NesyLink Documentation
This documentation is focused on using nesylink as a Gymnasium-compatible
reinforcement learning environment.
Recommended Reading Order
- guides/env-overview.md (open)
- Environment architecture, package boundaries, built-in tasks, and the
Gymnasium usage flow.
- guides/map-creation.md (open)
- How to create JSON maps with rooms, objects, exits, and validation rules.
- reference/game-content.md (open)
- Current player capabilities, items, monsters, map object meanings, exit
requirements, and built-in map content.
- reference/rewards.md (open)
- Built-in reward modules, reward signals, weight overrides, and custom
rewards.
- guides/training-config.md (open)
- Practical training configuration patterns for random rollouts, PPO-style
loops, and Dreamer-style wrappers.
- reference/env-api.md (open)
- Exact
reset, step, action, observation, and info contracts.
- reference/tasks-and-validators.md (open)
- Python task registry and task-level environment construction.
Documentation Scope
Keep docs centered on:
- creating environments
- understanding the architecture
- authoring maps and tasks
- configuring rewards
- training RL agents
- reading observations, events, and episode metadata
Avoid adding evaluation process notes, contributor workflow logs, or unrelated
project history here unless they directly help a user run or extend the
environment.