How to Blend Hybrid Retrieval with Structured Data

Level up from “hello world” to production-grade. In this live-recorded, developer-first deep-dive, Andriy Burkov and Alejandro Cantu take MindsDB’s open-source Knowledge Bases beyond the basics — showing you how to tailor everything from chunking rules to ranking strategies, blend vector and keyword retrieval for pinpoint accuracy, and query relational tables the same way you query text.

We crack open a real database, wire up a Knowledge Base in one SQL command, then live-tune it to handle both unstructured docs and structured rows — returning answers you can trace back to exact records.

What you’ll learn

• Customize end-to-end — override chunk sizes, embedding models, ranking weights, and more.
• Query structured data — ask natural-language questions over your tables and get row-level citations.
• Debug with transparency — inspect retrieval scores, traces, and explainers in real time.
• Ship to production — clone the notebook, swap in your data, and deploy advanced semantic + SQL search inside your existing workflows. No external vector store or ML ops needed.