Skip to main content

SDK for scientific agents

Project description

Science Agent SDK

这是DP Tech的Science Agent SDK,提供了一个命令行工具dp-agent,用于管理科学计算任务。同时提供了Python SDK用于开发自定义的科学计算应用。

安装

Linux/Mac

curl -sSL https://raw.githubusercontent.com/dptech-corp/science-agent-sdk/refs/heads/feat/master/install.sh | bash

Windows

在PowerShell中运行:

Invoke-WebRequest -Uri https://raw.githubusercontent.com/dptech-corp/science-agent-sdk/refs/heads/feat/master/install.sh -OutFile install.sh
bash install.sh

CLI 使用方法

安装后,您可以使用以下命令:

获取资源

# 获取基础代码结构
dp-agent fetch scaffolding

# 获取配置文件
dp-agent fetch config

fetch config 命令会下载 .env 配置文件并替换部分动态变量,如 MQTT_DEVICE_ID。 注意:出于安全考虑,此功能仅在内网环境可用。其他环境需要手动配置。

运行命令

# 运行实验环境
dp-agent run lab

# 运行云环境
dp-agent run cloud

# 运行代理
dp-agent run agent

# 调试模式
dp-agent run debug

SDK 快速入门

Science Agent SDK 提供了两种主要的开发模式:实验室模式(Lab)和云模式(Cloud)。

基础结构

安装完成并运行 dp-agent fetch scaffolding 后,您将获得以下基础项目结构:

your-project/
├── lab/                # 实验室模式相关代码
│   ├── __init__.py
│   └── tescan_device.py  # 设备控制示例
├── cloud/              # 云模式相关代码
│   └── __init__.py
└── main.py            # 主程序入口

实验室模式开发

实验室模式主要用于控制本地实验设备。以下是一个基于 Tescan 设备的示例:

from typing import Dict, TypedDict
from dp.agent.lab.device import Device, action, BaseParams, SuccessResult

class TakePictureParams(BaseParams):
    """拍照参数"""
    horizontal_width: str

class PictureData(TypedDict):
    """照片数据"""
    image_id: str

class PictureResult(SuccessResult):
    """拍照结果"""
    data: PictureData

class MyDevice(Device):
    device_name = "my_device"
    
    @action("take_picture")
    def take_picture(self, params: TakePictureParams) -> PictureResult:
        """拍照动作
        
        Args:
            params: 拍照参数
                - horizontal_width: 图片水平宽度
        """
        hw = params.get("horizontal_width", "default")
        return PictureResult(
            message=f"Picture taken with {self.device_name}",
            data={"image_id": "image_123"}
        )

云端开发

云模式基于 MCP (Message Control Protocol) 实现,用于处理远程设备控制和任务调度。register_mcp_tools 通过 python 的自省和反射机制实现了设备控制的自动注册,无需重复实现操作定义。 以下展示如何创建设备并注册到 MCP 服务器:

"""
Example of using the science-agent-sdk cloud functionality.
"""
import signal
import sys
from dp.agent.cloud import mcp, get_mqtt_cloud_instance
from dp.agent.lab.device import TescanDevice, register_mcp_tools
from dp.agent.lab.device.device import register_mcp_tools

def signal_handler(sig, frame):
    """Handle SIGINT signal to gracefully shutdown."""
    print("Shutting down...")
    get_mqtt_cloud_instance().stop()
    sys.exit(0)

def main():
    """Start the cloud services."""
    print("Starting Tescan Device Twin Cloud Services...")
    
    # Register signal handler
    signal.signal(signal.SIGINT, signal_handler)
    
    # Create device instance
    device = TescanDevice(mcp, device)
    
    # Register device tools
    register_mcp_tools(device)
    
    # Start MCP server
    print("Starting MCP server...")
    mcp.run(transport="sse")

if __name__ == "__main__":
    main()

配置说明

.env 文件中配置必要的环境变量:

MQTT_INSTANCE_ID=your_instance_id
MQTT_ENDPOINT=your_endpoint
MQTT_DEVICE_ID=your_device_id
MQTT_GROUP_ID=your_group_id
MQTT_AK=your_access_key
MQTT_SK=your_secret_key

主要功能

  • 设备控制接口(Lab模式)

    • 设备初始化
    • 命令执行
    • 状态监控
  • 云端任务处理(Cloud模式)

    • 任务队列管理
    • 计算资源调度
    • 结果回传

更详细的API文档请参考代码中的注释。

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

science_agent_sdk-0.1.3.tar.gz (34.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

science_agent_sdk-0.1.3-py3-none-any.whl (41.2 kB view details)

Uploaded Python 3

File details

Details for the file science_agent_sdk-0.1.3.tar.gz.

File metadata

  • Download URL: science_agent_sdk-0.1.3.tar.gz
  • Upload date:
  • Size: 34.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.12.2

File hashes

Hashes for science_agent_sdk-0.1.3.tar.gz
Algorithm Hash digest
SHA256 dee13e874be6082509578c8ab4c887b633e3e68afccefeabd331e3990764dc8a
MD5 44d9e9187cd58cf0c6c4d796c8225a84
BLAKE2b-256 cb6a6a6ec10efc9928d3ad29c4f2b1338eee24ff97a70c992f488624a1c51318

See more details on using hashes here.

File details

Details for the file science_agent_sdk-0.1.3-py3-none-any.whl.

File metadata

File hashes

Hashes for science_agent_sdk-0.1.3-py3-none-any.whl
Algorithm Hash digest
SHA256 5b3b4f58fdf9ac929fd594b8b55b8e70da306f4a827d55d5dcd69e293ece26c4
MD5 e4b1090e2a6ed4c6f30068c7f35811a7
BLAKE2b-256 5c273db8c29bf987423f2499e854127e11063ace37a0c2589ca17cbd8a512f85

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page