A monadic framework for reinforcement learning built on top of dotty-cps-async.

import cps.rl.*

• Scored Logic Monad - Monadic abstraction for weighted search with scores (CpsScoredLogicMonad)
• RL Abstractions - Core traits for environments, agents, and models (RLEnvironment, RLAgentBehavior, RLModelControl)
• Tensor Support - Type-safe tensor operations with scope management (TensorType, TensorScope, TensorRepresentation)
• Priority Queue Data Structures - Efficient heap implementations including PairingHeap, FingerTree, and scaled variants
• Cross-Platform - Supports JVM, Scala.js, and Scala Native

Add to your build.sbt:

libraryDependencies += "io.github.dotty-cps-async" %%% "rl-logic" % "<version>"

• Scala 3.3.7+
• dotty-cps-async 1.3.0
• DJL (Deep Java Library) for neural network support (JVM only)

Apache 2.0