Deploying Amazon Kendra - Amazon Kendra
OverviewPrerequisitesSetting up the exampleMain search pageSearch componentResults componentFacets componentPagination component

Deploying Amazon Kendra

When it comes time to deploy Amazon Kendra search to your website, we provide source code that you can use with React to get a head start on your application. The source code is provided with no charge under a modified MIT license. You can use it as is or change it for your own needs. The provided React app is an example to help you get started. It's not a production ready app.

To deploy a search application with no code and generate an endpoint URL to your search page with access control, see Amazon Kendra Experience Builder.

The following example code adds Amazon Kendra search to an existing React web application:

The examples are modeled after the search page of the Amazon Kendra console. They have the same features for searching and displaying search results. You can use the whole example, or you can choose just one of the features for your own use.

To see the three components of the search page in the Amazon Kendra console, choose the code icon (</>) from the right menu. Hover your pointer over each section to see a brief description of the component and to get the URL of the component's source.

Topics

Overview

You add the example code to an existing React web application to activate search. The example code includes a Readme file with steps to set up a new React development environment. The example data in the code example can be used to demonstrate a search. The search files and components in the example code are structured as follows:

  • Main search page (Search.tsx)—This is the main page that contains all of the components. This is where you integrate your application with the Amazon Kendra API.

  • Search bar—This is the component where a user enters a search term and calls the search function.

  • Results—This is the component that displays the results from Amazon Kendra. It has three components: Suggested answers, FAQ results, and recommended documents.

  • Facets—This is the component that shows the facets in the search results and allows you to choose a facet to narrow the search.

  • Pagination—This is the component that paginates the response from Amazon Kendra.

Prerequisites

Before you begin, you need the following:

  • Node.js and npm installed. Node.js version 19 or older is required.

  • Python 3 or Python 2 downloaded and installed.

  • SDK for Java or AWS SDK for JavaScript to make API calls to Amazon Kendra.

  • An existing React web application. The example code includes a Readme file with steps on how to set up a new React development environment, including using required frameworks/libraries. You can also follow the quick start instructions in the React documentation on creating a React web app.

  • The required libraries and dependencies configured in your development environment. The example code includes a Readme file that lists the required libraries and package dependencies. Note that sass is required, as node-sass is deprecated. If you previously installed node-sass, uninstall this and install sass.

Setting up the example

A complete procedure for adding Amazon Kendra search to a React application is in the Readme file included in the code example.

To get started using kendrasamples-react-app.zip

  1. Make sure you have completed the Prerequisites, including downloading and installing Node.js and npm.

  2. Download kendrasamples-react-app.zip and unzip.

  3. Open your terminal and go to aws-kendra-example-react-app/src/services/. Open local-dev-credentials.json and provide your credentials. Do not add this file to any public repository.

  4. Go to aws-kendra-example-react-app and install the dependencies in package.json. Run npm install.

  5. Launch a demo version of your app on your local server. Run npm start. You can stop the local server by entering on your keyboard Cmd/Ctrl + C.

  6. You can change the port or host (for example, IP address) by going to package.json and update the host and port: "start": "HOST=[host] PORT=[port] react-scripts start". If you use Windows: "start": "set HOST=[host] && set PORT=[port] && react-scripts start".

  7. If you have a registered website domain, you can specify this in package.json after your app name. For example, "homepage": "https://mywebsite.com". You must run npm install again to update new dependencies, and then run npm start.

  8. To build the app, run npm build. Upload the contents of the build directory to your hosting provider.

    Warning

    The React app is not production ready. It's an example of deploying an app for Amazon Kendra search.

Main search page

The main search page (Search.tsx) contains all of the example search components. It includes the search bar component for output, the results components to display the response from the Query API, and a pagination component for paging through the response.

Search component

The search component provides a text box to enter query text. The onSearch function is a hook that calls the main function in Search.tsx to make the Amazon Kendra Query API call.

Results component

The results component shows the response from the Query API. The results are shown in three separate areas.

  • Suggested answers—These are the top results returned by the Query API. It contains up to three suggested answers. In the response, they have the result type ANSWER.

  • FAQ answers—These are the frequently asked questions results returned by the response. FAQs are added to the index separately. In the response, they have the type QUESTION_ANSWER. For more information, see Questions and answers.

  • Recommended documents—These are additional documents that Amazon Kendra returns in the response. In the response from the Query API, they have the type DOCUMENT.

The results components share a set of components for features like highlighting, titles, links, and more. The shared components must be present for the result components to work.

Facets component

The facets component lists the facets available in the search results. Each facet classifies the response along a specific dimension, such as author. You can refine the search to a specific facet by choosing one from the list.

After you select a facet, the component calls Query with an attribute filter that restricts the search to documents that match the facet.

Pagination component

The pagination component allows you to display the search results from the Query API in multiple pages. It calls the Query API with the PageSize and PageNumber parameters to get a specific page of results.