We evaluated pgrust v0.2 and v0.3 against the same held-out corpus: 164,000 test cases from 50 open-source Postgres drivers, ORMs, and applications that we didn't use while fixing v0.3. v0.2 diverged from Postgres on 1,454 cases. v0.3 diverged on 11.
Our goal is 100% compatibility with Postgres.
We used Antithesis and workers.dev for simulation testing, Detail.dev for LLM-assisted code analysis, home-built differential fuzzers, and Kani for formal verification.
We're running pgrust internally, but there are still bugs. pgrust is not production ready. Do not put data you care about in it. Keep Postgres as the source of truth for data you can't afford to lose.
Try the browser demo, or try a real workload in a non-critical environment. If it behaves differently from Postgres, file an issue with a reproduction.