Skip to content

Student Reading Guide

This page is the curated reading path through the documentation for students joining the project. Everything it links to lives in this docs site; read in the suggested order rather than alphabetically.

What's Here

docs/
├── index.md                  ← docs home
├── mission.md                ← why this project exists
├── student-guide.md          ← you are here
│
├── overview/                 ← scope, architecture, glossary, roadmap
├── concepts/                 ← gridworld, agents, observations, rewards,
│                                actions, wrappers, MARL and RL foundations
├── flows/                     ← init flow, step flow, training loop
├── algorithms/                ← deep dives: IQL, CQL/Mixed, DQN, AC, A2C, A3C
├── guides/                    ← quickstart, custom plug-ins, config recipes,
│                                using evaluate
├── tutorials/                 ← first-experiment walkthrough
├── specs/                     ← plug-in and algorithm contracts
├── reference/                 ← API reference, config reference, papers
└── api/                       ← auto-generated API pages (mkdocstrings)

Day 1 Reading Order

Read these in order before doing anything else.

  1. mission.md — what this project is, what it is not, and why it exists
  2. overview/architecture.md — the five layers (core, plugins, wrapper, baselines, orchestration) and what each one does
  3. overview/glossary.md — terms you will see throughout the code and docs
  4. guides/quickstart.md — install dependencies and run your first training job
  5. tutorials/first-experiment.md — a guided first experiment, end to end

Week-by-Week Reading Guide

Week Read Purpose
1 concepts/rl-foundations.md The RL vocabulary everything else assumes
1 concepts/gridworld.md · concepts/agents.md Understand what the env simulates
1 flows/init-flow.md · flows/step-flow.md Trace the code path before writing any
1 flows/training-loop.md Understand the Bellman update and epsilon decay
1–2 concepts/observations.md · concepts/rewards.md · concepts/actions.md Before touching IQL/CQL experiments
1–2 concepts/marl.md Single-agent MDP → Markov game transition
2–4 algorithms/index.md and the deep-dive for your algorithm Before implementing or modifying a baseline
2–4 specs/algorithm-spec.md The contract a new algorithm must satisfy
2–4 guides/config-recipes.md Setting up ablation experiments
2–4 guides/using-evaluate.md Collecting and interpreting metrics
Any guides/custom-observation.md When adding a new observation builder
Any guides/custom-reward.md When adding a new reward function
Any guides/custom-action.md When adding a new action space
Any concepts/wrappers.md When speed/stamina mechanics matter to your experiment
Any reference/config-reference.md When a YAML key is unclear
Any reference/api-reference.md When a method signature is unclear
Any reference/papers.md The literature behind each algorithm

I want to run the environment for the first time → guides/quickstart.md

I want to understand what my algorithm is receiving → concepts/observations.md → flows/step-flow.md

I want to understand why an agent gets a certain reward → concepts/rewards.md → flows/training-loop.md

I want to add a new observation builder → guides/custom-observation.md → specs/observation-builder-spec.md

I want to add a new reward function → guides/custom-reward.md → specs/reward-function-spec.md

I want to add a new action space → guides/custom-action.md → specs/action-space-spec.md

I want to implement a new learning algorithm (PPO, Nash Q, ...) → specs/algorithm-spec.md → flows/training-loop.md → reference/api-reference.md

I want to change experiment parameters → reference/config-reference.md → guides/config-recipes.md

I want to evaluate a trained policy → guides/using-evaluate.md

I want to understand the MARL theory behind what I am building → concepts/marl.md → overview/architecture.md

How to contribute what you build

When your reward function, observation builder, or experiment is ready to share, follow contributing.md and git-workflow.md: branch off STRP, keep core/ untouched, and open your PR against STRP.