sneks-engine 1.0.4

Sneks engine

Sneks Engine

A Python game engine for creating snake-like games with customizable rules, graphics, and AI submissions.

Architecture

src/sneks/
├── engine/           # Core engine components
│   ├── core/         # Primitives: Cell, Bearing, Direction, Action
│   ├── entity/       # Game entities: Snek, Food, Static/Dynamic
│   ├── config/       # Configuration dataclasses
│   ├── manager/      # Cell and bearing calculations
│   ├── validator/    # Submission validation tests
│   ├── world.py      # Base World class (game loop)
│   ├── graphics.py   # Pygame rendering
│   ├── recorder.py   # Frame capture and video encoding
│   ├── runner.py     # Game execution and scoring
│   └── scoring.py    # Score normalization
└── games/            # Game variants
    ├── classic/      # Traditional snek
    ├── space/        # Asteroids + snek
    ├── toroidal/     # Infinite growth mode
    └── example/      # Example submission

Key Components

• World: Manages the game loop, entities, and board state. Subclass this to create new game modes.
• Entity: Base class for all game objects. DynamicEntity moves and grows; StaticEntity maintains shape.
• Snek: The AI interface that submissions implement. Receives game state and returns an Action.
• Config: Singleton configuration system. Each game mode extends Config with custom settings.

Creating a New Game Mode

1. Create a config class extending Config:

from dataclasses import dataclass
from sneks.engine.config.base import Config


@dataclass(frozen=True)
class MyGameConfig(Config):
    title: str = "My Game"
    turn_limit: int = 5000

2. Create a world class extending World:

from sneks.engine.world import World
from sneks.engine.scoring import Criteria


class MyGame(World):
    criteria = Criteria.AGE | Criteria.LENGTH  # Scoring criteria

    def __init__(self, config: MyGameConfig) -> None:
        super().__init__(config=config)

    # Override hooks as needed:
    # - load_entities(): Add custom entities
    # - before_step(): Pre-step logic
    # - after_step(): Post-step logic (collisions, growth)

Core Concepts

Bearing vs Direction vs Action

The engine uses a velocity-based movement model:

• Bearing: Current velocity as Bearing(x, y) — cells moved per tick in each axis. A snek at Bearing(2, -1) moves 2 cells right and 1 cell down each tick.
• Direction: Cardinal directions (UP, DOWN, LEFT, RIGHT) used for querying neighbors and obstacles.
• Action: Acceleration input returned by get_next_action(). Actions modify the bearing — Action.UP increments bearing.y by 1.

Sneks start stationary at Bearing(0, 0). Returning Action.RIGHT twice results in Bearing(2, 0), moving 2 cells right per tick.

Toroidal Board

The game board wraps — moving off the right edge appears on the left. This affects:

• Cell neighbors (e.g., Cell(0, 0).get_left() wraps to the rightmost column)
• Distance calculations (shortest path may cross edges)
• Collision detection

Relative Coordinates

Submissions receive positions relative to their head, which is always Cell(0, 0):

• self.occupied contains obstacles offset from the head
• self.food contains food positions offset from the head
• self.body[0] is always Cell(0, 0) (the head)

This simplifies AI logic — you don't need to know absolute board position.

Vision Range

Sneks can only see obstacles within Config().snek.vision_range cells. Food is always visible regardless of distance.

Recording

Install the recording dependency:

pip install -e ".[record]"

Enable recording in your config:

from sneks.engine.config.graphics import GraphicsConfig

MyGameConfig(
    graphics=GraphicsConfig(
        record=True,
        record_prefix="./output",  # Output directory
        record_fps=12,             # Video framerate
    )
)

Frames are saved to output/pics/ during gameplay. Call recorder.animate_game() after the run to encode to MP4 in output/movies/. The encoder uses ffmpeg via imageio_ffmpeg.