VexDB-THU/VexDB-Lite

The pitch here is integration, not replacement. Rather than asking a team to adopt a standalone vector store and figure out sync, VexDB-Lite slots in as a plugin to a database the app already trusts. That framing matters for teams where adding a new service is a bigger friction point than adding an extension. Cross-platform build means it is not locked to a specific OS or cloud environment. The research origin (THU) suggests the indexing work is not just a naive wrapper over FAISS — there is likely real algorithmic work underneath, though production maturity of an academic release always warrants scrutiny. Reservation: the repo is early and documentation is sparse; if you need a production-grade vector store today, Pgvector or Qdrant are better-understood options. This is worth watching for teams who hate running a separate vector service and are willing to take an early dependency. -> Best for: SaaS team of 2-5 wanting vector search without a standalone service.