SerpentAI is a Python framework for creating AI agents that learn to play video games, targeting AI researchers and hobbyists. It provides a flexible, plugin-based sandbox for experimenting with various AI techniques, from deep reinforcement learning to computer vision, without requiring specific game APIs or proprietary environments.

How It Works

SerpentAI acts as a bridge between a game and an AI agent, capturing game state (e.g., screen pixels, memory) and translating agent actions (e.g., key presses) into game inputs. Its plugin architecture allows for easy integration of new games and agent algorithms, promoting portability and community sharing. The framework emphasizes native execution and user-provided games, avoiding dependencies on Docker or VNC.

Quick Start & Requirements

• Install via pip install serpentai.
• Requires Python 3.8+.
• Supports Linux and Windows. macOS support is deprecated.
• Official documentation is available on the GitHub Wiki.

Highlighted Details

• Supports diverse AI approaches: deep reinforcement learning, computer vision, image processing, trigonometry.
• Plugin-based design for game support and agents, enabling portability and distribution.
• Aims for ease of use with an installer and GUI in future versions.
• Revival in May 2020 targets Python 3.8+, reduced dependencies, and improved usability.

Maintenance & Community

• Development resumed in May 2020.
• Business contact: info@serpentai.ai.

Licensing & Compatibility

• Permissive license (details not specified in README, but implied to be suitable for commercial use and closed-source linking).

Limitations & Caveats

• macOS support is officially deprecated due to reliance on CUDA and Apple's hardware/software ecosystem decisions.
• The framework is described as "simple yet powerful," but the extent of its capabilities and stability for complex games is not detailed.