Shannon-Data/ShannonBase

Embeds machine learning models (LightGBM, XGBoost) directly within stored procedures and natively supports vector data types for RAG/embedding workloads, eliminating ETL overhead. Features a dual-engine architecture with Rapid (in-memory columnar store) for analytical queries and InnoDB for transactional workloads, using ML-based cost optimization to route queries intelligently. Extends MySQL compatibility with multimodal data support (GIS, JSON, vectors) and a lightweight JavaScript engine for procedural logic.

165 stars.