSource({
  connector: 'mysql',
  host: 'localhost',
  port: 3306,
  database: 'test',
  password: 'pass',
  user: 'root',

Define models

Models are defined as regular JavaScript classes. If you want your model to be persisted in a database, your model must have an id property and inherit from Entity base class.

TypeScript version:

import {Entity, model, property} from '@loopback/repository';

export class Account extends Entity {
  @property({id: true})
  id: number;

  @property({required: true})
  name: string;

JavaScript version:

import {Entity, ModelDefinition} from '@loopback/repository';

export class Account extends Entity {}

Account.definition = new ModelDefinition({
  name: 'Account',
  properties: {
    id: {type: 'number', id: true},
    name: {type: 'string', required: true},

Define repositories

Use DefaultCrudRepository class to create a repository leveraging juggler bridge and binding your Entity-based class with a datasource you have configured earlier. It’s recommended that you use Dependency Injection to retrieve your datasource.

TypeScript version:

import {DefaultCrudRepository, juggler} from '@loopback/repository';
import {Account, AccountRelations} from '../models';
import {DbDataSource} from '../datasources';
import {inject} from '@loopback/core';

export class AccountRepository extends DefaultCrudRepository<
> {
  constructor(@inject('datasources.db') dataSource: DbDataSource) {
    super(Account, dataSource);

JavaScript version:

import {DefaultCrudRepository} from '@loopback/repository';
import {Account} from '../models/account.model';
import {db} from '../datasources/db.datasource';

export class AccountRepository extends DefaultCrudRepository {
  constructor() {
    super(Account, db);

Controller Configuration

Once your DataSource is defined for your repository, all the CRUD methods you call in your repository will use the Juggler and your connector’s methods unless you overwrite them. In your controller, you will need to define a repository property and create a new instance of the repository you configured your DataSource for in the constructor of your controller class as follows:

export class AccountController {
    @repository(AccountRepository) public repository: AccountRepository,
  ) {}

Defining CRUD methods for your application

When you want to define new CRUD methods for your application, you will need to modify the API Definitions and their corresponding methods in your controller. Here are examples of some basic CRUD methods:

  1. Create API Definition:

      "/accounts/create": {
        "post": {
          "x-operation-name": "createAccount",
          "requestBody": {
            "description": "The account instance to create.",
            "required": true,
            "content": {
              "application/json": {
                "schema": {
                  "type": "object"
          "responses": {
            "200": {
              "description": "Account instance created",
              "content": {
                "application/json": {
                  "schema": {
                    "$ref": "#/components/schemas/Account"

    Create Controller method:

    async createAccount(accountInstance: Account) {
      return this.repository.create(accountInstance);
  2. Find API Definition:

      "/accounts": {
        "get": {
          "x-operation-name": "getAccount",
          "responses": {
            "200": {
              "description": "List of accounts",
              "content": {
                "application/json": {
                  "schema": {
                    "type": "array",
                    "items": {
                      "$ref": "#/components/schemas/Account"

    Find Controller method:

    async getAccount() {
      return this.repository.find();

Don’t forget to register the complete version of your OpenAPI spec through app.api().

Please See Testing Your Application section in order to set up and write unit, acceptance, and integration tests for your application.

Access KeyValue Stores

We can now access key-value stores such as Redis using the KeyValueRepository.

Define a KeyValue Datasource

We first need to define a datasource to configure the key-value store. The easiest option is to to generate the datasource automatically using lb4 datasource command and selecting Redis key-value connector (supported by StrongLoop).

