Skip to content

JDtapcode/integration-example-recombee

Repository files navigation

Core integration

Last modified Issues Contributors Forks MIT License

Stack Overflow Discord

Kontent.ai logo for dark/light scheme. recombee logo

DemoPrerequisitiesGet startedRecombee data structureGetting recommendationsDeleting itemsDevelopingContributorsLicenseResources

This repository contains an example implementation of integration between Recombee AI-powered content personalization and recommendation engine, and Kontent.ai headless CMS. It comes with an optional custom element for indexing your content in Recombee, as well as two Netlify functions for the initial setup and subsequent processing of content changes via a webhook.

Demo

recombee-custom-element

Prerequisities

To run this integration, you'll need a Kontent.ai project + a Recombee account.

Contact Recombee to recieve a special offer for Kontent.ai customers!

Get Started

  1. Quick Deploy

    Netlify has made this easy. If you click the deploy button below, it will guide you through the process of deploying it to Netlify and leave you with a copy of the repository in your account as well.

    Deploy to Netlify

    After you deploy the project into Netlify, you'll find two functions there:

    1. recombee-init-function is the function that imports or refreshes all content of selected content type from Kontent.ai into Recombee.
    2. recombee-sync-webhook is the function you want to call via a webhook from Kontent.ai to notify the integration that there's been a change to your content and that Recombee database has to be potentialy updated.
  2. Get your credentials from Recombee

    In order for the integration to work, you'll need a set of credentials for Recombee. In order to obtain those, just login into your Recombee account, select your database and you'll want to copy the API Identifier and Private token that can be found under the Settings section. These credentials will be neccessary for the following steps.

recombee-keys

  1. Create a new Webhook in Kontent.ai to Sync your data with Recombee

    The next step is creating a new webhook in Kontent.ai.

    Fill out the following into the webhook's URL address field:

    https://<YOUR_NETLIFY_APP>/.netlify/functions/recombee-sync-webhook?apiId=(Recombee API Identifier)&types=(Kontent.ai content types to be processed by the webhook)&languages=(Kontent.ai languages to be processed by the webhook)

    You can include multiple Kontent.ai types and languages the webhook will be looking out for and synchronizing - to do so, just include them as a comma-separated list.

    webhook-setup

    Subsequently, set the Triggers to published data. Then choose Specific events and select only Content item events. (You can also unselect Content item metadata changed events as they won't affect the resulting data in the index.)

    The resulting triggers setup should look like this:

    webhook-triggers-setup

    Save the webhook and copy the generated secret as it will be required as a parameter in the following step.

  2. Configure your Netlify functions

    To setup both function, all you need to do is set the following environment variables for your Netlify site. You can find our tutorial on configuring Netlify functions here.

    Variable Value
    RECOMBEE_API_KEY your Recombee private token
    KONTENT_SECRET your Kontent.ai webhook secret

    The Recombee private key is used by the function to synchronize your recommendation database.

    The Kontent.ai webhook secret is used to ensure integrity of the webhook payload. It has been created in the previous step.

    After your function has been redeployed (or your environment variables has been propagated), the functions are ready to be called.

    You can observe the real-time Function log on the same screen you'd find your function's endpoint (Functions -> your function).

  3. Register a content type into your Recombee database

    The initialization of your Recombee recommendation database with your content is done through the recombee-init-function. Simply make a POST request towards the function's endpoint URL (https://<YOUR_NETLIFY_APP>/.netlify/functions/recombee-init-function) with the following payload:

    {
      "projectId":"{Kontent.ai project ID}",
      "language":"{Kontent.ai language codename}",
      "contentType":"{Kontent.ai content type codename}",
      "recombeeApiId":"{Recombee's API Id}",
    }
    

    The function processes all published content of the given content type and language from your project and creates or updates your recommendation database in Recombee.

    Please keep in mind, that this example integration does not fully support modular content, i.e. it only includes codenames of linked items, not the whole content structure and subsequent links as it makes sense to choose mostly flat types for recommendations (i.e. articles, blogposts, ...).

    Alternatively use a custom element to manage your registered content types

    Simply add a new custom element into your content model (it will be a good idea to create some kind of meta content type to store some of the project's settings, including this element).

    Use your netlify's URL for the base page as the Hosted code URL and a following settings to setup the custom element:

    {
      "recombeeApiId": "{Recombee's API Id}"
    }
    

    The custom element allows you to (re)register your content types into your Recombee database. It also saves codenames of types you have already registered through it. Check out the demo gif above! All of your content model properties (including images) will be modeled and transfered into the recommendation database.

    If you are working with content in multiple language variants, adding additional variants to Recombee is just as easy as simply switching to the desired language variant and registering the content type through a custom element included in that variant (see gif below).

    recombee-multilingual

    Please note, that the element actually stores the types that have been registered in Recombee as its value, so if you use it in multiple places, it might not display the registered content types correctly.

Recombee data structure

In order to process data from Kontent.ai and save them into Recombee, the functions map your elements from Kontent.ai directly into Recombee under the same name. The linked items are saved as a set (array) of codenames. Every item will also have codename, type, language, collection, and last_modified properties on top of the specific elements. Knowing the structure of your data will allow you to use the full potential of Recombee's filtering and boosting features. You can also sync multiple content types with one Recombee database.

If you need help with your use-case, feel free to contact us on our Discord, or reach out to Recombee directly.

recombee-data

Sending interactions and recieving recommendations

Now that you have your recombee database populated with your content, it's time to start sending content interactions into Recombee. The recommendation engine works with provided information about your content, as well as with how your visitor interact with said content. You can send multiple different types of user-content interactions (simple view, purchase, ...). To learn more check out the Recombee documentation on interactions.

There are also multiple ways to get recommendations for your website once you are all setup. The most simple and streight forward is using the Recombee widget that you can create directly from the Recombee's UI. Once you set it up, you'll recieve a generated HTML code that you can just plug into your website and you are good to go (it also automatically tracks and sends view interactions back into Recombee). For more please refer to Recombee documentation on getting recommendations.

recombee-data

Deleting archived or unpublished items

Sometimes a content item has to be archived, or unpublished. Once this item becomes unavailable, it should be also removed from the recommendation engine so your recommendations are always valid. This integration, by default, removes unpublished and archived items from your recombee database, however, by doing this it also removes all of the previous item interactions that are being used to construct your recommendation model. To avoid this, Recombee suggests creating a dedicated model property that will be used to determine if the item is valid or invalid. You'll then need to use this property to request only valid recommendations. Read more about this topic in Recombee's documentation.

Developing

Netlify Dev is highly recommended for local development.

# Initial project setup
$ npm install

# Run locally
$ netlify dev

The custom element is created with Vite.

Contributors

We have collected notes on how to contribute to this project in CONTRIBUTING.md.

contributors

License

MIT

Additional Resources

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages