Performance
Process boundaries, measured round trips, and bounded memory use.
Tablify is written in Rust on GPUI and ships as a native binary. A typical install is around 30 MB, against a few hundred megabytes for a comparable Electron app.
Measured round trips
Tablify runs the UI, a backend service, and each database driver as separate processes that talk over a private local channel. On Apple Silicon, a full round trip through that path measures a p95 of 1.83 ms, and 1.61 ms when measured against the driver directly. The performance budgets for those paths are 8 ms and 5 ms.
Bounded by design
Tablify puts explicit limits on every queue, pool, result set, cache, history, log buffer, and telemetry buffer. Large exports stream to disk as background jobs instead of staying in memory. Every asynchronous result carries the session and operation that created it, so Tablify discards stale results rather than showing them.
Crash isolation
Each database driver is a separate process. If a driver crashes, it fails its pending work and cannot take down the app.