fsrs-kotlin

A tiny, dependency-free Kotlin Multiplatform implementation of FSRS-5 (Free Spaced Repetition Scheduler) — the modern, open scheduling algorithm behind Anki's FSRS and many spaced-repetition apps.

• Pure & side-effect-free. No I/O, no clock, no platform code — just the math. You pass in elapsed days and a rating; you get back the next memory state and interval.
• Multiplatform. JVM, Android, and iOS (arm64 + simulator) out of the box.
• Zero dependencies. commonMain pulls in nothing but the Kotlin stdlib.
• Tested against the algorithm's provable identities (e.g. R(S, S) = 0.9, interval at 0.9 retention = S), not hand-copied magic numbers.

// settings.gradle.kts → dependencyResolutionManagement { repositories { mavenCentral() } }
dependencies {
    implementation("ru.lorddarthart.fsrs:fsrs:0.1.0")
}

import ru.lorddarthart.fsrs.Fsrs
import ru.lorddarthart.fsrs.Rating

val fsrs = Fsrs() // default weights + 0.9 desired retention

// First review of a brand-new card:
val first = fsrs.review(current = null, rating = Rating.Good)
println(first.memory)        // MemoryState(stability=…, difficulty=…)
println(first.intervalDays)  // days until it's next due

// A later review, given how many days actually elapsed:
val next = fsrs.review(current = first.memory, rating = Rating.Again, elapsedDays = 12.0)

fsrs.retrievability(elapsedDays, stability) and fsrs.nextInterval(stability) are exposed too if you need the raw forgetting curve.

FSRS-5 with the power forgetting curve R(t, S) = (1 + FACTOR · t / S) ^ DECAY (DECAY = -0.5, FACTOR = 19/81, so stability S is by definition the number of days until retrievability decays to 90%). Difficulty uses FSRS-5 linear damping and mean reversion; stability grows on recall and shrinks on lapse.

Apache License 2.0.