rephrase readme
This commit is contained in:
parent
cab0ab6a1c
commit
87929060c8
|
|
@ -68,4 +68,7 @@ htmlcov/
|
|||
# Misc
|
||||
*.bak
|
||||
*.tmp
|
||||
*.temp
|
||||
*.temp
|
||||
|
||||
|
||||
test.ipynb
|
||||
22
README.md
22
README.md
|
|
@ -5,24 +5,24 @@
|
|||
Minimalist LLM framework in [100 lines](minillmflow/__init__.py). Express popular paradigms like agents, task decomposition, RAG, and more.
|
||||
|
||||
- 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 build LLM applications with [this prompt](minillmflow/docs/prompt)
|
||||
- We **strongly recommened** using LLMs (e.g., Claude, ChatGPT, Cursor) to develop LLM applications with [this prompt](minillmflow/docs/prompt)
|
||||
|
||||
## Why miniLLMFlow?
|
||||
|
||||
The future of programming will be heavily LLM-assited, and LLMs:
|
||||
|
||||
1. **👍 Excel at Feature Implementation**:
|
||||
Give proper context/docs, they know how to use APIs, tools, text chunking, prompt wrapping, etc.
|
||||
These shouldn't be part of an LLM framework, as they're hard to maintain, update, and optimize.
|
||||
1. **😀 Shine at Feature 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.
|
||||
|
||||
2. **👎 Suck at Paradigm Design**:
|
||||
Paradigms like MapReduce, DAG workflows, and recent agents are powerful for reasoning about problems.
|
||||
However, designing elegant paradigms is challenging, and LLMs tend to write redundant code.
|
||||
2. **☹️ Struggle with Paradigm Design**:
|
||||
Paradigms like MapReduce, task decomposition, and agents are powerful, even for LLMs.
|
||||
However, designing these elegantly remains challenging for LLMs.
|
||||
|
||||
Can we build a framework that:
|
||||
(1) removes redunant feature implementations,
|
||||
(2) but keep core paradigms for LLMs to program against?
|
||||
Turns out that we just need 100 lines of codes.
|
||||
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.
|
||||
|
||||
<div align="center">
|
||||
<img src="./docs/minillmflow.jpg" width="400"/>
|
||||
|
|
|
|||
Loading…
Reference in New Issue