5 ms·
Building AI Memory at 10M+ Nodes: Architecture, Failures, and Lessons
- Manik_agg 9mo agoAuthor here. We've been building CORE (open source) for the past year. Happy to answer questions about the architecture, reification approach, or what broke at scale.
- emschwartz 9mo agoHow did you decide on pgvector versus a dedicated vector database? I’m curious because it didn’t sound like your architecture included Postgres before (though maybe I’m wrong about that).
- Manik_agg 9mo agoHey we already had PostgreSQL so no new infrastructure to manage, it was easy way to see if vector database change has any value. It also has good enough performance - handles 10M vectors with HNSW indexes adequately open source - leverages existing infrastructure for future migration. we've created a vector service, easy to swap later if needed