Getting Started

Prerequisites

Before you begin, ensure you have the following installed:

Bun (opens in a new tab)
Basic understanding of JSX

Installation

To get started with micro-agi, you need to install the package in your project.

bun add micro-agi

micro-agi works seamlessly with AI.JSX (opens in a new tab), a framework for building AI applications using JavaScript and JSX. To take full advantage of micro-agi, make sure you have AI.JSX (opens in a new tab) installed in your project.

Creating an AI Agent Team

With micro-agi, you can create a team of AI agents, each with specific roles and tasks.

The following example is taken from micro-agi-starter (opens in a new tab) repo. Feel free to clone it and run it against Ollama locally.

Here's a simple example to get you started:

/** @jsxImportSource ai-jsx */
import * as AI from "ai-jsx";
import Agent from "micro-agi/core/components/agent";
import Task from "micro-agi/core/components/task";
import Team from "micro-agi/core/components/team";
 
const App = async ({ topic }: { topic: string }) => {
  return (
    <Team process="sequential">
      <Agent
        agentType="mrkl"
        role="Writer"
        goal="Write articles about a topic"
        backstory="You are a very experienced writer. You've written thousands of article in your career."
        model="mistral"
        provider="ollama"
      >
        <Task
          onStart={async () => {
            console.log("Started writing article about", topic);
          }}
          onDone={async () => {
            console.log("Done writing article about", topic);
          }}
        >
          Write an article about {topic}. Your result in markdown format.
        </Task>
      </Agent>
    </Team>
  );
};
 
const renderContext = AI.createRenderContext();
const result = await renderContext.render(<App topic="Apple" />);
await Bun.write(`./result.json`, result);

Introduction Features