The Shop sends two types of data to Inbox.Wine:
‍
- Price List: This contains the inventory of wines available, including details like the name, varietal, region, price, and possibly wine description.
‍
- User Purchase Data: This includes the historical purchase data of users, which can entail which wines have been bought, the frequency of purchases, and any user ratings or feedback if available.

Inbox.Wine takes the data provided by the Shop to create custom AI Models.
‍
The price list helps in understanding the variety of wines available and potentially their characteristics.

The user purchase data is used to identify patterns and preferences in the wine selections of different customers.
‍
This custom AI model is trained to understand correlations between different wines, customer preferences, and possible recommendations.

Once the model is trained, Inbox.Wine provides an API Endpoint.
‍
This API is the interface through which the Shop Website can communicate with the Inbox.Wine service to get recommendations.

The Shop Website integrates with the Inbox.Wine API.

When a user is browsing the website and looking at different wines, the website can send a Request for Recommendations to the API Endpoint.

This request would typically include the ID of the wine being viewed, or the user ID to get personalized recommendations based on their history.

The Inbox.Wine service processes the request, with the AI Model generating the appropriate recommendations.
‍
These recommendations are then sent back to the Shop Website through the API Endpoint.
‍
Process takes only subsecond, so it's could be integrated inline even inside product page.

Finally, the Shop Website displays these recommendations to the user.

For instance, when a user is viewing a particular wine, the website can show the most similar wines or personalized recommendations alongside it.

This can also extend to the homepage, where the user sees a Netflix-style list of recommended wines based on their previous browsing and purchase history.