A Puyo Puyo environment for practice and reinforcement learning.

Mipukiyo is a Puyo Puyo implementation that aims to be controlled by turns and programmatically.

Turns allow a human to practice the game at their own pace, while the programmatic controls are intended to be used as a RL environment for research.

Currently, not deployed anywhere. Please run with npm.

npm install

npm start

• Human play by turn (to plan chains) is possible in Training mode.
• RL environment is not yet implemented.

Attempts to mimic a normal Puyo Puyo experience:

• Pieces fall at normal speed
• Soft drop enabled
• Full animations

Intended for policy training and evaluation, but currently usable for turn-based practice:

• Instant piece placement (no falling)
• Fast/minimal animations

Movement & Rotation:

• W/A/S/D - Move pieces
• Q/E - Rotate left/right

• Implement basic puyo functionality.
• Verify / Improve consistency with original Puyo Puyo rules

• Restart, scoring, saving setups

• Allow game to be played by discrete turns (no falling)

• Implement RL environment and baseline policy.

This project is licensed under the MIT License - see the LICENSE file for details.