Javascript client for the Elastic App Search Api

Downloads in past


58108.7.02 days ago4 years agoMinified + gzip package size for @elastic/app-search-javascript in KB


Elastic App Search Logo

CircleCI buidl

A first-party JavaScript client for building excellent, relevant search experiences with Elastic App Search.


Getting started 🐣

Install from a CDN

The easiest way to install this client is to simply include the built distribution from the jsDelivr CDN.
<script src="https://cdn.jsdelivr.net/npm/@elastic/app-search-javascript@8.7.0/dist/elastic_app_search.umd.js"></script>

This will make the client available globally at:

Install from NPM

This package can also be installed with npm or yarn.
npm install --save @elastic/app-search-javascript

The client could then be included into your project like follows:
// CommonJS
var ElasticAppSearch = require("@elastic/app-search-javascript");

// ES
import * as ElasticAppSearch from "@elastic/app-search-javascript";


This client is versioned and released alongside App Search.
To guarantee compatibility, use the most recent version of this library within the major version of the corresponding App Search implementation.
For example, for App Search 7.3, use 7.3 of this library or above, but not 8.0.
If you are using the SaaS version available on swiftype.com of App Search, you should use the version 7.5.x of the client.

Browser support

The client is compatible with all modern browsers.
Note that this library depends on the Fetch API.
This is not supported by Internet Explorer. If you need backwards compatibility for Internet Explorer, you'll need to polyfill the Fetch API with something like https://github.com/github/fetch.


Setup: Configuring the client and authentication

Using this client assumes that you have already an instance of Elastic App Search up and running.
The client is configured using the searchKey, endpointBase, and engineName parameters.
var client = ElasticAppSearch.createClient({
  searchKey: "search-mu75psc5egt9ppzuycnc2mc3",
  endpointBase: "",
  engineName: "favorite-videos"

\ Please note that you should only ever use a Public Search Key within Javascript code on the browser. By default, your account should have a Key prefixed with search- that is read-only. More information can be found in the documentation.

Swiftype.com App Search users:

When using the SaaS version available on swiftype.com of App Search, you can configure the client using your hostIdentifier instead of the endpointBase parameter. The hostIdentifier can be found within the Credentials menu.
var client = ElasticAppSearch.createClient({
  hostIdentifier: "host-c5s2mj",
  searchKey: "search-mu75psc5egt9ppzuycnc2mc3",
  engineName: "favorite-videos"

List of configuration options:

| Option | Required | Description | | ----------------- | -------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | hostIdentifier | No | Your Host Identifier, should start with host-. Required unless explicitly setting endpointBase | | searchKey | No | Your Public Search Key. It should start with search-.

NOTE: This is not technically required, but in 99% of cases it should be provided. There is a small edge case for not providing this, mainly useful for internal App Search usage, where this can be ommited in order to leverage App Search's session based authentication. | | engineName | Yes | | | endpointBase | No | Overrides the base of the App Search API endpoint completely. Useful when proxying the App Search API, developing against a local server, or a Self-Managed or Cloud Deployment. Ex. "http://localhost:3002" | | cacheResponses | No | Whether or not API responses should be cached. Default: true. | | additionalHeaders | No | An Object with keys and values that will be converted to header names and values on all API requests |

API Methods

This client is a thin interface to the Elastic App Search API. Additional details for requests and responses can be found in the documentation.