The generated datasource file has two parts:

  1. A config object with the configuration options provided via CLI.
  2. A DataSource class accepting the configuration via Dependency Injection, with a fall-back to the default config.
import {inject} from '@loopback/core';
import {juggler, AnyObject} from '@loopback/repository';

const config = {
  name: 'redis',
  connector: 'kv-redis',
  host: '',
  port: 6379,
  password: '',
  db: 0,

export class RedisDataSource extends juggler.DataSource {
  static dataSourceName = 'redis';
  static readonly defaultConfig = config;

    @inject('datasources.config.redis', {optional: true})
    dsConfig: AnyObject = config,
  ) {

Define a KeyValueRepository

The KeyValueRepository binds a model such as ShoppingCart to the RedisDataSource. The base DefaultKeyValueRepository class provides an implementation based on loopback-datasource-juggler.

import {DefaultKeyValueRepository} from '@loopback/repository';
import {ShoppingCart} from '../models/shopping-cart.model';
import {RedisDataSource} from '../datasources/redis.datasource';
import {inject} from '@loopback/core';

export class ShoppingCartRepository extends DefaultKeyValueRepository<ShoppingCart> {
  constructor(@inject('datasources.redis') ds: RedisDataSource) {
    super(ShoppingCart, ds);

Perform Key Value Operations

The KeyValueRepository provides a set of key based operations, such as set, get, delete, expire, ttl, and keys. See KeyValueRepository for a complete list.

// Please note the ShoppingCartRepository can be instantiated using Dependency
// Injection
const repo: ShoppingCartRepository =
  new ShoppingCartRepository(new RedisDataSource());
const cart1: ShoppingCart = givenShoppingCart1();
const cart2: ShoppingCart = givenShoppingCart2();

async function testKV() {
  // Store carts using userId as the key
  await repo.set(cart1.userId, cart1);
  await repo.set(cart2.userId, cart2);

  // Retrieve a cart by its key
  const result = await repo.get(cart1.userId);


Persist Data without Juggler [Using MySQL database]

LoopBack 4 gives you the flexibility to create your own custom Datasources which utilize your own custom connector for your favorite back end database. You can then fine tune your CRUD methods to your liking.

Example Application

You can look at the account-without-juggler application as an example.

  1. Implement the CrudConnector interface from @loopback/repository package. Here is one way to do it

  2. Implement the DataSource interface from @loopback/repository. To implement the DataSource interface, you must give it a name, supply your custom connector class created in the previous step, and instantiate it:

    export class MySQLDs implements DataSource {
      name: 'mysqlDs';
      connector: MySqlConn;
      settings: Object;
      constructor() {
        this.settings = require('./mysql.json'); // connection configuration
        this.connector = new MySqlConn(this.settings);
  3. Extend CrudRepositoryImpl class from @loopback/repository and supply your custom DataSource and model to it:

    import {CrudRepositoryImpl} from '@loopback/repository';
    import {MySQLDs} from './datasources/mysqlds.datasource';
    import {Account} from './models/account.model';
    export class NewRepository extends CrudRepositoryImpl<Account, string> {
      constructor() {
        const ds = new MySQLDs();
        super(ds, Account);

You can override the functions it provides, which ultimately call on your connector’s implementation of them, or write new ones.

Configure Controller

The next step is to wire your new DataSource to your controller. This step is essentially the same as above, but can also be done as follows using Dependency Injection:

  1. Bind instance of your repository to a certain key in your application class

    class AccountMicroservice extends Application {
      private _startTime: Date;
      constructor() {
        const app = this;
  2. Inject the bound instance into the repository property of your controller. inject can be imported from @loopback/core.

    export class AccountController {
      private repository: NewRepository;

Example custom connector CRUD methods

Here is an example of a find function which uses the node-js mysql driver to retrieve all the rows that match a particular filter for a model instance.

public find(
  modelClass: Class<Account>,
  filter: Filter<Account>,
  options: Options
): Promise<Account[]> {
  let self = this;
  let sqlStmt = "SELECT * FROM " +;
  if (filter.where) {
    let sql = "?? = ?";
    let formattedSql = "";
    for (var key in filter.where) {
      formattedSql = mysql.format(sql, [key, filter.where[key]]);
    sqlStmt += " WHERE " + formattedSql;
  debug("Find ", sqlStmt);
  return new Promise<Account[]>(function(resolve, reject) {
    self.connection.query(sqlStmt, function(err: any, results: Account[]) {
      if (err !== null) return reject(err);

Example Application

You can look at the account application as an example.