Exploring the headless CMS functionality of AEM 6.5

Reading Time: 5 minutes

Adobe Experience Manager (AEM) is one of the leading enterprise content management system (CMS), formerly knows as Day CQ. The initial version was introduced in 2002, at a time when web projects were mostly implemented as static, server-side rendered websites. Content as well as styling information were mixed up within the same HTML document. Nowadays traditional websites are being more and more replaced and complemented by (mobile) applications which come up with their own presentation layer. Thus there is a need for a more flexible approach that separates content from styling, and that is able to deliver the data to multiple channels in a universal format. This requirement led to the emergence of so called headless content management systems, which usually supply the data in a RESTful manner (as JSON) to their consumers. This blog post summarizes the headless CMS extension provided by AEM.

Headless CMS in AEM

The headless CMS extension for AEM was introduced with version 6.3 and has been continuously improved since then, it mainly consists of the following components:

  • Content Services: Provides the functionality to expose user-defined content through a HTTP API in JSON format. This allows to deliver data to 3rd party clients other than AEM.
  • Content Fragments: Allows the user to insert/edit content as structured data entities. The schema of each content fragment is defined by a corresponding Content Fragment Model.

Note that AEM follows a hybrid approach, e.g. content fragments can either be delivered as JSON through the content services API, or embedded within a traditional HTML page. Visit Adobe’s headless CMS website for further information.

Example Project

There is a tutorial provided by Adobe where the concept of content services is explained in detail. It describes how to model the entries of a FAQ list by using content fragments, and how to expose this data through a API as JSON. The complete article can be found here.

The example is based on the existing We.Retail demo project that comes with the installation file of AEM. In summary, the following steps have to be performed:

  • First content fragment models should be enabled for the We.Retail project. From the AEM welcome page, go to ToolsConfiguration Browser, open the properties of the We.Retail configuration and ensure that the Content Fragment Models property has been selected.
  • Navigate to Tools → Assets → Content Fragment Models → We.Retail to create or edit content fragment models. Select the FAQ model and click on the edit button to open the Content Fragment Model Editor. Here you can edit the model structure by adding/removing elements using drag and drop.
  • The model can be used to create new content fragments which contain the actual data . To do this, navigate to Assets → Files → We.Retail → English → FAQs → Company. There is already content available here: Each entry of the FAQ list is modeled as a single fragment. The picture below shows the editor view for a FAQ content fragment.
  • To access the data through content services from outside, the FAQ items have to be embedded within a content page. Content fragments can be added to the page by drag and drop in the same way as any standard content component.
  • The content of the page can now be delivered as JSON via the “model” selector. The code section below shows an extract of the response of the FAQ page /content/we-retail/language-masters/en/faq.model.json
":items": {
        "contentfragment": {
          "title": "The Company Name",
          "description": "",
          "model": "we-retail/models/faq",
          "elements": {
            "question": {
              "title": "Question",
              "dataType": "string",
              "value": "What's with the name \"We.Retail\"?",
              ":type": "string"
            "answer": {
              "title": "Answer",
              "dataType": "string",
              "paragraphs": [
              "value": "<p>We're not sure, but it sounds good!<br>\n</p>\n",
              ":type": "text/html"
          ":items": {},
          "elementsOrder": [
          ":itemsOrder": [],
          ":type": "weretail/components/content/contentfragment"
        "contentfragment_1811741936": {
          "title": "History",
  • Finally the data can now be consumed by any 3rd party application. The screenshot below shows an example made with vue.js, where the FAQ list is loaded from AEM content services by XHR request.


AEM content services provide a flexible way to deliver structured content to multiple channels, the data as well as its corresponding schema can be created and modified without any need for a deployment. However, the main focus of AEM is still on the authoring of backend-side rendered websites, but content services may be a good addition for environments where AEM is already in use.

Project story: Automate AEM deployments for a Swiss bank

Reading Time: 5 minutes

A large bank in St. Gallen, Switzerland had the need to improve the AEM deployment process for its various staging environments. It was one of my first projects for 47 North Labs and was settled to run for 6 months starting in October 2018. The following blog gives a short project overview.

Getting started

Starting to work for a new customer is always exciting to me because every company and team has a unique mindset and culture. Usually, it takes a few days or weeks to get to know the new teammates. But this time it was completely different, as I already worked with each of the three team members and their supervisor together in one before my previous company. It was nice to meet old colleagues and we had a very good start.

Deployment process before automation.
Source: www.dreamstime.com

Technology Stack

The technology stack was already defined and the servers ordered. But it took a while until the infrastructure was ready and for the time being, I worked on my local machine.

Jenkins was set as the central tool for build orchestration, deployments, and various DevOps tasks. All the pipeline source code is stored in GitLab and the main business application we’re dealing with is Adobe Experience Manager (AEM).