For the query term lion, a search call is constructed as follows:
var options = {
  search_fields: { title: {} },
  result_fields: { id: { raw: {} }, title: { raw: {} } }

  .search("lion", options)
  .then(resultList => {
    resultList.results.forEach(result => {
      console.log(`id: ${result.getRaw("id")} raw: ${result.getRaw("title")}`);
  .catch(error => {
    console.log(`error: ${error}`);

Note that options supports all options listed here: https://swiftype.com/documentation/app-search/guides/search.
In addition to the supported options above, we also support the following fields:
| Name | Type | Description | | ------------------------------ | ------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | disjunctiveFacets | ArrayString | An array of field names. Every field listed here must also be provided as a facet in the facet field. It denotes that a facet should be considered disjunctive. When returning counts for disjunctive facets, the counts will be returned as if no filter is applied on this field, even if one is applied. | | disjunctiveFacetsAnalyticsTags | ArrayString | Used in conjunction with the disjunctiveFacets parameter. Queries will be tagged with "Facet-Only" in the Analytics Dashboard unless specified here. |
The search method returns the response wrapped in a ResultList type:
ResultList {
  rawResults: [], // List of raw `results` from JSON response
  rawInfo: { // Object wrapping the raw `meta` property from JSON response
    meta: {}
  results: [ResultItem], // List of `results` wrapped in `ResultItem` type
  info: { // Currently the same as `rawInfo`
    meta: {}

ResultItem {
  getRaw(fieldName), // Returns the HTML-unsafe raw value for a field, if it exists
  getSnippet(fieldName) // Returns the HTML-safe snippet value for a field, if it exists

Query Suggestion

var options = {
  size: 3,
  types: {
    documents: {
      fields: ["name"]

  .querySuggestion("cat", options)
  .then(response => {
    response.results.documents.forEach(document => {
  .catch(error => {
    console.log(`error: ${error}`);

Multi Search

It is possible to run multiple queries at once using the multiSearch method.
To search for the term lion and tiger, a search call is constructed as follows:
var options = {
  search_fields: { name: {} },
  result_fields: { id: { raw: {} }, title: { raw: {} } }

  .multiSearch([{ query: "node", options }, { query: "java", options }])
  .then(allResults => {
    allResults.forEach(resultList => {
      resultList.results.forEach(result => {
          `id: ${result.getRaw("id")} raw: ${result.getRaw("title")}`
  .catch(error => {
    console.log(`error: ${error}`);

Clickthrough Tracking

    query: "lion",
    documentId: "1234567",
    requestId: "8b55561954484f13d872728f849ffd22",
    tags: ["Animal"]
  .catch(error => {
    console.log(`error: ${error}`);

Clickthroughs can be tracked by binding client.click calls to click events on individual search result links.
The following example shows how this can be implemented declaratively by annotating links with class and data attributes.
document.addEventListener("click", function(e) {
  const el = e.target;
  if (!el.classList.contains("track-click")) return;

    query: el.getAttribute("data-query"),
    documentId: el.getAttribute("data-document-id"),
    requestId: el.getAttribute("data-request-id"),
    tags: [el.getAttribute("data-tag")]

  Item 1

Running tests

The specs in this project use node-replay to capture responses.
The responses are then checked against Jest snapshots.
To capture new responses and update snapshots, run:
nvm use
REPLAY=record npm run test -u

To run tests:
nvm use
npm run test



You will probably want to install a node version manager, like nvm.
We depend upon the version of node defined in .nvmrc.
To install and use the correct node version with nvm:
nvm install

Dev Server

Install dependencies:
nvm use
npm install

Build artifacts in dist directory:
# This will create files in the `dist` directory
npm run build

Add an index.html file to your dist directory
    <script src="elastic_app_search.umd.js"></script>


Run dev server:
# This will serve files in the `dist` directory
npm run dev

Navigate to and execute JavaScript commands through the browser Dev Console.


nvm use
npm run build


nvm use
npm run publish


What if I need write operations?

App Search has a first-party Node.js client which supports write operations like indexing.

Where do I report issues with the client?

If something is not working as expected, please open an issue.

Where can I learn more about App Search?

Your best bet is to read the documentation.

Where else can I go to get help?

You can checkout the Elastic App Search community discuss forums.

Contribute 🚀

We welcome contributors to the project. Before you begin, a couple notes...

License 📗

Apache 2.0 © Elastic
Thank you to all the contributors!