@getgrowly/agents

A modular agent system with multi-agent support for Growly applications, built on LangGraph.

Features

• 🛠️ Modular Tool System - Easily register and manage tools
• 🤖 Multi-Agent Support - Create complex agent workflows
• 🧩 Extensible Architecture - Add custom tools and agents
• 🔄 Event-Driven - Listen to agent events in real-time
• 🚀 Production Ready - Built with TypeScript and battle-tested patterns
• 📚 Resource Integration - Dynamic access to websites, contracts, documents, and text resources

Installation

npm install @getgrowly/agents
# or
yarn add @getgrowly/agents

Quick Start

Basic Usage

import { ChatOpenAI } from '@langchain/openai';

import { ReactAgent, ToolRegistry } from '@getgrowly/agents';

// Initialize tool registry
const toolRegistry = new ToolRegistry();

// Set up the language model
toolRegistry.setModel(
  new ChatOpenAI({
    modelName: 'gpt-4',
    temperature: 0,
  })
);

// Create an agent
const agent = new ReactAgent(
  {
    agentId: 'my-agent',
    systemPrompt: 'You are a helpful assistant',
    verbose: true,
  },
  toolRegistry
);

// Execute the agent
const result = await agent.execute('Hello, world!');
console.log(result.output);

Resource-Enabled Agent

import { createAgent } from '@getgrowly/agents';

// Create an agent with attached resources
const agent = await createAgent({
  agentId: 'my-agent',
  systemPrompt: 'You are a helpful assistant with access to resources',
  resources: [
    {
      id: 'resource-1',
      name: 'Uniswap Documentation',
      type: 'link',
      value: { url: 'https://docs.uniswap.org/' },
    },
    {
      id: 'resource-2',
      name: 'USDC Contract',
      type: 'contract',
      value: {
        address: '0xA0b86a33E6441b8c4C8C8C8C8C8C8C8C8C8C8C8C',
        network: 1,
        abi: [
          /* contract ABI */
        ],
      },
    },
  ],
  firecrawlService: firecrawlService, // For website content extraction
});

// Agent can now use resource tools:
// - get_resource_details
// - get_website_content
// - get_contract_abi
// - get_document_content

Multi-Agent System

import { MultiAgentSystem, ReactAgent } from '@getgrowly/agents';

// Create agents
const researcher = new ReactAgent(
  {
    agentId: 'researcher',
    systemPrompt: 'You are a research assistant',
  },
  toolRegistry
);

const writer = new ReactAgent(
  {
    agentId: 'writer',
    systemPrompt: 'You are a technical writer',
  },
  toolRegistry
);

// Create multi-agent system
const multiAgent = new MultiAgentSystem(
  {
    agents: {
      researcher: { role: 'Research Assistant', goal: 'Gather information' },
      writer: { role: 'Technical Writer', goal: 'Write clear documentation' },
    },
    workflow: {
      type: 'sequential',
      steps: ['researcher', 'writer'],
    },
  },
  toolRegistry
);

// Execute the workflow
const result = await multiAgent.execute('Research and write about AI agents');

Documentation

Core Concepts

For detailed information about the resource integration feature, see README_RESOURCES.md.

Agents

Agents are the core building blocks that process input and generate responses. The package provides a BaseAgent class that can be extended to create custom agents.

Tools

Tools are functions that agents can use to perform specific tasks. They can be registered with the ToolRegistry and made available to agents.

Multi-Agent Systems

Create complex workflows by connecting multiple agents together. The system supports different workflow types:

• Sequential: Linear flow from one agent to another
• Hierarchical: Controller agent delegates to specialist agents
• Collaborative: Fully connected agents that can communicate with each other

API Reference

See the API documentation for detailed information about all classes and methods.

Contributing

Contributions are welcome! Please read our contributing guidelines before submitting pull requests.

License

MIT