kyo-test

kyo-test is the Kyo project's own test framework. You write a suite by extending Test[S] and registering cases with two infix operators on string names. "name" - { ... } declares a GROUP whose body runs immediately at registration, so its nested -/in calls fire and build the tree (loops, helper defs, and conditionals inside a group all register their children). "name" in { ... } declares a LEAF, and the leaf body is the test: a Kyo computation of type Unit < (S & Async & Abort[Any] & Scope) that is captured, not run, until the runner executes it.

Because the body is an ordinary Kyo value, an asynchronous test and a synchronous one share the same shape: a bare () or assert(...) auto-lifts into the effectful type, with no Future or toFuture shim. Inside a leaf you assert with a single power-assert that renders a subexpression diagram on failure.

The type parameter S is an additive extra effect row. The baseline Async & Abort[Any] & Scope is always present and cannot be dropped; Test[Any] is the common case (baseline only), spelled explicitly because Scala 3 has no default type arguments. A suite that needs an extra effect (a database Env[Db] or a Var) declares it in S and discharges it with .handle at the leaf or group level, so the runner only ever sees a baseline-shaped leaf. Decorators chain off the name before the operator ("slow op".retry(3).timeout(5.seconds) in { ... }). Property-based and snapshot testing are opt-in by extending PropertyTest[S] or SnapshotTest[S] instead of Test[S]. The whole surface compiles and runs on JVM, JS, and Scala Native.

import kyo.*
import kyo.test.*

class BankTest extends Test[Any]:
    "a fresh account has zero balance" in {
        val account = Account(1, "ada", balanceCents = 0L)
        assert(account.balanceCents == 0L)
    }
end BankTest

A suite is a class; its body runs once at construction and registers every case. You point sbt at the module and the runner discovers the class, instantiates it, and executes each leaf.

Writing a suite

A suite is the unit you register and run. Extend Test[S], pick S for the extra effects your leaves need, and fill the class body with - groups and in leaves.

Test[S] and the baseline

Test[Any] is the suite you reach for first: its leaves use only the always-present baseline Async & Abort[Any] & Scope. You write Test[Any] explicitly because Scala 3 has no default type argument; the Any is the empty extra row, not "any effect".

The baseline is unioned into every leaf body regardless of S, so you cannot narrow it away. Abort[Any] (not Abort[Throwable]) is deliberate: a leaf may abort with any value, and the runner converts a non-Throwable abort into a LeafAborted failure at its boundary. Async lets a leaf suspend; Scope lets a leaf acquire resources that are released when the leaf ends.

When a leaf needs more (a database handle, mutable state, a seeded RNG), declare it in S and discharge it with .handle (see Per-test and per-suite setup).

Groups (-) and leaves (in): two operators, no inference

"name" - { body } is ALWAYS a group: the body runs at registration so its nested -/in calls fire and build the tree. "name" in { body } is ALWAYS a leaf: the body is the test and is deferred until the runner executes it. There is no compile-time inference; the operator you choose says exactly what the node is.

class TransfersTest extends Test[Any]:
    "transfers" - { // group: this block runs now, registering children
        "deposit increases the balance" in {
            val updated = Account(1, "ada", balanceCents = 0L).copy(balanceCents = 500L)
            assert(updated.balanceCents == 500L)
        }
        "a new account starts empty" in {
            assert(Account(2, "bob", 0L).balanceCents == 0L)
        }
    }
end TransfersTest

Because a group body is ordinary Scala that runs at registration, you can generate leaves with loops, helper defs, and conditionals; every in they reach registers a child:

class HostParseTest extends Test[Any]:
    "parse" - {
        for host <- List("a.com", "b.org") do
            s"accepts $host" in {
                assert(parse(s"https://$host/").isSuccess)
            }
    }
end HostParseTest

The leaf body is an ordinary Kyo value, so a synchronous and an asynchronous test have the same type. A bare () or assert(...) auto-lifts; an Async body composes with .map and is accepted through the baseline Async:

class AsyncBalanceTest extends Test[Any]:
    "balance settles after a sleep" in {
        Async.sleep(10.millis).andThen {
            assert(Account(1, "ada", 0L).balanceCents == 0L)
        }
    }
end AsyncBalanceTest

