Gwent RL PettingZoo API¶

This site documents the PettingZoo-facing reinforcement-learning API for the standalone Gwent research environment. It covers both public AEC wrappers:

• gwent.rl.pettingzoo: a match-only two-agent AEC environment.
• gwent.rl.pettingzoo_deckbuilding: a meta AEC environment where each agent can choose a baseline deck or build a deck before the match begins.

The wrappers sit outside the dependency-free gwent.rl action and observation helpers. Importing the PettingZoo modules is the point where PettingZoo, Gymnasium, and NumPy become runtime dependencies.

Install And Build¶

Use the repository's uv environment:

uv sync
uv run zensical serve
uv run zensical build --strict --clean

Generated reference pages are rendered from the live Python wrappers:

uv run python tools/render_pettingzoo_docs.py
uv run python tools/render_pettingzoo_docs.py --check

API Map¶

External API References¶

The docs here follow the current PettingZoo AEC shape documented by Farama:

• PettingZoo AEC API
• PettingZoo environment tests

The site itself is built with Zensical:

• Zensical setup basics
• Zensical build command