Compatibility · metrics4s · Sendline Open Source

What the artifacts promise

Before 1.0.0, a minor release may break the API. Every break is listed in CHANGELOG.md with what to change. versionScheme := "semver-spec" says so to sbt and to Scala Steward.
From 1.0.0 on, MiMa checks binary compatibility against the previous release on every build. Every deliberate break inside a major version is then an exclusion in build.sbt and an entry here, so nothing is waived silently.
New members of Effect, Dispatch and the backend SPI carry default implementations wherever one exists, so an instance of your own keeps compiling across minor releases.
Configuration types (MetricsConfig) are builders with a private constructor, not case classes, so adding an option is not a binary break.

	Version
Scala	`3.9.0` (built on the LTS); the `it` module compiles on `3.10.0` against it
JDK, to run	17 or newer: every module is compiled to Java 17 class files. `metrics4s-rapid` needs 21, because rapid runs on virtual threads, and `metrics4s-kyo` needs 25, because Kyo ships Java 25 class files
JDK, to build	25 or newer, checked when the build loads
Prometheus Java client	`1.10.0`
cats-effect	`3.7.1`
ZIO	`2.1.27`
Kyo	`1.0.0-RC7`
rapid	`2.9.9`
http4s	`0.23.38`
tapir	`1.14.1`
sttp client	`4.1.1`
MongoDB driver	`5.14.0`, and `mongo4s` `5.1.1` in the integration tests. Drivers `5.6.0` to `5.13.0` carry JAVA-6313
Prometheus server	`v3.5.0` in the integration tests
Grafana	`12.0.0` in the integration tests; the dashboards use schema version 39

The metric names are unchanged, so dashboards and alerts keep working. Code changes:

commons	metrics4s
`Counter.builder().name(...).labelNames("outcome").register(registry)`	`Counter("...", "...").labelled[Outcome]`, registered once with `Metrics[F].register`
`labelValues(outcome.value).inc()`	`inc(Outcome.Ok)`
`Histograms.seconds(...)` and `record(time)`	`Histogram.seconds(...)` and `observe(time)`
`histogram.timed(fa)` (`metrics-effect`)	`histogram.timed(fa)`
`CallbackGauge.resource(builder, source)`	`Metrics[F].observeAllF(source, timeout)(Observation(...))`
`CatsRuntimeMetrics.resource(registry)`	`metrics4s.runtime.cats.CatsRuntimeMetrics.resource()` with a `given Metrics[F]`
`ZioRuntimeMetrics.resource(registry)`	`metrics4s.runtime.zio.ZioRuntimeMetrics.scoped`
`KyoRuntimeMetrics.resource(registry)`	`metrics4s.runtime.kyo.KyoRuntimeMetrics.scoped`
`MetricsTestkit.snapshot(registry)`	`MetricsTestkit().series`, or the typed `value` and `histogram`
`PrometheusMetrics.start(port)`	`PrometheusExporter.allocated(port, registry)` plus `PrometheusMetrics.jvm(registry)`