Assertions

Everything in this section is something you call inside a leaf body to make a claim or steer the leaf's outcome. The single power-assert is the workhorse; the rest cover expected exceptions, eventual consistency, preconditions, and compile-time checks.

assert: the power-assert

There is one assertion. assert(cond) throws AssertionFailed on false, carrying a diagram of every subexpression's value so you see why the condition was false without re-running under a debugger. assert(cond, msg) appends your message to that diagram.

class AccountAssertTest extends Test[Any]:
    "balance math" in {
        val a = Account(1, "ada", balanceCents = 250L)
        assert(a.balanceCents * 2 == 500L)
        assert(a.owner == "ada", "owner should survive the copy")
    }
end AccountAssertTest

fail, cancel, and assume: steering the outcome

fail(msg) ends the leaf as a failure unconditionally. The fail(cause: Throwable) overload preserves a caught exception (and its stack trace) as the failure's cause, for when an observed exception is itself the reason the test failed.

cancel(msg) ends the leaf as cancelled, not failed (throwing TestCancelled): cancellation says the test could not run (a missing dependency, the wrong platform), not that the system misbehaved. assume(cond, msg) is the conditional form: it cancels the leaf when cond is false and otherwise does nothing.

class PreconditionTest extends Test[Any]:
    "settlement requires a connection" in {
        val connected = false
        assume(connected, "no DB connection; skipping")
        fail("unreachable when the assumption holds")
    }
end PreconditionTest

When you need a conditional outcome, assert is the failure form and assume is the skip form: use assert(x) when a false x means a bug, and assume(x) when a false x means the test does not apply here.

intercept: expected exceptions

When the behavior under test is "this should throw", assert it with intercept[E]. It evaluates the body, asserts that it throws an E, and returns the caught exception so you can assert on its fields. interceptMessage[E](msg) additionally requires the thrown exception's message to equal msg.

class WithdrawTest extends Test[Any]:
    "overdraw is rejected" in {
        val e = intercept[InsufficientFunds] {
            throw InsufficientFunds(accountId = 7)
        }
        assert(e.accountId == 7)
    }
end WithdrawTest

interceptThrown[E] and interceptThrownMessage[E](msg) are the same assertions but return Unit; reach for them when you only care that the exception was thrown, not for its value.

assertEventually: eventual consistency

For state that becomes correct asynchronously (a balance that posts after settlement, a queue that drains), assertEventually(cond) re-evaluates cond every 10ms until it yields true. It returns a leaf body, so it is the last expression in the in block.

class SettlementTest extends Test[Any]:
    "posted balance settles" in {
        AtomicLong.init(0L).map { posted =>
            assertEventually(posted.get.map(_ == 0L))
        }
    }
end SettlementTest

typeCheck: compile-time assertions

To assert that a code string does or does not compile, use the compile-time checks. typeCheck(code) asserts it compiles; typeCheckFailure(code)(expected) asserts it does not compile and that the error contains expected; typeCheckFailure(code) asserts only that it fails. These pin API contracts that ordinary tests cannot reach, like "this illegal construction is rejected".

class TypeContractTest extends Test[Any]:
    "an Int does not satisfy a String parameter" in {
        typeCheckFailure("""def f(s: String): String = s; f(42)""")("Found")
    }
end TypeContractTest

Asserting inside a helper: AssertScope

Every assertion needs the leaf's AssertScope: the per-leaf object the runner uses to record failures (including those raised by a detached or leaked fiber) and to count how many assertions the leaf evaluated. The in operator supplies it implicitly, which is why the assert family compiles inside a leaf body but not in a - group, the suite class body, or a plain method.

To assert from a shared helper, give the helper a (using AssertScope) parameter so it joins the calling leaf's scope. Its assertions then count toward that leaf, so a leaf whose only checks live in such a helper still satisfies the no-assertion check.

import kyo.*
import kyo.test.*

class HelperAssertTest extends Test[Any]:
    def assertValidAccount(a: Account)(using AssertScope): Unit =
        assert(a.id > 0)
        assert(a.owner.nonEmpty)

    "a fresh account is valid" in {
        assertValidAccount(Account(1, "ada", balanceCents = 0L))
    }
end HelperAssertTest

Decorating tests

A decorator is metadata you chain onto the name before the -/in operator. It controls selection, retries, timing, lifecycle, and platform without changing the body. Each decorator produces a TestBuilder, the carrier the operator consumes; you rarely name TestBuilder yourself, you just chain methods and end with - or in.

