Event-based personalized search + recommendations

[eostis]

Describe your feature request

The most requested feature nowadays is a real-time recommendation system, based on user events, like clicks, add to wish-list, add to basket, or orders, views, video display ....

It is so difficult to set in motion that only paid SaaS services are considered in general. Open source/local recsys are either too slow or not good enough (lack of real time, lack of good models for specific goals ...).

Features in the API:

• Ingest catalog
• Ingest events
• Ingest users
• dashboards
• create scenarios from a selection of models (items-to-item, items-to-user, ...), boosts, and filters
• call scenarios to get recommendations
• call scenarios to get personalized search (reranking?)

Code of Conduct

• I have read and agree to the Weaviate's Contributor Guide and Code of Conduct

[bobvanluijt]

Sooner than you might think 😌

[eostis]

You've got your first use case!

[eostis]

I forgot to mention search personalization! Done.

[eostis]

Perhaps integrating the new UBI format?

• https://github.com/opensearch-project/user-behavior-insights
• https://www.youtube.com/watch?v=8_4VRtMWtSY

[alexanderdean]

Exciting! Flagging this one to @yalisassoon and @miike

[miike]

Happy to share our (Snowplow) experience here if it helps as one of our use cases has been proxying behavioural events (primarily for ecomm and media rec systems) to recommendation systems like AWS Personalize and Vertex AI search.

[bobvanluijt]

Thanks all!

We are planning a data to release this to our cloud console, we will make sure to let you know when it's live.