# Creating AI Agents with Eliza

---
Table of Contents:

[TOC]

## Introduction

_Duration: 2-3 minutes_

Eliza is a powerful multi-agent simulation framework designed to create, deploy, and manage autonomous AI agents. Built with TypeScript, it provides a flexible and extensible platform for developing intelligent agents that can interact across multiple platforms while maintaining consistent personalities and knowledge.

### Key Features

- Multi-Agent Architecture
- Character System
- Memory Management
- Platform Integration
- Plugin System

## Setup & Prerequisites

_Duration: 3-4 minutes_

### Required Software

- Node.js 23+
- pnpm
- Python 2.7+ (if needed)
- WSL 2 (for Windows users)

### Initial Setup

Clone the starter repository

git clone https://github.com/ai16z/eliza-starter.git

Install dependencies and start

pnpm i && pnpm build && pnpm start

## Creating Your First Agent

_Duration: 8-10 minutes_

### Character Configuration

1. Create a new character file
2. Define basic personality traits
3. Set up model preferences
4. Configure platform connections

### Basic Configuration Example

const myAgent = {
    name: "MyFirstAgent",
    personality: {
        traits: ["helpful", "friendly", "knowledgeable"],
        background: "An AI assistant focused on helping with technical tasks"
    },
    modelConfig: {
        provider: "OPENAI",
        model: "gpt-4"
    }
};

## Understanding Eliza Plugins

_Duration: 3-4 minutes_

### What are Plugins?

Plugins are modular extensions that add new capabilities to your Eliza agent. Each plugin consists of three main components:

1. **Actions**: Specific tasks the agent can perform
2. **Evaluators**: Logic for assessing responses and behavior
3. **Providers**: Sources of information or services

### Plugin Architecture

interface Plugin {
    name: string;           // Unique identifier for the plugin
    description: string;    // Brief description of functionality
    actions?: Action[];     // Custom actions provided by plugin
    evaluators?: Evaluator[]; // Response evaluation logic
    providers?: Provider[]; // Information/service providers
}

### How Plugins Work

1. **Registration**: Plugins are registered when configuring your agent
2. **Discovery**: The agent automatically discovers available actions
3. **Execution**: Actions are triggered based on context and conversation
4. **Response**: Results are processed and returned to the conversation

### Common Plugin Types

- **Integration Plugins**: Connect to external services (APIs, databases)
- **Capability Plugins**: Add new functionalities (web search, image generation)
- **Utility Plugins**: Provide helper functions and tools
- **Platform Plugins**: Enable platform-specific features

### Plugin Development Flow

1. Create plugin structure
2. Define actions/evaluators/providers
3. Build and test locally
4. Package and distribute

For example, here's a basic plugin structure:

// Basic plugin template
const myPlugin: Plugin = {
    name: "my-plugin",
    description: "Adds custom functionality",
    actions: [
        {
            name: "MY_ACTION",
            description: "Performs a specific task",
            handler: async (runtime, message) => {
                // Implementation
            }
        }
    ]
};

## Adding Plugins - Web Search Example

_Duration: 10-12 minutes_

### Plugin Installation

pnpm add @ai16z/plugin-web-search

### Plugin Configuration

const myAgent = {
    plugins: ["@ai16z/plugin-web-search"]
};

### Web Search Plugin Structure

import { Action, Plugin } from "@ai16z/eliza";

const webSearch: Action = {
    name: "WEB_SEARCH",
    similes: [
        "SEARCH_WEB",
        "INTERNET_SEARCH",
        "LOOKUP",
        "QUERY_WEB"
    ],
    description: "Perform a web search to find information related to the message.",
    validate: async (runtime, message) => {
        return !!runtime.getSetting("TAVILY_API_KEY");
    },
    handler: async (runtime, message, state, options, callback) => {
        // Implementation details
    }
};
export const webSearchPlugin: Plugin = {
    name: "webSearch",
    description: "Search web",
    actions: [webSearch],
    evaluators: [],
    providers: []
};

### Configuration Steps

1. Add API keys to .env file:

TAVILY_API_KEY=your_api_key_here

2. Register plugin in your agent configuration:

import { webSearchPlugin } from "@ai16z/plugin-web-search";
const agentConfig = {
// ... other config
plugins: [webSearchPlugin]
};

### Testing Web Search

Example interactions:

// Example search queries
[
    {
        user: "user1",
        content: {
            text: "Find the latest news about SpaceX launches."
        }
    },
    {
        user: "agentName",
        content: {
            text: "Here is the latest news about SpaceX launches:",
            action: "WEB_SEARCH"
    }
    }
]

## Advanced Features

_Duration: 5-6 minutes_

### Memory Management

- Conversation history
- Knowledge base
- Context awareness

### Multi-platform Support

- Discord integration
- Twitter integration
- Telegram support

### Custom Actions and Providers

const customAction: Action = {
    name: "CUSTOM_ACTION",
    description: "Description of your custom action",
    handler: async (runtime, message) => {
        // Implementation
    }
};

## Best Practices & Tips

_Duration: 3-4 minutes_

### Character File Organization

- Keep configurations modular
- Use clear naming conventions
- Separate concerns appropriately

### Plugin Development Guidelines

- Follow TypeScript best practices
- Implement proper error handling
- Document your code thoroughly

### Security Considerations

- Protect API keys
- Validate user inputs
- Monitor resource usage

## Conclusion & Next Steps

_Duration: 2-3 minutes_

### Resources

- [Official Documentation](https://ai16z.github.io/eliza/)
- [GitHub Repository](https://github.com/ai16z/eliza)
- [Discord Community](https://discord.gg/ai16z)

### Further Learning

- Advanced plugin development
- Custom model integration
- Platform-specific optimizations

## Video Timestamps

- 0:00 - Introduction
- 2:30 - Setup & Prerequisites
- 5:30 - Creating Your First Agent
- 13:30 - Adding Plugins
- 23:30 - Advanced Features
- 28:30 - Best Practices
- 31:30 - Conclusion

## Troubleshooting Common Issues

1. Installation Problems
- Check Node.js version
- Verify pnpm installation
- Confirm environment variables

2. Plugin Issues
- Verify API keys
- Check plugin compatibility
- Review error logs

3. Runtime Errors
- Monitor memory usage
- Check network connectivity
- Validate configuration files