class DecoratedTest extends Test[Any]:
    "slow settlement".retry(3).timeout(5.seconds) in {
        assert(Account(1, "ada", 0L).balanceCents == 0L)
    }
end DecoratedTest

Selection: focus, ignore, only, tagged, slow

.focus restricts the run to focused leaves only (everything else reports Skipped), the way you isolate one test while iterating. .ignore (optionally .ignore(reason)) marks a leaf as ignored: its body never runs and it reports Ignored, recording the optional reason; use it both to disable a test and to mark one whose body is not written yet. .only(cond) registers the leaf only when cond is true at registration; a false condition reports Skipped. .tagged("name", ...) attaches tags for filtering at run time; .slow is shorthand for .tagged("slow").

class SelectionTest extends Test[Any]:
    "wip".focus in { assert(1 == 1) }
    "broken".ignore in { assert(1 == 2) }
    "ci-only".only(sys.env.contains("CI")) in { assert(true) }
    "heavy".tagged("slow", "db") in { assert(true) }
end SelectionTest

Lifecycle: pendingUntilFixed

.pendingUntilFixed (optionally .pendingUntilFixed(reason)) DOES run the body and inverts the outcome: a still-failing body reports Pending, a now-passing body reports Failed. The failure is a tripwire telling you the underlying bug is fixed and the marker should be removed. To skip a test's body entirely, whether it is disabled or not yet written, reach for .ignore / .ignore(reason) from the selection group above.

class LifecycleTest extends Test[Any]:
    "blocked on issue #123".pendingUntilFixed in {
        assert(brokenFeatureWorks())
    }

    def brokenFeatureWorks(): Boolean = false
end LifecycleTest

Resilience and timing: timeout, retry, flaky, times

.timeout(d) bounds this leaf to d, overriding the suite default. .retry(n) retries up to n times with no backoff; .retry(schedule) retries on a Schedule you supply. .flaky is the curated form for known-flaky leaves: 3 retries with linear 100ms backoff, plus a "flaky" tag so CI can filter them. .times(n) repeats the body n times (every run must pass); .nonFlaky (or .nonFlaky(n)) is the readable alias for that and the inverse of .flaky: it runs the leaf 100 times (or n) and requires every run to pass, to prove a leaf is not flaky.

class ResilienceTest extends Test[Any]:
    "settles within a window".timeout(2.seconds) in {
        assert(true)
    }
    "occasionally races".flaky in {
        assert(true)
    }
    "retry on a custom policy".retry(Schedule.exponential(50.millis, factor = 2.0).take(4)) in {
        assert(true)
    }
end ResilienceTest

When a Passed or Failed result reports attempts > 1, a retry decorator rescued (or exhausted on) that leaf; attempts == 1 means it settled on the first try.

Platform: jvm/js/native/wasm, notX, onlyX

Three families select platforms. .jvm / .js / .native / .wasm add a platform to the include set (additive). .notJvm / .notJs / .notNative / .notWasm remove one. .onlyJvm / .onlyJs / .onlyNative / .onlyWasm restrict to exactly one. On a non-matching platform the leaf is compile-excluded: the body is never emitted, no code for the leaf reaches the platform's output, and the leaf is absent entirely. It produces no TestResult and is never reported. Filters compose: a second filter intersects with the first, so .notNative.notWasm excludes the leaf on both Native and WebAssembly while leaving it on the JVM and Scala.js.

class PlatformTest extends Test[Any]:
    "uses java.nio".onlyJvm in {
        assert(true)
    }
    "not on Scala Native".notNative in {
        assert(true)
    }
    "stack-unsafe on AOT targets".notNative.notWasm in {
        assert(true)
    }
end PlatformTest

.onlyJvm is semantically .jvm alone, but it reads unambiguously as single-platform intent; prefer it when you mean "exactly JVM" rather than "also JVM".

Per-test and per-suite setup

A leaf that uses an effect beyond the baseline (Env or Var) must discharge it before the runner sees the leaf. Two surfaces cover this: .handle discharges per leaf or per group, and aroundLeaf wraps every leaf in the suite.

