From a9e46b61d3d2c167d53e3381051a6dbe273beb64 Mon Sep 17 00:00:00 2001 From: Zachary Huang <33015448+zachary62@users.noreply.github.com> Date: Thu, 26 Dec 2024 17:56:45 -0500 Subject: [PATCH] Update README.md --- README.md | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) diff --git a/README.md b/README.md index afb1340..1a40325 100644 --- a/README.md +++ b/README.md @@ -1,28 +1,30 @@ -

miniLLMFlow

+

Mini LLM Flow

![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg) -Minimalist LLM framework in [100 lines](minillmflow/__init__.py). Express popular paradigms like agents, task decomposition, RAG, and more. +A 100-line minimalist LLM framework for agents, task decomposition, RAG, etc. -- Install via ```pip install minillmflow```. Or just copy the [source codes](minillmflow/__init__.py) (it's only 100 lines) -- We **strongly recommened** using LLMs (e.g., Claude, ChatGPT, Cursor) to develop LLM applications with [this prompt](minillmflow/docs/prompt) +- Install via ```pip install minillmflow```, or just copy the [source](minillmflow/__init__.py) (only 100 lines) -## Why miniLLMFlow? +- **Pro tip:** Build LLM apps with LLMs assistants (ChatGPT, Claude, etc.) via [this prompt](minillmflow/docs/prompt) -The future of programming will be heavily LLM-assited, and LLMs: +## Why Mini LLM Flow? -1. **😀 Shine at Feature Implementation**: +The future of LLM app development will be **heavily LLM-assited**: users specify requirements, and LLMs build, test, and maintain on their own. Current LLM assistants: + +1. **😀 Shine at Low-level Implementation**: With proper docs, LLMs can handle APIs, tools, chunking, prompt wrapping, etc. -These are hard to maintain and optimize, so they don’t belong in the framework. +These are hard to maintain and optimize for a general-purpose framework. -2. **☹️ Struggle with Paradigm Design**: -Paradigms like MapReduce, task decomposition, and agents are powerful, even for LLMs. +2. **☹️ Struggle with High-level Paradigms**: +Paradigms like MapReduce, task decomposition, and agents are powerful for development. However, designing these elegantly remains challenging for LLMs. -To enable LLMs to develop LLM applications, a framework should -(1) remove redunant feature implementations but -(2) keep core paradigms to build on. -It turns out that such a framework only needs 100 lines of code. +To enable LLMs to develop LLM app, a framework should +(1) remove specialized low-level implementations, and +(2) keep high-level paradigms to program against. +Hence, I built this framework that lets LLMs focus on what matters. It turns out 100 lines is all you need. +