Product Recommendations SDK
If you need to programmatically access product recommendations for your storefront, use the Product Recommendations JavaScript SDK. The SDK is a web services API wrapper that allows you to fetch recommendations programmatically in the browser. With the SDK, you do not need to manage the full lifecycle or understand the complexity of the web services API.
Installing
Dependencies
Before installing the Product Recommendations JavaScript SDK, make sure you have installed the Catalog SaaS Export module.
Product Recommendations JavaScript SDK
This SDK can be pulled down from a CDN or installed as a module from NPM.
CDN
The SDK is published on a CDN in versioned directories.
Copied to your clipboard<script src="https://magento-recs-sdk.adobe.net/v1/index.js"></script>
NPM
Copied to your clipboardnpm install @magento/recommendations-js-sdk
Initializing the client
To programmatically fetch and render recommendations on your web site, you must first initialize your client by calling new RecommendationsClient()
.
Example usage
Copied to your clipboardconst client = new RecommendationsClient()
When you initialize the client, your store's environmentId
, instanceId
, storeCode
, storeViewCode
, and websiteCode
values are automatically retrieved by the SDK.
Registering recommendations
With the client initialized, register the recommendations you want by calling the client.register()
function and specifying the recommendation type.
Example usage
The following example registers a recommendation with a type of most-viewed
.
Copied to your clipboardclient.register({name: "Most Viewed Products",type: "most-viewed",})
The following example shows how to filter a recommendation that has a base price of less than \$200.
Copied to your clipboardclient.register({name: "Most Viewed, Under $200",type: "most-viewed",filter: "prices.maximum.regular: <200",})
Attributes
The client.register()
function contains the following inputs.
Input | Description |
---|---|
name | The user-specified name of the recommendation unit |
type | Options: most-viewed , most-purchased , most-added-to-cart , trending , just-for-you , viewed-viewed , viewed-bought , bought-bought , and more-like-this |
filter | String used to filter the results. If you are setting a filter based on price, you must use the base currency specified for your store. Currency conversion is currently not supported when filtering |
search | Defines the search criteria for your custom recommendation. This input contains the signal attribute. In non-custom recommendations, the values specified in this attribute are the types defined above. However, in a custom recommendation, the value is "query" . search might also contain a key attribute, such as "categories:(159 OR 377)" . The key attribute is not required by all custom recommendations. Some types require you to be on a product page, which would mean you know the current SKU. Other types are more broad, and do not require you to have any specific filter data as they are site wide. Site wide types do not require the key attribute. Possible key values are: user_purchase_history , cart , current_pdp , user_view_history , and <custom query> |
boost | User-specified value that indicates the rank of a specific recommendation |
Filtering
Currently, you can filter on categories and prices.
Category
To include specific categories:
categories: (<url-key-1> OR <url-key-2> OR ...)
To exclude specific categories (note the -
at the beginning):
-categories: (<url-key-1> OR <url-key-2> OR ...)
Prices
To filter based on a specific price point (note that these filters use $50
as the price point):
prices.maximum.final: <50
prices.maximum.regular: <=50
prices.minimum.final: >50
prices.minimum.regular: >=50
Registering custom recommendations
Instead of using one of the built-in types, the SDK provides a way to define a custom type by calling the client.register()
function and passing in specific search criteria. For example, you can pass in a search query, such as "categories:(159 OR 377)"
to register a recommendation that has a category ID of either 159
or 377
.
Fetching recommendations
You can fetch the registered recommendations by calling the client.fetch()
function.
Example usage
Copied to your clipboardconst {status, data} = await client.fetch()
The following shows an example of the fetched recommendations. This example is intentionally truncated.
Copied to your clipboard{"units": [{"unitId": "45687","unitName": "test-recs","searchTime": 10,"totalResults": 3,"results": [{"rank": 1,"score": 0.38299224,"sku": "35123","name": "Pursuit Lumaflex™ Tone Band","shortDescription": null,"type": "simple","categories": ["gear","gear/fitness-equipment"],"weight": 0.0,"weightType": null,"currency": "USD","image": {"label": "","url": "http://magento2sc.local/pub/media/catalog/product/cache/fbb00452bcc1f45faf89264b683c708f/u/g/ug02-bk-0.jpg"},"smallImage": {"label": "","url": "http://magento2sc.local/pub/media/catalog/product/cache/fbb00452bcc1f45faf89264b683c708f/u/g/ug02-bk-0.jpg"},"thumbnailImage": null,"swatchImage": null,"parents": [],"url": "http://magento2sc.local/pursuit-lumaflex-trade-tone-band.html","prices": {"maximum": {"finalAdjustments": [],"final": 16.0,"regular": 16.0,"regularAdjustments": []},"minimum": {"finalAdjustments": [],"final": 16.0,"regular": 16.0,"regularAdjustments": []}}}...
Attributes
The client.fetch()
function contains the following inputs.
Input | Description |
---|---|
ids | Specifies the IDs of the recommendations. If unspecified, all recommendations are fetched; otherwise, only those recommendations you specify are fetched |
limit | (Optional) Specifies the number of recommendations to fetch. The maximum is 25 |
offset | (Optional) Specifies where in the recommendations array to begin fetching the recommendations |
currentSku | The SKU of the product on the current product page |
cartSkus | The SKUs of the products within the cart |
userViewHistorySkus | List of SKUs the user recently viewed |
userViewHistory | List of recent user views |
userPurchaseHistory | List of recent user purchases |
Injecting recommendations on your web site
To display recommendations on your web site, you need to define a template for the recommendation then render the recommendation on the web site.
Defining a template
The SDK works with the mustache.js
template.
Rendering a recommendation on your web site
The client.render()
function creates a string of HTML you can then place on your web site.
Attributes
The client.render()
function contains the following inputs.
Input | Description |
---|---|
template | A mustache.js HTML template string. Uses reserved variable names such as {{#rec-items}} : {{url}}, {{image}}, {{title}}, {{price}} , which are keys returned in the unit object |
unit | An object returned by the client.fetch method that contains the results |
Using the SDK from start to finish
The following example shows a sample workflow beginning with importing the SDK to rendering the recommendation.
Copied to your clipboardimport RecommendationsClient from "@magento/recommendations-js-sdk"// create the clientconst client = new RecommendationsClient()// register pre-built recommendation unitclient.register({name: "Most Viewed Products",type: "most-viewed",})// retrieve recommendations for all unitsconst {status, data} = await client.fetch()// render the markupconst markup = client.render({unit: data.units[0],})// insert the markupdocument.body.insertAdjacentHTML("beforeend", markup)