.handle: discharge an extra effect

.handle[S1](h) chains off a name (or a decorated TestBuilder) and peels S1 down to baseline. The handler h is a result-PRESERVING polymorphic function with type [A] => (A < (S1 & Async & Abort[Any] & Scope)) => A < (Async & Abort[Any] & Scope): it takes a body that still needs S1 and returns one that needs only baseline, without changing the result value. The suite declares the extra effect in S so the leaf body can use it.

class DbLeafTest extends Test[Env[Db]]:
    "reads the balance".handle[Env[Db]](
        [A] => (body: A < (Env[Db] & Async & Abort[Any] & Scope)) => Env.run(Db("jdbc:test"))(body)
    ) in {
        Bank.balance(1).map(b => assert(b == 0L))
    }
end DbLeafTest

Chain .handle to peel several effects; each call peels one row down toward the row the previous call left:

class TwoEffectTest extends Test[Var[Int] & Env[Db]]:
    "uses Env then Var"
        .handle[Env[Db]]([A] => (b: A < (Env[Db] & Async & Abort[Any] & Scope)) => Env.run(Db("jdbc:test"))(b))
        .handle[Var[Int] & Env[Db]](
            [A] => (b: A < (Var[Int] & Env[Db] & Async & Abort[Any] & Scope)) => Var.run(0)(b)
        ) in {
        Env.get[Db].andThen(Var.update[Int](_ + 1)).unit
    }
end TwoEffectTest

The terminal operator rejects an under-discharged body at compile time: a body whose residual row still needs an effect beyond what you discharged is not a subtype of the baseline-shaped type the operator requires, and the compiler reports a lift failure naming the leaked effect.

kyo-Random is not a row effect: every operation suspends Sync, which the baseline already provides through Async, so you use it directly inside a leaf without declaring anything in S. For a deterministic RNG, wrap the body in Random.withSeed(seed) { ... } (it returns < Sync). The property runner's RNG (below) is a separate, pure seed and does not touch kyo-Random.

aroundLeaf and the suite hooks

When every leaf in a suite needs the same wrapper (open a connection, set a config), override aroundLeaf instead of repeating .handle at each in. It is the kyo-test equivalent of ScalaTest's withFixture: a single point that wraps each raw leaf body before the runner applies retry, timeout, and scope discharge.

class ConnectedSuite extends Test[Any]:
    override def aroundLeaf[A](body: A < (Async & Abort[Any] & Scope))(using Frame): A < (Async & Abort[Any] & Scope) =
        Sync.defer(()).andThen(body) // e.g. open and close a connection around each leaf

    "leaf runs inside the wrapper" in {
        assert(true)
    }
end ConnectedSuite

Other per-suite hooks you can override: name (the suite display name), timeout (the default per-leaf bound), randomize / randomSeed (shuffle leaf order), and config (the suite's own RunConfig, e.g. override def config = super.config.sequential).

Property-based testing

Property testing is opt-in: extend PropertyTest[S] instead of Test[S] and call forAll in the class body. forAll(gen) { value => ... } runs the body over numSamples samples (default 100) drawn from gen; on the first failing sample it shrinks toward a minimal counterexample before reporting.

import kyo.*
import kyo.test.*
import kyo.test.prop.*

class DepositPropertyTest extends PropertyTest[Any]:
    forAll(Gen.long) { amount =>
        val before = 0L
        val after  = before + amount - amount
        assert(after == before)
    }
end DepositPropertyTest

forAll comes in arities 1 through 4; the higher arities take that many generators and pass that many arguments:

import kyo.test.prop.*

class UrlPropertyTest extends PropertyTest[Any]:
    forAll(Gen.string, Gen.string) { (host, path) =>
        assert(Url("https", host, path).host == host)
    }
end UrlPropertyTest

Gen: generators with shrinking

A Gen[A] samples a value together with the full lazy tree of its shrink candidates, so the combinators carry shrinking automatically. The built-ins are const, int, long, double, string, boolean, list, listOfN, map, oneOf, and frequency. Combine them with map, flatMap, and filter, and derive a generator for any case class or sealed trait with Gen.derive[A].

