AgentScope AI Dev

AgentScope AI Dev provides a local development environment for building, testing, and deploying AI-driven agents efficiently. This module allows developers to create custom agents, integrate tools, and simulate real-world interactions within a controlled setup before deploying them in production.

Prerequisites

Before setting up AgentScope AI Dev, ensure you have the following:

• Node.js (>=16.0)

• Python (>=3.8)

• Docker (Optional, for containerized development)

• Git

• Access to AgentScope AI's SDK

Installation

To install AgentScope AI Dev, run the following command:

npm install -g agentscope-ai-dev

Alternatively, if using Python:

pip install agentscope-ai-dev

Initializing a Project

To start a new AgentScope AI development project:

agentscope init my-project
cd my-project

This will create a directory structure with the necessary configuration files.

Configuring the Development Environment

Inside my-project, you’ll find the configuration file:

{
  "agent_name": "MyCustomAgent",
  "model": "gpt-4",
  "tools": ["retrieval", "data_analysis"],
  "workflows": ["data_pipeline"]
}

Modify this file to define your agent’s properties and the tools it can access.

Running the Development Server

To start the local development server:

agentscope dev

This will launch a local instance where you can interact with your agents in a sandbox environment.

Testing Agents

Use the CLI or an API request to test agent responses:

agentscope test "What is the latest market trend in AI?"

Or via API:

import requests

response = requests.post("http://localhost:8000/agent", json={"query": "Analyze market trends in AI."})
print(response.json())

Debugging and Logging

Enable verbose mode for detailed logs:

agentscope dev --verbose

Logs will be available in logs/dev.log.

Deploying to Production

Once testing is complete, deploy the agent using:

agentscope deploy --env production

This will push the agent to the production environment configured in your AgentScope AI account.