AI SDK Agents

A powerful and flexible library for building AI agents using the Vercel AI SDK. This project provides a simple yet powerful way to create and manage AI agents with customizable flows, tools, and contexts.

Features

• 🤖 Simple agent creation and management
• 🔄 Customizable agent flows
• 🛠️ Extensible tool system
• 📝 Context management
• 🔄 Streaming support

Installation

npm install ai-sdk-agents
# or
yarn add ai-sdk-agents
# or
pnpm add ai-sdk-agents

Prerequisites

This package requires the Vercel AI SDK and zod as a peer dependency:

npm install ai zod
# or
yarn add ai zod
# or
pnpm add ai zod

Usage

Here's a basic example of how to use the library:

import { agent, ChatFlow, Context } from 'ai-sdk-agents';
import { openai } from '@ai-sdk/openai';

export type PlannerContext = Context & {
  today: string;
};

const sqlAgent = agent({
  model: openai('gpt-4o'),
  system: (ctx: PlannerContext) => plannerPrompt(ctx.today),
  description: 'This agent can help you write SQL queries',
  tools: { sqlExecutor: sqlExecutorTool },
  asTool: {
    input: z.object({
      question: z.string().describe('The user query to analyze'),
    }),
    getPrompt: ({ question }) => ({
      prompt: `question: ${question}.
      - Make sure you return no more than 200 rows of data.`,
    }),
  },
});

export const routerAgent = agent({
  system: (ctx: InsightContext) => routerPrompt(ctx),
  model: openai('gpt-4o'),
  maxSteps: 5,
  tools: { sqlAgent },
  toolChoice: 'auto',
});

const flow = new ChatFlow({ agent: routerAgent });
const ctx = { history: [], today: new Date().toISOString().split('T')[0] };
const { result } = await flow.run(ctx);

Project Structure

src/
├── agent.ts           # Core agent implementation
├── context.ts         # Context management
├── flow.ts             # Flow control and management
├── flows/              # Predefined flow implementations
│   └── chat-flow.ts    # Chat flow implementation
├── index.ts           # Main entry point
├── tools.ts           # Tool definitions
├── types.ts           # Type definitions
├── utils.ts           # Utility functions
└── prompts.ts         # Prompt handling

Development

Setup

1. Clone the repository
2. Install dependencies:

   yarn install

Available Scripts

• yarn test - Run tests using Vitest
• yarn lint - Run ESLint with auto-fix
• yarn format - Format code using Prettier

Tests

Comprehensive tests are located in the tests/ directory.

Examples

Example projects are located in the examples/ directory. (Note: If the directory is currently named exmaples/, please rename it to examples/.)

Contributing

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

License

This project is licensed under the MIT License - see the LICENSE.md file for details.

Acknowledgments

• Built on top of the Vercel AI SDK
• Heavily inspired by (openai agent framework)[https://github.com/openai/openai-agents-python]
• Uses Vitest for testing

TODO

• Add more tests
• Add more examples
• Add more documentation
• Inter-Agent memory