AgentScope AI Engine

AgentScope AI Engine is the core execution framework that powers intelligent agent workflows. It provides a robust environment for developing, deploying, and managing AI-driven agents capable of processing tasks, automating workflows, and integrating with various external services.

This document outlines the architecture, key components, and usage of the AgentScope AI Engine for local development.

---

Architecture

The AgentScope AI Engine follows a modular architecture designed for scalability and flexibility. It consists of the following core components:

1. Agent Runtime: Manages the execution of AI agents, ensuring optimal task processing and resource allocation.

2. Task Scheduler: Handles the prioritization and execution of agent tasks.

3. Memory Store: Maintains agent state and history for contextual decision-making.

4. Tool Integrations: Allows agents to leverage external APIs and tools to enhance their capabilities.

5. Logging & Debugging: Provides real-time monitoring and debugging utilities for developers.

---

Installation & Setup

To get started with AgentScope AI Engine, follow these steps:

1. Install Dependencies

Ensure that you have Node.js installed. Then, install AgentScope AI Engine via npm:

npm install -g agentscope-engine

2. Initialize a New Project

Create a new directory for your project and initialize the engine:

mkdir my-agentscope-project
cd my-agentscope-project
agentscope init

3. Configure Environment Variables

Set up a .env file to configure API keys and other environment settings:

OPENAI_API_KEY=your_openai_api_key
DATABASE_URL=your_database_url

---

Defining an Agent

Agents in AgentScope AI Engine are defined using a simple configuration format. Below is an example of creating a basic AI agent:

src/agents/myAgent.ts

import { Agent } from "@agentscope/core";
import * as tools from "../tools/weatherInfo";

export const myAgent = new Agent<typeof tools>({
  name: "Weather Agent",
  instructions:
    "You are a helpful assistant that provides current weather information.",
  model: {
    provider: "OPEN_AI",
    name: "gpt-4",
    toolChoice: "required",
  },
  tools: {
    weatherInfo: tools.weatherInfo,
  },
});

---

Running the AgentScope AI Engine

To start the AgentScope AI Engine and execute workflows:

agentscope start

For debugging and real-time logs:

agentscope start --debug

---

Testing & Debugging

AgentScope AI Engine supports testing using Jest or Vitest. Below is an example of a simple test:

tests/weatherAgent.test.ts

import { myAgent } from "../src/agents/myAgent";

test("Weather Agent fetches data successfully", async () => {
  const response = await myAgent.generate("What's the weather like in New York?");
  expect(response.text).toContain("temperature");
});

Run the tests using:

npm test

---

Deployment

To deploy the AgentScope AI Engine, you can use Docker or cloud-based services.

Using Docker

Create a Dockerfile:

FROM node:18
WORKDIR /app
COPY . .
RUN npm install
CMD ["npm", "start"]

Build and run the container:

docker build -t agentscope-engine .
docker run -p 3000:3000 agentscope-engine

Cloud Deployment

For cloud deployment, consider using services like AWS Lambda, Google Cloud Functions, or Vercel for serverless execution.