Why database-level audit

Teams focus on user-facing features. Application-level audit is easy to miss — and changes made via API, UI, or direct SQL do not always get recorded the same way.

PostgreSQL Audit captures row-level changes in the database, so history is complete regardless of how data was modified. Install early and history exists from the first meaningful data, not from the day audit finally became urgent.

Foundation is complete product value on its own: inspect, verify, and reconstruct change history with SQL — compliance, forensics, and debugging without a separate UI. The Application layer is additive: UIs and APIs that surface the same history when you need broader self-serve access, without a new backend audit programme.

ML training data provenance: When ML models train on your database, tamper-evident history supports reconstructing training-relevant state at a point in time, with integrity verification on the audit trail. Useful for documentation, reproducibility, and model debugging — data provenance, not an “AI audit” product.

Product and implementation details are in the GitHub repository. Questions or early-access interest: send feedback.