Bevy AI Agent - AI-Powered Game Development Assistant

Bevy AI Agent is a comprehensive library and CLI tool that leverages cutting-edge AI models (GPT-4, Claude-3, Gemini) to accelerate Bevy game development through natural language code generation, intelligent feature addition, and automated code optimization.

Features

AI Model Integration

• OpenAI: GPT-4, GPT-4-Turbo, GPT-3.5-turbo
• Anthropic: Claude-3-opus, Claude-3-sonnet, Claude-3-haiku
• Google: Gemini-pro, Gemini-pro-vision

Dynamic Code Generation

• No more rigid templates - AI generates contextual, unique code
• Understands complex game concepts and mechanics
• Creates complete, working Bevy applications from scratch
• Context-aware feature additions that integrate seamlessly

Game Development Features

• Natural Language Game Creation: Describe your game in plain English
• Intelligent Feature Addition: Add complex systems with AI assistance
• Code Analysis & Optimization: AI-powered code review and improvements
• Smart Dependency Management: Automatically detects and manages Cargo dependencies
• Template System: Built-in templates for common game types
• Project Management: Track AI conversations and generated files

Installation

Prerequisites

Linux systems require additional audio development libraries:

# Ubuntu/Debian
sudo apt-get install libasound2-dev libudev-dev pkg-config

# Fedora/RHEL
sudo dnf install alsa-lib-devel systemd-devel pkgconf-pkg-config

# Arch Linux
sudo pacman -S alsa-lib systemd pkgconf

CLI Tool

cargo install bevy-agent

Library

Add to your Cargo.toml:

[dependencies]
bevy-agent = "0.1.0"
tokio = { version = "1.0", features = ["full"] }

Quick Start

1. Configure API Keys

# Configure OpenAI (recommended)
bevy-agent config --openai-key sk-...

# Or use Anthropic
bevy-agent config --anthropic-key ant-...

# Or use Google
bevy-agent config --google-key AIz...

# Set default model
bevy-agent config --default-model gpt-4

2. Create Your First AI-Generated Game

# Create a 2D platformer
bevy-agent create "2D platformer with physics and collectibles"

# Create a 3D space shooter
bevy-agent create "3D space shooter with asteroids and power-ups"

# Create a puzzle game
bevy-agent create "puzzle game with grid-based mechanics and level progression"

3. Add Features to Existing Games

cd your-game-project
bevy-agent add "inventory system with drag-and-drop UI"
bevy-agent add "magic system with spell combinations and cooldowns"
bevy-agent add "multiplayer support with networking"

4. Improve and Optimize Code

# Improve performance
bevy-agent improve performance

# Enhance readability
bevy-agent improve readability

# Add better structure
bevy-agent improve structure

Advanced Usage

Complex Game Creation

# Roguelike with procedural generation
bevy-agent create "roguelike dungeon crawler with procedural generation, turn-based combat, and character progression"

# Physics-based puzzle game
bevy-agent create "physics-based puzzle game like Portal with teleportation mechanics and environmental puzzles"

# Tower defense with AI
bevy-agent create "tower defense with pathfinding enemies, upgradeable towers, and resource management"

Model Selection

# Use specific models for different tasks
bevy-agent create "space game" --model gpt-4
bevy-agent add "multiplayer" --model claude-3-opus
bevy-agent improve performance --model gpt-4-turbo

Code Analysis and Debugging

# Get AI explanations of your codebase
bevy-agent explain

# Debug specific issues
bevy-agent debug "compilation error in movement system"

# Analyze specific files
bevy-agent explain --file src/player.rs

Project Management

# View project information
bevy-agent project info

# Show project statistics
bevy-agent project stats

# View AI conversation history
bevy-agent project history

# Clean up generated files
bevy-agent project clean

Build Operations

# Build your game
bevy-agent build build

# Run your game
bevy-agent build run

# Check for errors
bevy-agent build check

# Run clippy lints
bevy-agent build clippy

# Format code
bevy-agent build format

Template System

# List available templates
bevy-agent template list

# Show template details
bevy-agent template show platformer_2d

# Create game from template
bevy-agent create "my platformer" --template platformer_2d

Library Usage

Basic Usage

use bevy_agent::{BevyAIAgent, AIConfig, ModelType};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let config = AIConfig::from_env()?;
    let agent = BevyAIAgent::new(config).await?;
    
    let response = agent
        .generate_game("A simple 2D shooter")
        .with_model(ModelType::GPT4)
        .execute()
        .await?;
    
    println!("Generated code:\n{}", response.content);
    Ok(())
}

Advanced Project Management

use bevy_agent::{Project, BevyAIAgent, AIConfig};
use std::path::PathBuf;

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let config = AIConfig::load_or_create()?;
    let agent = BevyAIAgent::new(config).await?;
    
    // Create a new project
    let mut project = Project::init(
        PathBuf::from("my-game"),
        "My Awesome Game",
        "A revolutionary puzzle platformer",
        agent
    ).await?;
    
    // Generate initial game code
    let response = project.generate_game(
        "2D puzzle platformer with physics-based mechanics"
    ).await?;
    
    // Add features
    project.add_feature("inventory system with crafting").await?;
    project.add_feature("save/load system").await?;
    
    // Build the project
    project.manager().build().await?;
    
