Vector Search is now freely available in MongoDB Community Edition - but "available" and "production-ready" are not the same thing. This session, the first public deep-dive into Percona Search for MongoDB, walks through what it actually takes to run vector search at enterprise scale on self-managed infrastructure.

We open where every Percona Server for MongoDB user eventually arrives: your application needs to find documents by meaning, not just by exact field values. We build the mental model from scratch - full-text search for the keyword gap, vector search for the semantic gap, hybrid search that combines both, and RAG that closes the loop to natural-language Q&A - all grounded in MongoDB data patterns your team already works with.

Then we get honest about the gaps, as running vector search in production on self-managed MongoDB means solving two layers of problems at once:

1. The embedding pipeline: choosing among 100+ models, integration, and eventually building MLOps queuing infrastructure just to ingest at scale.
2. Enterprise readiness: upstream Vector Search is still in Public Preview, with no production-grade authentication integration, no observability into service and index health, and no automated deployment story for Day 1 and Day 2 operations.

We'll walk through a real reference deployment showing how to address these challenges in practice: from existing database deployment to extending it with Vector Search capabilities powered by automatic embeddings, enterprise authentication, and PMM monitoring, covering your vector indexes alongside your data.

The session closes with a live demo of the complete Percona Search for MongoDB stack with a demo application.