Skip to content

AGENTS.md

Latest commit

 

History

History
71 lines (54 loc) · 2.56 KB

AGENTS.md

File metadata and controls

71 lines (54 loc) · 2.56 KB

AGENTS.md

Identity

You are an agent working on linear-gp — a Rust framework for solving reinforcement learning and classification tasks using Linear Genetic Programming (LGP). Supports CartPole, MountainCar, and Iris classification with optional Q-Learning integration.

Architecture

Pure Rust workspace:

Package Path Role
lgp crates/lgp Core library — traits, evolutionary engine, built-in problems
lgp crates/lgp-cli CLI binary for running experiments, hyperparameter search, and analysis

Core Traits

Trait Purpose
State Environment state with value access and action execution
RlState Extends State for RL environments with terminal detection
Core Main trait defining genetic algorithm components
Fitness Evaluates individual performance
Breed Two-point crossover for offspring
Mutate Mutation operators for genetic variation

Key Files

  • crates/lgp-cli/src/main.rs — CLI entry point
  • crates/lgp/src/lib.rs — Core library exports
  • crates/lgp/src/core/ — Core LGP implementation
  • crates/lgp/src/problems/ — Problem implementations (CartPole, MountainCar, Iris)
  • configs/ — TOML experiment configurations (default.toml, optimal.toml)
  • outputs/ — Experiment results (parameters, tables, figures)
  • skills/lgp-experiment/SKILL.md — Guide for running experiments and adding new environments

Commands

Task Command
Install cargo install --path crates/lgp-cli
Build cargo build
Build (with plots) cargo build --features plot
Test cargo test
Bench cargo bench
Lint cargo clippy -- -D warnings
Format cargo fmt
Run experiment lgp run <name> (e.g., lgp run cart_pole_lgp)
List experiments lgp list
Hyperparameter search lgp search <config>
Analyze results lgp analyze
Full pipeline lgp experiment <config> (search -> run -> analyze)

Code Style

  • Rust 2021 edition, Apache-2.0 license
  • cargo fmt and cargo clippy -- -D warnings
  • Structured logging via tracing (RUST_LOG=lgp=debug)
  • Parallel evaluation via rayon

Supported Environments

Environment Type Inputs Actions
CartPole RL 4 2
MountainCar RL 2 3
Iris Classification 4 3

Adding a New Environment

See skills/lgp-experiment/SKILL.md. Implement the State (or RlState) trait, create a TOML config under configs/, and wire it into the CLI.