Miniagent
What is Miniagent
MiniAgent is a lightweight and easy-to-use framework for building LLM agents from scratch in just 5 minutes. It helps beginners quickly understand the principles of agents and the interaction between LLMs and tools.
Use cases
Use cases for MiniAgent include building educational tools for learning about LLMs, developing chatbots that utilize various LLMs, and creating custom agents for specific tasks such as calculations or retrieving information.
How to use
To use MiniAgent, clone the repository, install the dependencies, configure your API keys and model settings in a .env file, and run a simple example script to validate your LLM connection.
Key features
Key features of MiniAgent include quick learning of LLM agent principles, model-agnostic compatibility with various LLMs, minimal design focusing on core functionalities, lightweight integration without heavy dependencies, and easy tool registration and invocation through natural language.
Where to use
MiniAgent can be used in various fields such as education for teaching LLM concepts, software development for creating custom agents, and research for experimenting with LLM interactions.
Clients Supporting MCP
The following are the main client software that supports the Model Context Protocol. Click the link to visit the official website for more information.
Overview
What is Miniagent
MiniAgent is a lightweight and easy-to-use framework for building LLM agents from scratch in just 5 minutes. It helps beginners quickly understand the principles of agents and the interaction between LLMs and tools.
Use cases
Use cases for MiniAgent include building educational tools for learning about LLMs, developing chatbots that utilize various LLMs, and creating custom agents for specific tasks such as calculations or retrieving information.
How to use
To use MiniAgent, clone the repository, install the dependencies, configure your API keys and model settings in a .env file, and run a simple example script to validate your LLM connection.
Key features
Key features of MiniAgent include quick learning of LLM agent principles, model-agnostic compatibility with various LLMs, minimal design focusing on core functionalities, lightweight integration without heavy dependencies, and easy tool registration and invocation through natural language.
Where to use
MiniAgent can be used in various fields such as education for teaching LLM concepts, software development for creating custom agents, and research for experimenting with LLM interactions.
Clients Supporting MCP
The following are the main client software that supports the Model Context Protocol. Click the link to visit the official website for more information.
Content
MiniAgent
- 🚀「5分钟从零实现LLM Agent！」 |⚡️English Version
为什么选择MiniAgent？
帮助初学者，快速理解Agent原理并实现一个Agent，直观了解LLM与工具交互的细节。
如果这个项目对你有帮助，请给我们一个⭐️！你的支持是我们持续改进的动力。
特色功能
MiniAgent是一个轻量级且易于使用的LLM Agent框架。以下是选择MiniAgent的理由：
- 快速学习：快速学习LLM Agent工作原理,直观了解LLM与工具交互的细节
- 模型无关：适用于任何LLM，即使没有原生函数调用能力，兼容OpenAI、DeepSeek、Anthropic等LLM
- 极简设计：专注于核心功能的清晰、可读代码
- 轻量级：无重依赖，易于集成和扩展
- 简单工具集成：通过自然语言轻松注册和调用工具
快速开始
环境要求
- Python 3.8+
- 任何LLM的有效API密钥（OpenAI、DeepSeek等）
安装
# 克隆仓库
git clone https://github.com/yourusername/MiniAgent.git
cd MiniAgent
# 安装依赖
pip install -r requirements.txt
配置
从.env.example文件创建项目根目录下的.env文件：
# API密钥
LLM_API_KEY=your_api_key_here
# 模型配置
LLM_MODEL=deepseek-chat
LLM_API_BASE=https://api.deepseek.com/v1
LLM_TEMPERATURE=0.7
验证LLM连接
在使用MiniAgent之前，请验证您的LLM设置：
# 快速验证
python validate_llm.py
基础示例
运行一个简单示例：
python examples/simple_example.py
创建你自己的Agent
from miniagent import MiniAgent
from miniagent.tools import load_tools, register_tool
# 定义自定义工具（可选）
@register_tool
def calculator(expression: str) -> float:
"""计算数学表达式的结果"""
return eval(expression)
# 创建Agent, 使用.env中的配置
agent = MiniAgent()
# 加载工具
tools = load_tools(["calculator", "get_current_time"])
# 运行Agent
response = agent.run(
query="现在几点了？同时计算123 × 456。",
tools=tools
)
print(response)
示例
查看examples/目录获取更多详细示例：
simple_example.py：与任何LLM的基本用法custom_tools_example.py：使用自定义工具
许可证
Star历史
Dev Tools Supporting MCP
The following are the main code editors that support the Model Context Protocol. Click the link to visit the official website for more information.