The primitive generators bias toward edge values rather than drawing purely uniformly. Gen.int and Gen.long mostly draw uniformly within [-size, size] but occasionally inject the type boundaries Int.MinValue/Int.MaxValue and Long.MinValue/Long.MaxValue, which fall OUTSIDE [-size, size]; both still shrink toward 0. Gen.double injects 0.0, -0.0, 1.0, -1.0, NaN, the two infinities, and Double.MinValue/Double.MaxValue as edges, so a property over doubles that assumes finiteness must guard for NaN and the infinities itself. Gen.string, Gen.list, and Gen.map bias toward the empty and the singleton (or maximum-length) case. Gen.oneOf and Gen.frequency shrink toward earlier choices (lower index is treated as simpler), and Gen.derive[A] on a sealed trait shrinks toward earlier constructors.

For independent components, compose generators applicatively: Gen.zipWith(ga, gb)(f) combines two generators with f, and Gen.zip/Gen.zip3/Gen.zip4 return tuples of two, three, or four generators. Each component shrinks independently, so the counterexample minimizes component by component.

import kyo.*
import kyo.test.*
import kyo.test.prop.*

class ZipPropertyTest extends PropertyTest[Any]:
    val pair: Gen[(Int, Int)] = Gen.zipWith(Gen.int, Gen.int)((x, y) => (x, y))

    forAll(pair) { case (x, y) =>
        assert(x + y == y + x)
    }
end ZipPropertyTest

import kyo.*
import kyo.test.prop.*

class AccountPropertyTest extends PropertyTest[Any]:
    val accounts: Gen[Account] = Gen.derive[Account]

    forAll(accounts) { account =>
        assert(account.copy(owner = account.owner).owner == account.owner)
    }

    forAll(Gen.int.map(_.abs)) { n =>
        assert(n >= 0)
    }
end AccountPropertyTest

When forAll fails, it throws PropertyFailedException, carrying the original failing sample, the shrunk minimal counterexample, the underlying cause, and the seed so you can reproduce the exact run. Copy that seed into forAllSeeded(seed, gen) { body } (arities 1 through 4, mirroring forAll) to pin the failing run and replay it deterministically, regardless of the suite's randomize/nonRandomSeed.

import kyo.*
import kyo.test.*
import kyo.test.prop.*

class ReplayPropertyTest extends PropertyTest[Any]:
    // A run reported "Seed: 987654321"; pin it to replay the exact run.
    forAllSeeded(987654321L, Gen.list(Gen.int)) { xs =>
        assert(xs.sum == xs.foldLeft(0)(_ + _))
    }
end ReplayPropertyTest

Shrink: custom generators

When you build a Gen by hand and want the same shrinking the built-ins use, call the standalone algorithms in Shrink: Shrink.int, Shrink.long, Shrink.double, Shrink.string, and Shrink.list(chunk, elemShrink). These are the exact functions the built-in generators delegate to. Shrink.double drives toward 0.0 and toward integral values (2.7 shrinks to its integral neighbor 2.0 first, then halves toward 0.0); there is no lower floor, so it reaches 0.0 exactly. NaN shrinks to 0.0, and the infinities shrink through Double.MaxValue toward 0.0.

import kyo.Chunk
import kyo.test.prop.Shrink

val intShrinks: Chunk[Int]       = Shrink.int(100)    // 50, 25, 12, 6, 3, 1, 0
val doubleShrinks: Chunk[Double] = Shrink.double(2.7) // 2.0 (integral neighbor), then halving toward 0.0; no floor

Write meaningful tests: no-assertion enforcement

A leaf that runs to completion having evaluated zero assertions is itself a bug: it proves nothing. kyo-test enforces this at run time. After a leaf body joins, the runner counts how many assertions it evaluated; a leaf that would otherwise pass with a count of zero is flipped to Failed with the message leaf passed without evaluating any assertion. The whole assert family counts: assert, fail, intercept, assertEventually, typeCheck, assertSnapshot, and a property forAll. cancel and assume do not count, because they steer the outcome rather than make a claim.

This is a run-level setting, failOnNoAssertion, on by default. ScalaTest forced this at compile time through its Assertion return type; kyo-test cannot, because a leaf body is an effectful Unit < (S & Async & Abort[Any] & Scope), so the check runs after the body completes.

