7 posts filed under kit.
The mongreldb-kit CLI is a pile of verbs on purpose, and the discipline that makes it useful is sorting them by pipeline stage: read-only checks go in CI, one-way doors go in deploy, and the operator verbs stay in the terminal.
MongrelDB Kit ships fixture create/load for moving rows in and out of a database as JSON, and generate types for emitting TypeScript, Rust, and Python row definitions from the schema, so the two artifacts that rot fastest stop being hand-maintained copies.
MongrelDB Kit splits schema safety into two commands: schema validate answers 'is this file correct' in CI without a database, and diff answers 'does the live database match' at deploy, with generate migration to close the gap.
MongrelDB Kit exposes the engine's storage knobs as one-line calls per table: transaction spill thresholds, mutable-run spill, compaction zstd level, result cache budget, and the deferred-versus-eager index build policy.
A synchronous native database call that takes forty milliseconds stalls every request your Node process is serving, so MongrelDB Kit ships genuinely non-blocking async twins for the hot read and write paths, and this post covers which ones are real and which ones are not.
An embedded database has no daemon to scrape and no slow query log to tail, so MongrelDB Kit puts the storage engine's own per-table counters directly on the database handle.
A durable commit costs about a thousand times more than the write itself, so MongrelDB Kit ships WriteBuffer, an opt-in micro-batching surface that amortizes one fsync across a thousand rows.
