A proxy for The Metropolitan Museum of Art website. It queries a recommendation engine to give on-page, instant & personalized recommendations based on the visitor's browsing history. All on Fastly Compute.
This code was lightly modified from @triblondon's original Met proxy at fastly/compute-recommender-met-demo. Check out the excellent explanation video on Fastly Developers Live.
Go to https://edgeml-recommender.edgecompute.app/art/collection/search/1 and start browsing around.
As you browse, your personalised recommendations will be displayed on-page, under the ✨ For you: other artworks matching your interests heading.
Open the developer console to see the recommendation engine backend response time:
✨ Recommendations generated in 46.39ms ✨`
That's how quickly it selected artwork recommendations from the Met Museum's half-a-million-strong collection based on your recent browsing history, using ML inference on Fastly Compute. For context, the average time it takes for a human to blink is around 400 milliseconds!
Note: The difference between the console-logged time and the network request time to the
/recommendendpoint accounts for requests to the Met Museum's Collection API, to load object descriptions and images.You can make requests directly to the recommendation engine using comma-separated object IDs from the Met's collection:
https://edgeml-recommender-engine.edgecompute.app/?ids=1,2,3