How to plan and build a plugin for the AX Semantics NLG Cloud
AX Semantics relies on data to automate content. The common integration for data import is to connect a data storage application like a shop or PIM system to AX Semantics. Also quite common for content delivery is to extract the content to a CMS System. Both ways are fully automatable and can be covered by one plugin to provide a overall roundtrip for data management and content publishing.
The roundtrip from data to content on the AX NLG Cloud
The standard case on how to implement AX Semantics via our API encompases import from your data into our system, generation of content and export of the generated content back into your system:
A few tips in regard to handling your data
Now that the 3 necessary steps on the platform are clear from a technical standpoint, let’s see how to best shape the data sources environment on the AX NLG Cloud. In the data sources you have access to data containers, called “collections”.
Starting to implement the roundtrip
Get your AX Semantics account by creating one here:
The trial period is three days. If you need more time to develop your plugin, please contact us via in App Chat to get your developer license.
On the platform you can find your request token and links to the API documentation by going on your profile settings and selecting the menu entry “API Access”.
Next, it is mandatory to get a grasp on how the AX NLG Cloud works.
Go to the point “Learning” in your profile and do the “AX Semantics Seminar”.
Now that you have Access to the API and a first understanding on how the plugin might be designed, you are set for implementing the AX API Integration into your plugin.
Designing the data import
First of all, start with a full sync from your storage application to the AX Semantics platform. For that, create a Collection in your Project and import every dataset into that collection. When the data is imported for the first time, you can start to implement more (optional) advanced features for the data import:
- That includes filtered exports from your PIM for different data subsets.
- That includes importing different data subsets into different collections on the AX NLG Cloud.
- That includes a highly flexible data model that can be altered regularly in composition and structure.
Please check back with your content manager wether one or more of these cases is necessary for the user of your customer.
Crafting a data model
The next step after the data import is the matching / processing of the data in the ruleset on the AX NLG Cloud. This should be done by yourself and by one of your colleagues from your content team. Here is a small guide on how the data model crafting should be pursued.
Implementing the content generation
The initial content generation is quite easy to implement. If you are using Collections or the direct API, every method comes equipped with an option to automatically trigger the generation of content when a dataset is imported for the first time. This also can be expanded to generate content whenever a dataset is changed. It is highly advisable to first concentrate on automatic content generation for your plugin. At a later stage it might be feasible to implement a content preview which would be implemented using our “direct API” for generation in real time.
Publishing the generated content
In a best case scenario, the generated content is then ready for publishing right after the content generation. In that case you can simply extract the generated content from webhook messages or the API response (direct API) and sort it into your system.
[optional] Reviewing the generated content
This is an optional feature that implements a QA loop inside your plugin. Please check with your editorial Colleagues wether this is necessary. These are (optional!) cases to keep in mind for the QA process of the automated content:
If the initial content is not meeting expectations, the returned content needs to be
- A. altered by hand or
- B. regenerated.
Please check with a colleague from content management if and how this feature should be build into your plugin. Regeneration of a single piece of content can be achieved with this API Call.
The generated content needs to be stored for the editorial review, which is achieved by implementing draft and published collections. Depending wether your customers like the resulting content, there are four options to cover:
- Discarding the piece of content (In case your ruleset is not producing adequate results yet.)
- Regenerating the piece of content (In case you wish to see a different composition for the generated content)
- Saving the generated content to your system (In case you would like to revisit that certain result, or save manual changes.)
- Publishing the content to the intended platform (In case the review was completely successful)
The ruleset is always producing consistent results, in accordance to the defined rules. Mass operations are necessary:
- Mass-Regeneration (In case you want to apply, current changes to the ruleset to the respective batch of data)
- Mass-Publishing (The final action when the content is matching your current requirements)
If these things are present in your Plugin the customer is highly autonomous during editorial review and publishing of the generated content.
That is an overview of the technical information and considerations that should greatly help you with building your new Application with AX Semantics. If you need a second opinion or are stuck somewhere or have a finished puzzle piece that need’s review, feel free to reach out and book a Call with me.