A relatively large amount of work was needed for the initial setup like enabling connectivity to the relevant systems, basic shared library, and getting to know the internal processes. Read more about Jenkins behind a corporate proxy as an example for this setup: https://www.47northlabs.com/knowledge-base/update-jenkins-plugins-behind-a-corporate-proxy/

Implemented Pipelines

The bank has two different AEM projects: one for the corporate website and another for their intranet. They require a slightly different deployment pipeline and both have three environments: development, staging, and production.

Besides the deployment pipelines, there are pipelines for copying content from the production to the development environment and restoring a complete production environment into the staging environment in order to have an exact copy and a good baseline for approvals.

Many auxiliary jobs like start/stop AEM + Dispatcher, checking the health of instances, fetch last backup time, and execute Groovy scripts are used in the deployment pipelines as well in an independent executable job.

An example of a Jenkins Pipeline
Source: https://jenkins.io
An example of a Gitlab Pipeline
Source: https://docs.gitlab.com/ee/ci/pipelines.html

Advantages of automation

The automation of the various processes brought faster deployments. But more important transparency and centralized logs about what exactly happened and higher quality as repetitive tasks are always executed.

One example is the backup check, which needed coordination and forced to long waiting times. Now an API is used and the automation pipeline has instant feedback about the last backup time and shows a note if a backup is missing. Before, such a step might have been skipped in order to save some time.

With each built pipeline, some more little and reusable helpers were introduced which made it then again easier and faster to create the next pipelines. Think of a construction kit.

Deployment process with automation.
Source: www.wikimedia.org

Project finished ➞ client is very happy

After several months of close collaboration, more and more pipelines have been implemented and are used to support the crucial deployment processes countless times.

I enjoyed building-up the AEM automation and believe it’s a very good aid for higher quality and further extensions.

After a warm welcome and six months of working together, it was time to say good-bye as the project had a fixed time span. The client’s team was very kind and gave me even some great presents to remember the exciting time in St. Gallen.

Present from client: Swiss beer, chocolate, bratwurst and biber

How to get a service reference or BundleContext with no OSGi context

Reading Time: 2 minutes


In Adobe Experience Manager (AEM) projects developers are working a lot in services, filters, servlets and handlers. All of these classes are OSGi components and services using the Felix SCR annotations or the newer OSGi DS annotations. But sometimes you need an OSGi service or the BundleContext also in non OSGi / DS controlled class


You can use the OSGi FrameworkUtil [1] to get the reference to the bundle context from any object. The code below shows how to get reference to the BundleContext and the service.

Most of the time, you have the SlingHttpServletRequest ready to pass:

import org.apache.sling.api.SlingHttpServletRequest;
import org.apache.sling.api.scripting.SlingBindings;
import org.apache.sling.api.scripting.SlingScriptHelper;
import org.osgi.framework.BundleContext;
import org.osgi.framework.FrameworkUtil;
import org.osgi.framework.ServiceReference;

import java.util.Objects;

public class ServiceUtils {

     * Gets the service from the current bundle context.
     * Return null if something goes wrong.
     * @param <T>     the generic type
     * @param request the request
     * @param type    the type
     * @return        the service
    @SuppressWarnings({"unchecked", "rawtypes"})
    public static <T> T getService(SlingHttpServletRequest request, Class<T> type) {
        SlingBindings bindings = (SlingBindings) request.getAttribute(SlingBindings.class.getName());
        if (bindings != null) {
            SlingScriptHelper sling = bindings.getSling();

            return Objects.isNull(sling) ? null : sling.getService(type);
        } else {
            BundleContext bundleContext = FrameworkUtil.getBundle(type).getBundleContext();
            ServiceReference settingsRef = bundleContext.getServiceReference(type.getName());

            return (T) bundleContext.getService(settingsRef);

Or you just use the class that was loaded over a bundle classloader.

package com.sanitas.aem.core.utils;

import org.osgi.framework.Bundle;
import org.osgi.framework.BundleContext;
import org.osgi.framework.FrameworkUtil;
import org.osgi.framework.ServiceReference;

public class ServiceUtils {
     * Gets the service from the current bundle context.
     * Return null if something goes wrong.
     * @param <T>     the class that was loaded over a bundle classloader.
     * @param type    the service type.
     * @return        the service instance.
    @SuppressWarnings({"unchecked", "rawtypes"})
    public static <T> T getService(Class clazz, Class<T> type) {
        Bundle currentBundle = FrameworkUtil.getBundle(clazz);
        if (currentBundle == null) {
            return null;
        BundleContext bundleContext = currentBundle.getBundleContext();
        if (bundleContext == null) {
            return null;
        ServiceReference<T> serviceReference = bundleContext.getServiceReference(type);
        if (serviceReference == null) {
            return null;
        T service = bundleContext.getService(serviceReference);
        if (service == null) {
            return null;
        return service;


[1] https://osgi.org/javadoc/osgi.core/7.0.0/org/osgi/framework/FrameworkUtil.html