Mini LLM Flow最小化的 100 行 LLM 框架，用于构建自我编程代理工作流程

4.8 (6)

审阅者 Daniel Nikulshyn·更新 2026年7月

Open Source Developer Tools LLM Framework Agents Minimalist Workflow Orchestration Educational

概览

Mini LLM Flow 是一个轻量级的开源框架，能够将 LLM 编排压缩到大约 100 行代码。它提供了链式提示、状态管理以及构建代理工作流所需的核心构件，避免了更大型框架带来的额外负担。该项目的核心思想是，采用最小化的抽象更容易让 LLM 本身去理解、扩展并生成相应代码。这使其非常适合自编程代理（self-programming agents）的实验，在这些实验中模型会对自己的 workflow logic 进行推理和修改。开发者可以将其用作学习工具、定制代理系统的基础，或者作为更重型编排库的精简版替代。

主要功能

核心代码大约100行
促进链式和流程控制
兼容代理流式工作流
设计用于 LLM 自我编程
最小的依赖
开放且容易分叉

价格

模型: Free
分类: AI Agents Frameworks
评分: 4.8 / 5 (6)

使用场景

学习代理工作流基本知识

研究一个压缩到约100行代码的紧凑代码库，以了解如何促进链式、状态和代理编排工作，无需涉及大的框架。

创建自定义轻量级代理系统

将最小核心代码分叉为自定义代理工作流的基础，避免使用更大的编排库的重复依赖和固化。

实验自我编程代理

利用最小的抽象，使得 LLM 可以读、reason 和更可靠地生成他们自己的工作流代码变体。

快速构造 LLM.pipeline

使用剥夺的基本构件快速启动促进链并流程控制，为原型设计证明概念之前，在更大堆栈上承诺。

优点 & 缺点

优点

极小且可读的代码库
很容易让 LLM 理解和扩展
没有重复的依赖或固化
很好的教育资源，用于代理设计

缺点

与更大框架相比，被内置的特性有限
需要手动设置更多复杂的用例
较小的用户社区和生态

对决战绩

在万神殿中参与了 1 对决。

第1

第2

第3

Last battle

#1
AI Agents Frameworks Showdown — July 7, 2026
Jul 7, 2026 · #1 of 10

评测

4.8

6 个评分的平均值。

登录以留下评测。

Daniel Schmidt

Apr 29, 2026

Years in this space

I've evaluated a lot of these over the years. What stands out here is minimal dependencies — handled better than most — and no heavy dependencies or lock-in. Limited built-in features compared to larger frameworks is my one real gripe. Worth the time if this is your use case.

Leila Hassan

Jan 30, 2026

Years in this space

I've evaluated a lot of these over the years. What stands out here is open and easily forkable — handled better than most — and extremely small and readable codebase. Worth the time if this is your use case.

Sanjay Gupta

Jan 8, 2026

Years in this space

I've evaluated a lot of these over the years. What stands out here is prompt chaining and flow control — handled better than most — and extremely small and readable codebase. Smaller community and ecosystem is my one real gripe. Worth the time if this is your use case.

Gunnar Eriksson

Nov 18, 2025

Years in this space

I've evaluated a lot of these over the years. What stands out here is designed for LLM self-programming — handled better than most — and extremely small and readable codebase. Worth the time if this is your use case.

Linda Petersen

Nov 13, 2025

Solid for our team

We rolled this out across the team last quarter and extremely small and readable codebase. Support for agent-style workflows fits neatly into how we already work, and prompt chaining and flow control removed a step we used to do by hand. but it has held up under daily use.

Yuki Mori

Oct 14, 2025

Skeptical, then convinced

I went in skeptical — most tools in this space overpromise. It actually delivers on minimal dependencies, and no heavy dependencies or lock-in caught me off guard. still, I'd recommend giving it a real trial.