Integrate Amazon Kendra with AEM as a cloud

[Prashardan]

Hi Team

I have a requirement to integration Amazon Kendra Search solution with AEM as a cloud. I have seen the documentation on Amazon however would want to know inputs if anyone integrated recently.

Basically to start with I want to do POC on AEM local cloud sdk.

Any pointers would be really helpful.

Thanks in advance.

[VishalKa5]

Hi @Prashardan ,

Steps for POC:

1. Set up a Kendra index in AWS.

2. Push AEM content (pages or fragments) to Kendra using its API or connector.

3. Build a search UI in AEM that sends queries to Kendra and shows results.

4. Handle authentication using AWS IAM or API keys.

Thanks & Regards,

Vishal

[ShafaliTh]

@SantoshSai @VishalKa5 @Prashardan 
For our local setup, we're currently using AWS session credentials for the POC.
From a security and authentication standpoint, do you have any recommendations or best practices for establishing a secure connection between the application and AEM Cloud using the AWS SDK? Specifically, what would be the suggested approach at the application level, and how should we proceed to ensure a secure and scalable integration

[SantoshSai]

I would suggest:

• Use IAM Role with Web Identity (Recommended)

  • Use AWS STS (AssumeRoleWithWebIdentity) to get temporary credentials securely.

  • No hardcoded credentials in AEM.

• Use Environment Variables / Secrets Manager

  • Store role ARN and config securely.

  • Never embed secrets in code.

• Least Privilege IAM Policy

  • Grant only kendra:Query and needed actions to the IAM role.

[SantoshSai]

Hi @Prashardan,

1. Create Amazon Kendra Index

• Go to AWS Console > Amazon Kendra

• Create an Index and note down the Index ID

Reference: https://docs.aws.amazon.com/kendra/latest/dg/getting-started.html

2. Add AWS SDK Dependency in core/pom.xml

<dependency>
  <groupId>software.amazon.awssdk</groupId>
  <artifactId>kendra</artifactId>
  <version>2.25.14</version>
</dependency>

Reference: https://sdk.amazonaws.com/java/api/latest/software/amazon/awssdk/services/kendra/KendraClient.html
3. Create Sling Servlet to Query Kendra

Use KendraClient in a servlet to query based on user input.

Basic steps:

• Read query param

• Call KendraClient.query(...)

• Parse results and return JSON

4. Test Endpoint

Access your servlet via:

http://localhost:4502/bin/kendra/search?q=your+query

5. Security Note

For local, use AWS credentials via ~/.aws/credentials.

Reference: https://docs.aws.amazon.com/sdk-for-java/latest/developer-guide/credentials.html

More Useful References:

• Amazon Kendra Overview

• Kendra Query API

• AWS SDK for Java v2

A Sample Servlet to Query Amazon Kendra that might helps you:

@SlingServletPaths("/bin/kendra/search")
public class KendraSearchServlet extends HttpServlet {

    @Override
    protected void doGet(HttpServletRequest req, HttpServletResponse resp)
            throws ServletException, IOException {

        String queryText = req.getParameter("q");
        Region region = Region.US_EAST_1;
        KendraClient kendraClient = KendraClient.builder()
                .region(region)
                .credentialsProvider(ProfileCredentialsProvider.create()) // Use env or IAM for production
                .build();

        QueryRequest queryRequest = QueryRequest.builder()
                .indexId("your-kendra-index-id")
                .queryText(queryText)
                .build();

        QueryResponse queryResponse = kendraClient.query(queryRequest);

        List<QueryResultItem> items = queryResponse.resultItems();
        JSONArray jsonArray = new JSONArray();

        for (QueryResultItem item : items) {
            JSONObject json = new JSONObject();
            json.put("title", item.documentTitle().text());
            json.put("excerpt", item.documentExcerpt().text());
            json.put("uri", item.documentURI());
            jsonArray.put(json);
        }

        resp.setContentType("application/json");
        resp.getWriter().write(jsonArray.toString());
    }
}

[kautuk_sahni]

@Prashardanwere you able to get this resolved? If you found a different way to fix it, sharing your approach would be a great contribution to the community. Your follow-up not only helps close the loop but also ensures others benefit from your experience. Thanks so much for being part of the conversation!