Trip Report: VillageSQL at preFOSDEM MySQL Days
At this year's preFOSDEM MySQL Days, Dominic Preuss and Tomas Ulin presented VillageSQL’s vision for how to solve one of the bottlenecks of upstream contributions in the MySQL ecosystem through an extension framework. This annual event brings together the MySQL community to share ideas and updates. You can view our presentation at https://www.youtube.com/watch?v=VjmO50G8E-M
Innovating in MySQL with Extensions (Tomas Ulin & Dominic Preuss, VillageSQL)
VillageSQL Server is now in Alpha. Since launch, we’ve seen tremendous interest by members of the community - contributing Issues, discussing on Discord, and filing PRs. You can get started by running the install script:
curl https://www.villagesql.com/shell | sh Permissionless innovation
In the current MySQL ecosystem, adding a feature like native vector support or a custom data type requires an upstream PR or a divergent, unmaintainable fork. This is "permissioned innovation": a model where the community’s velocity is gated.
VillageSQL enables permissionless innovation by decoupling the core engine from the extension layer—ensuring upstream parity while allowing for native C++ shared library loading. By inserting hooks into the MySQL engine, we create a framework that can register and run extensions without needing to modify the core server. This framework can be used to support a variety of ideas:
- Native AI Support: We demonstrated the vsql-ai extension, which brings vector operations directly into the engine.
- Extensible Logic: The system allows developers to register custom functions and types that behave like native components.
- Custom Data Types: Libraries like vsql-uuid and vsql-network-address resolve long-standing parity issues without core-code modifications.
- Native Performance via C++: We prioritized a C++17 API to ensure low-latency, native performance.
By moving toward a modular architecture, we are attempting to give the community tools to innovate. Our current roadmap focuses on expanding the API to include custom index types—enabling high-performance vector search and specialized indexing for agentic AI workloads. Our goal is to give the community the tools they’ve been waiting for. We’re excited to see what you build with this Alpha.