Used to check for browser translation.
用于检测浏览器翻译。
ブラウザの翻訳を検出する

🚀 Quick Start

The AIGNE Framework lets you build powerful AI Agents and workflows with minimal code. Follow the steps below to experience AI magic right away and enjoy the fun of "write a little, achieve a lot"! ✨

Environment Requirements#

Before starting the installation, please ensure your local environment meets the following conditions:

Installation 🛠️#

First step, install dependencies! With just one command, you'll instantly have all the capabilities to build AI Agents. Supports npm/yarn/pnpm, choose whichever you prefer.

The AIGNE Framework depends on the core package @aigne/core and model packages (such as @aigne/openai). You can choose the appropriate installation command based on your package manager. After installation, you can start building your own Agents.

npm install @aigne/core @aigne/openai

You can also use yarn or pnpm:

yarn add @aigne/core @aigne/openai
# or
pnpm add @aigne/core @aigne/openai

Create Your First Agent 🎉#

Next, we'll build a simple but fully functional AI Agent step by step. Each step below is an important part of the building process.

Import Necessary Modules#

First, we need to import the core components and model implementations of the AIGNE framework:

import { AIAgent, AIGNE, isAgentResponseDelta } from "@aigne/core";
import { OpenAIChatModel } from "@aigne/openai";

Here we import two core components:

  • AIAgent - Class for creating and configuring AI Agents
  • AIGNE - The main framework class responsible for coordinating Agents and models
  • OpenAIChatModel - OpenAI chat model implementation for handling actual AI interactions

Create AIGNE Instance#

Next, we need to create an AIGNE instance and configure it to use OpenAI's model:

const aigne = new AIGNE({
model: new OpenAIChatModel({
apiKey: process.env.OPENAI_API_KEY,
model: "gpt-4o-mini",
}),
});

This code:

  • Creates an AIGNE framework instance
  • Configures it to use OpenAI's gpt-4o-mini model
  • Gets the API key from the environment variable OPENAI_API_KEY (make sure you've set this environment variable)

You can choose different models as needed, such as gpt-4o, o1, or other OpenAI-supported models.

Create Agent#

Now, let's create a simple AI Agent:

const agent = AIAgent.from({
instructions: "You are a helpful assistant",
inputKey: "message",
});

Here:

  • Use the AIAgent.from() method to create an Agent
  • Set the Agent's behavior guidelines through the instructions parameter
  • This simple instruction tells the AI it should play the role of a helpful assistant

You can customize instructions as needed to make the Agent play different roles or focus on specific domain tasks.

Run Agent#

Now we can invoke the Agent to handle user requests:

const result = await aigne.invoke(agent, { message: "What is AIGNE?" });
console.log(result);
// Output: { message: "AIGNE is a platform for building AI agents." }

This code:

  • Uses the aigne.invoke() method to call the Agent
  • Passes in the previously created Agent instance and the user's question
  • Asynchronously waits for the response and outputs the result
  • The output result is contained in the $message field

The invoke method is the primary way to interact with Agents, returning a Promise containing the AI response.

Streaming Mode#

For long responses or scenarios requiring real-time display of generated content, AIGNE supports streaming output:

const stream = await aigne.invoke(
agent,
{ message: "What is AIGNE?" },
{ streaming: true },
);

let response = "";
for await (const chunk of stream) {
console.log(chunk);
if (isAgentResponseDelta(chunk) && chunk.delta.text?.message)
response += chunk.delta.text.message;
}
console.log(response);
// Output: "AIGNE is a platform for building AI agents."

This code:

  • Enables streaming output by setting the { streaming: true } option
  • Uses a for await...of loop to process each response chunk
  • Appends the delta.text.$message from each chunk to the final response
  • Finally outputs the complete response

Streaming output is particularly useful for building real-time chat interfaces or applications that need to progressively display long responses.

Complete Example#

Below is a complete example that includes all the above steps:

import { AIAgent, AIGNE, isAgentResponseDelta } from "@aigne/core";
import { OpenAIChatModel } from "@aigne/openai";

const aigne = new AIGNE({
model: new OpenAIChatModel({
apiKey: process.env.OPENAI_API_KEY,
model: "gpt-4o-mini",
}),
});

const agent = AIAgent.from({
instructions: "You are a helpful assistant",
inputKey: "message",
});

const result = await aigne.invoke(agent, { message: "What is AIGNE?" });
console.log(result);
// Output: { message: "AIGNE is a platform for building AI agents." }

const stream = await aigne.invoke(
agent,
{ message: "What is AIGNE?" },
{ streaming: true },
);

let response = "";
for await (const chunk of stream) {
console.log(chunk);
if (isAgentResponseDelta(chunk) && chunk.delta.text?.message)
response += chunk.delta.text.message;
}
console.log(response);
// Output: "AIGNE is a platform for building AI agents."

previous
What is AIGNE?
next
AIGNE CLI