WeCommerce demo uses Weaviate and hybrid search to find products by vibe, not keywords

victorialslocum · x · 2026-07-29

A demo fashion e-commerce site called WeCommerce shows how product discovery can work with semantic search instead of exact keyword matching.

The stack stores each product in Weaviate with embeddings, then uses hybrid search that combines BM25 keyword matching with vector similarity. Users can search by vibe phrases like “garden tea party” or “berlin techno rave”, then refine results with standard filters such as category, price, and collection. The same approach is used for recommendations, replacing “customers also bought” with nearest-neighbor retrieval in embedding space.

Original post →

More from Infra

Infra channel →