The tech stack behind quelvinetesvous.fr

You may have seen it going around on social media, at Winesee we've just launched the "What wine are you?" website and apps in partnership with Système U.

This week marks the start of the wine fair in U stores, and the wine recommendation technology based on your sensory profile that we develop at Winesee was chosen to guide consumers all over France in a national rollout (more than 900 stores).

The concept

In France, 80% of people buy their wine for the wrong reasons: price, label, medal, etc... At Winesee, our mission is to give power back to the consumer for their wine purchase, which is why we developed a recommendation engine based on the user's taste profile, so that they choose their wines according to their own preferences.

By discovering your food preferences, our algorithm will be able to recommend the best wine made for you among all the references analyzed one by one by our experts.

The algorithm and the data used come from 25 years of research by our neurophysiology expert applied to taste and emotions, and today allow us to offer an innovative, reliable, and consumer-centered recommendation.

The tech stack

For those interested, here's the tech stack set up for the project. Note that the first version of our recommendation engine was born in 2016, and the front-end technologies used are those of that time.

Front-End

• Framework: AngularJS
• CSS library: None
• Toolkit: Gulp
• Dependency manager: Bower

The native iOS and Android apps are web versions packaged inside WebViews. In the future, the front-end will be migrated to Vue or React, and the native versions could be rebuilt with Vue/React Native.

The compilation and archiving of the iOS app is done using the bitrise.io service, which I introduced a few weeks ago right here.

Back-End

• Environment: NodeJS
• Framework: ExpressJS
• DB: MongoDB (Mongoose)
• Testing: Expect.js
• Logging: Sentry

The back-end is built as a (stateless) Rest API, able to be replicated as needed in the cloud and documented with Swagger.

Since our API is called from several other applications, we developed an sdk decoupled from the AngularJS part, working with Axios for server calls.

Hosting

• Front-end: AWS S3 + Cloudfront (CDN)
• Back-End: Heroku + S3 for static files
• Database: MongoDB Atlas
• SMTP server: Amazon SES

The cloud hosting is designed to handle scaling without major difficulty, and the platform's stress tests were carried out using the https://loader.io platform, which I'll talk about in more detail in an upcoming article.

Conclusion

Our recommendation engine is only at the beginning of its rise, but I must admit I'm quite proud to see a project in which I had the chance to design the design, the user experience, the technical architecture, and the code rolled out all across France.

If you haven't tried it yet, now's the time, and it's right here: https://quelvinetesvous.fr, and feel free to share it to help us break the codes of the wine world!