    Ok(())
}

Custom AI Requests

use bevy_agent::{BevyAIAgent, AIConfig, ModelType};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    let config = AIConfig::from_env()?;
    let agent = BevyAIAgent::new(config).await?;
    
    let response = agent
        .request("Create a component for enemy AI behavior")
        .with_model(ModelType::Claude3Opus)
        .with_temperature(0.8)
        .with_max_tokens(2000)
        .with_context("This is for a 2D tower defense game")
        .execute()
        .await?;
    
    println!("AI Response: {}", response.content);
    Ok(())
}

Real-World Examples

Metroidvania Game

bevy-agent create "metroidvania with ability-gated progression, interconnected world, and backtracking mechanics"

Generates:

• Interconnected world systems
• Power-up mechanics
• Ability-gated progression logic
• Map and exploration systems

Dialogue System

bevy-agent add "dialogue system with branching conversations, character portraits, and voice acting support"

Creates:

• NPC interaction systems
• Dialogue tree management
• UI components for conversations
• Audio integration hooks

A* Pathfinding

bevy-agent add "A* pathfinding for enemy AI with dynamic obstacle avoidance"

Implements:

• Pathfinding algorithms
• Navigation mesh generation
• Dynamic obstacle detection
• AI behavior integration

Configuration

Configuration File

Bevy AI stores configuration in ~/.bevy-agent-config.json:

{
  "openai": {
    "api_key": "sk-...",
    "organization": null,
    "base_url": null
  },
  "anthropic": {
    "api_key": "ant-...",
    "base_url": null
  },
  "google": {
    "api_key": "AIz...",
    "base_url": null
  },
  "default_model": "gpt-4",
  "generation": {
    "temperature": 0.7,
    "max_tokens": 4000,
    "include_comments": true,
    "generate_tests": false,
    "bevy_version": "0.12",
    "rust_edition": "2021"
  },
  "project": {
    "track_conversations": true,
    "auto_format": true,
    "auto_dependencies": true,
    "default_template": "basic"
  }
}

Environment Variables

You can also configure using environment variables:

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="ant-..."
export GOOGLE_API_KEY="AIz..."
export bevy_agent_DEFAULT_MODEL="gpt-4"

Project Configuration

Each Bevy AI project includes a .bevy-agent.json file that tracks:

• Project metadata
• AI conversation history
• Generated files
• Dependencies
• Custom templates

🔮 Key Advantages Over Templates

• Unlimited Creativity: Not constrained by predefined templates
• Context-Aware: Understands your existing codebase
• Learning: Benefits from the latest AI model improvements
• Explanatory: Can explain and teach as it generates code
• Adaptive: Handles edge cases and unique requirements

Architecture

Core Components

1. AI Integration Layer (src/ai/)

   • Multi-provider AI client (OpenAI, Anthropic, Google)
   • Model-specific prompt optimization
   • Response parsing and code extraction

2. Project Management (src/project.rs)

   • Project lifecycle management
   • Conversation history tracking
   • File generation monitoring

3. Template System (src/game_templates.rs)

   • Built-in game templates
   • Custom template support
   • Context-aware code generation

4. CLI Interface (src/cli.rs)

   • Command-line interface
   • User interaction handling
   • Progress reporting

5. Utilities (src/utils.rs)

   • Code analysis tools
   • Build system integration
   • File system operations

Supported AI Models

Provider	Models	Context Length	Strengths
OpenAI	GPT-4, GPT-4-Turbo, GPT-3.5-turbo	8K-128K	Code generation, debugging
Anthropic	Claude-3-opus, Claude-3-sonnet, Claude-3-haiku	200K	Long context, reasoning
Google	Gemini-pro, Gemini-pro-vision	32K	Multimodal, fast inference

Built-in Templates

• Basic Game: Simple 3D scene with camera and lighting
• 2D Platformer: Physics-based platformer with player movement
• 3D FPS: First-person shooter with mouse look and movement
• Puzzle Game: Grid-based puzzle mechanics
• Strategy Game: RTS-style unit management and resources

Contributing

We welcome contributions! Please see our Contributing Guide for details.

Development Setup

System Dependencies

Linux (Ubuntu/Debian):

sudo apt-get update
sudo apt-get install -y libasound2-dev libudev-dev pkg-config

Other Linux distributions:

• Fedora/RHEL: sudo dnf install alsa-lib-devel systemd-devel pkgconf-pkg-config
• Arch Linux: sudo pacman -S alsa-lib systemd pkgconf

macOS and Windows: No additional system dependencies required.

Building the Project

git clone https://github.com/jbuehler23/bevy-agent.git
cd bevy-agent
cargo build
cargo test

Running Examples

# Set up API keys
export OPENAI_API_KEY="your-key-here"

# Run CLI examples
cargo run -- create "simple platformer"
cargo run -- add "particle effects"
cargo run -- improve performance

📄 License

This project is licensed under either of

• Apache License, Version 2.0, (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

🙏 Acknowledgments

• The Bevy team for creating an amazing game engine
• OpenAI, Anthropic, and Google for providing powerful AI models
• The Rust community for excellent tooling and libraries

🔗 Links

• Documentation
• Crates.io
• Examples
• Issue Tracker
• Discussions

Status

This project is actively developed and maintained. We aim to support the latest versions of Bevy and provide cutting-edge AI integration for game development.

Current Status:

• Core AI integration ✓
• Multi-provider support ✓
• Project management ✓
• Template system ✓
• CLI interface ✓
• Web interface (planned)
• IDE plugins (planned)
• Asset generation (planned)