When a leaf genuinely has nothing to assert (a smoke test that only needs to confirm the body runs without error), write succeed (or succeed("note") to record why) as the per-leaf opt-out. It is an alias for assert(true): it flows through the assert runtime, counts as one evaluation, and always passes:

import kyo.*
import kyo.test.*

class SmokeTest extends Test[Any]:
    "pure smoke: the pipeline starts without error" in {
        succeed("nothing to claim; just confirm the body ran")
    }
end SmokeTest

To turn the check off for a whole suite (for a suite whose leaves assert through a domain helper that the runtime does not see), override config:

import kyo.*
import kyo.test.*

class HelperAssertSuite extends Test[Any]:
    override def config: RunConfig = super.config.failOnNoAssertion(false)

    "asserts through a helper the runtime cannot count" in {
        assert(true)
    }
end HelperAssertSuite

Snapshot testing

Snapshot testing is opt-in: extend SnapshotTest[S] and call assertSnapshot(actual, name). It renders actual via its Render[A] instance and compares the text against a stored file under snapshotDir (default "test-snapshots"), in a subdirectory named after the suite. A Render[A] is derived automatically for case classes, so a domain value snapshots without extra wiring.

import kyo.*
import kyo.test.*
import kyo.test.snapshot.*

class StatementSnapshotTest extends SnapshotTest[Any]:
    "monthly statement" in {
        val statement = Account(1, "ada", balanceCents = 12_300L)
        assertSnapshot(statement, "monthly")
    }
end StatementSnapshotTest

Running and selecting tests

You run a module's suites through sbt; the runner discovers every Test subclass, instantiates it, and executes its leaves.

# Run every suite in a module
sbt 'kyo-coreJVM/test'

# Run a single suite
sbt 'kyo-coreJVM/testOnly kyo.ChannelTest'

CLI flags

The command-line entry point (kyo.test.runner.Cli) takes flags that map onto the run configuration:

Exit codes are 0 (all passed, or nothing ran), 1 (a leaf failed, was cancelled, or timed out), and 2 (argument parse error).

RunConfig, TestFilter, Verbosity: the programmatic equivalents

RunConfig is the in-code form of the flags, with copy-style builders. RunConfig.default is the starting point; .sequential is shorthand for parallelism = 1.

import kyo.*
import kyo.test.*

val config: RunConfig =
    RunConfig.default
        .parallelism(4)
        .filter(TestFilter(pathInclude = Chunk("transfers.*"), tagsExclude = Set("slow")))
        .verbosity(Verbosity.Quiet)

TestFilter selects leaves: pathInclude/pathExclude are globs against the dot-joined leaf path, tagsInclude/tagsExclude are exact tag names. Includes act as allow-lists; excludes apply after. TestFilter.empty runs everything. Verbosity is Quiet / Normal / Verbose. Two RunConfig flags toggle discovery-only modes: countOnly reports the leaf count without running bodies, and listOnly additionally prints every leaf path (implying count-only). strictStructure turns on strict leaf-name-path validation, rejecting duplicate or structurally invalid paths at registration time.

Outcomes and counters

Each leaf produces one TestResult: Passed, Failed, Cancelled, Pending, Ignored, TimedOut, or Skipped. Passed and Failed carry an attempts count recording retry usage. The runner aggregates these into a TestReport (with a SuiteReport per suite), which exposes counters: totalLeaves, passed, failed, cancelled, pending, ignored, timedOut, skipped, and totalDuration.

Reporters

A TestReporter is a lifecycle-event sink (onRunStart, onSuiteStart, onLeafStart, onLeafComplete, onSuiteComplete, onRunComplete). The built-ins are ConsoleReporter (human-readable), TapReporter (TAP version 13), JUnitXmlReporter (JVM-only), and CombinedReporter (fan-out to several). Construct them through the Reporters factory:

import kyo.test.*
import kyo.test.runner.*

val reporter: TestReporter =
    Reporters.combined(
        Reporters.console(Verbosity.Normal),
        Reporters.tap(java.lang.System.out)
    )

The runner itself is TestRunner: runReport(suiteClass, config) yields the whole run as a Kyo value (TestReport < (Async & Abort[Throwable] & Scope)), and runToFuture(...) produces the single Future at the sbt edge. You rarely call either directly; sbt and the CLI invoke them for you.

Build wiring

A project enables kyo-test through the SbtKyoTestPlugin AutoPlugin, which adds the kyo-test-runner Test dependency and registers the framework with sbt:

// project/plugins.sbt
addSbtPlugin("io.getkyo" % "sbt-kyo-test-publish" % "<version>")

// build.sbt
lazy val myProject = project.enablePlugins(SbtKyoTestPlugin)