Google Agent Development Kit (ADK) - TypeScript

A TypeScript implementation of Google's Agent Development Kit (ADK), providing a comprehensive framework for building AI agents. This SDK mirrors the structure and functionality of the official Python ADK while providing TypeScript-native features and type safety.

Features

• 🤖 Multiple Agent Types: LlmAgent, SequentialAgent, ParallelAgent, LoopAgent, and RemoteAgent
• 🔧 Tool Support: Built-in tools and custom tool creation
• 🧠 Multi-Model Support: Google Gemini and OpenAI models
• 💾 State Management: Session and memory services
• 🔐 Authentication: OAuth and API key support
• 📝 Type Safety: Full TypeScript support with comprehensive type definitions

Installation

npm install @waldzellai/adk-typescript

Usage

Basic Agent Usage

import { LlmAgent, RunConfig } from '@waldzellai/adk-typescript';
import { Runner } from '@waldzellai/adk-typescript';

// Create an agent
const agent = new LlmAgent({
  name: 'my-agent',
  description: 'A helpful assistant',
  model: 'gemini-1.5-pro',
  instruction: 'You are a helpful assistant.'
});

// Create a runner
const runner = new Runner({
  appName: 'my-app',
  agent: agent
});

// Run the agent
const userMessage = {
  role: 'user',
  parts: [{ text: 'Hello, agent!' }]
};

for await (const event of runner.runAsync('user-123', 'session-123', userMessage)) {
  console.log(`${event.author}: ${event.getContent()?.parts[0].text}`);
}

Agent Types

The ADK supports several types of agents:

1. LlmAgent: The standard agent that uses an LLM to generate responses.
2. SequentialAgent: Runs its sub-agents in sequence.
3. ParallelAgent: Runs its sub-agents in parallel.
4. LoopAgent: Runs its sub-agents in a loop until a condition is met.
5. RemoteAgent: Communicates with a remote endpoint to get responses.

Using OpenAI Models

This implementation supports OpenAI models like GPT-4 and o4-mini:

import { LlmAgent } from '@waldzellai/adk-typescript';
import { Runner } from '@waldzellai/adk-typescript';
import { OpenAI } from '@waldzellai/adk-typescript';

// Create an OpenAI model
const openaiModel = new OpenAI({
  model: 'o4-mini-high',
  apiKey: 'your-openai-api-key'
});

// Create an agent using the OpenAI model
const agent = new LlmAgent({
  name: 'openai-assistant',
  description: 'A helpful assistant powered by OpenAI',
  model: openaiModel,
  instruction: 'You are a helpful assistant that provides concise and accurate information.'
});

// Create a runner
const runner = new Runner({
  appName: 'openai-example',
  agent
});

// Run the agent
const userMessage = {
  role: 'user',
  parts: [{ text: 'Hello, agent!' }]
};

for await (const event of runner.runAsync('user-123', 'session-123', userMessage)) {
  console.log(`${event.author}: ${event.getContent()?.parts[0].text}`);
}

Using RemoteAgent

RemoteAgent allows you to communicate with a remote endpoint to get responses:

import { RemoteAgent, LlmAgent } from '@waldzellai/adk-typescript';
import { Runner } from '@waldzellai/adk-typescript';

// Create a remote agent
const remoteAgent = new RemoteAgent({
  name: 'remote-agent',
  description: 'A remote agent',
  url: 'http://example.com/agent'
});

// Create a parent agent that uses the remote agent
const rootAgent = new LlmAgent({
  name: 'root-agent',
  description: 'A parent agent',
  model: 'gemini-1.5-pro',
  instruction: 'You are a helpful assistant.',
  subAgents: [remoteAgent]
});

// Create a runner
const runner = new Runner({
  appName: 'my-app',
  agent: rootAgent
});

// Run the agent
const userMessage = {
  role: 'user',
  parts: [{ text: 'Hello, agent!' }]
};

for await (const event of runner.runAsync('user-123', 'session-123', userMessage)) {
  console.log(`${event.author}: ${event.getContent()?.parts[0].text}`);
}

Development

Prerequisites

• Node.js >= 16.0.0
• npm or yarn

Building

npm install
npm run build

Testing

npm test

Linting

npm run lint
npm run lint:fix  # Auto-fix linting issues

API Documentation

For detailed API documentation, please refer to the TypeScript definitions in the dist folder after building, or explore the source code in the src directory.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit your changes (git commit -m 'Add some AmazingFeature')
4. Push to the branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Acknowledgments

• Based on Google's Agent Development Kit (ADK) Python SDK
• TypeScript implementation by WaldzellAI