Core Concepts

Agents

Agents are the primary units of work in Zod. An agent is an AI model configured with specific skills, tools, and behavior.

import { Agent } from 'zod-framework';

const agent = new Agent({
  name: 'MyAgent',
  model: 'gpt-4',
  skills: ['conversation'],
  tools: ['filesystem'],
});

Agent Lifecycle

Agents go through a continuous loop:

1. Receive: The agent receives a message or trigger
2. Process: The agent processes the input using its model and skills
3. Act: The agent may use tools to perform actions
4. Respond: The agent generates a response
5. Wait: The agent waits for the next input

Skills

Skills define what an agent can do. They are modular capabilities that can be mixed and matched.

Tools

Tools are external capabilities that agents can use to interact with the world.

import { Tool } from 'zod-framework';

class CustomTool extends Tool {
  name = 'my-tool';
  description = 'A custom tool for specific tasks';
  
  async execute(input) {
    // Perform the tool's task
    return result;
  }
}

Built-in Tools

Zod ships with several useful tools:

• filesystem: Read and write files
• web-search: Search the internet
• database: Query databases
• terminal: Execute shell commands
• http: Make HTTP requests

Prompts

Prompts control how agents behave and respond. Zod supports several types:

System Prompts

Set the overall behavior for an agent:

const agent = new Agent({
  name: 'CodeReviewer',
  systemPrompt: `You are a code reviewer. Look for bugs, security issues, 
    and code quality problems. Be specific and provide actionable feedback.`,
});

Task Prompts

Define specific tasks for agents to perform:

await agent.task('Review the code in src/auth.js and provide feedback');

Message Prompts

Send conversational messages:

await agent.message('What are the security implications of this code?');

The Harness

The harness is the orchestration layer that manages agents, their lifecycle, and communication.

import { Harness } from 'zod-framework';

const harness = new Harness({
  maxAgents: 10,
  timeout: 30000,
  logLevel: 'info',
});

harness.register(agent);
harness.start();

Configuration

Zod can be configured through zod.config.js:

export default {
  agents: {
    // Agent configurations
  },
  harness: {
    // Harness settings
  },
  plugins: {
    // Plugin configurations
  },
};

Understanding these core concepts is essential for building effective multi-agent systems with Zod. Continue to the Agents guide to learn more about creating and managing agents.