GenDB 0.1.0
I pushed out the first release of GenDB today.
What it does
GenDB sits between your application and PostgreSQL as a wire protocol proxy -- silently acting as a test data manager.
An LLM analyzes your schema -- column names, types, foreign keys -- and picks appropriate data generators for each column. The result is a synthetic copy of your database that looks realistic without containing any production PII.
What's in this release:
- LLM-driven data generation -- point it at your database and it figures out what kind of data each column needs
- PostgreSQL proxy -- intercepts connections and routes queries to real or generated data per table
- GENDB SQL -- control everything through standard
CALL gendb.*()syntax from any PostgreSQL client - Full CRUD -- the synthetic schema contains real PostgreSQL tables, so reads and writes work normally
- Per-table routing -- switch individual tables between real and synthetic data without restarting anything
Where things stand
This is not production-ready code. The core flow works -- start the proxy, generate data, toggle routing -- but there are rough edges. The LLM analysis works best with descriptive column names and well-defined foreign keys. Tables with vague names or missing constraints will get less realistic data, and I'm sure there are bugs I haven't found yet.
I'm planning to focus on better generator customization and multi-schema support next.
If you want to try it out, check the Getting Started guide. Issues and feedback on GitHub.