As I delve deeper into the world of Large Language Models (LLMs), I find myself wondering: do LLMs have their own design patterns? Patterns that make them efficient, creative, or just plain smarter. Think LangGraph, LangExtract, and other innovative approaches that make our systems more intelligent. What lies beneath these patterns? And can we apply them easily?
I decided to take matters into my own hands and started a repository to collect the designs of current LLM products. This personal project aims to help me stay up-to-date with the latest design patterns and mechanisms for LLMs. Since most open-source LLM projects are built on Python, I’m gathering them all to showcase how modern Python AI apps and tools are built. This repository will serve as a knowledge base, tracing the development and creative usage methods of LLMs.
I began with Claude Code, which excels at fetching and analyzing repositories. I’ve added a few use cases, categorized information, and demonstrated frequent usage in workshops. As I continue to enrich this repository with more cases and workshops, I invite you to join me on this journey of discovery.
The repository is open for anyone to use as a knowledge base, and I encourage you to do so. You can find it on GitHub, along with the accompanying workshop and GitBook.
GitHub: https://github.com/liyedanpdx/llm-python-patterns
Workshop: https://github.com/liyedanpdx/llm-python-patterns/tree/main/workshops
GitBook: https://paradx.gitbook.io/llm-python-patterns/
By exploring and understanding these design patterns, we can unlock the full potential of LLMs and create more innovative, efficient, and intelligent systems. Join me in this exploration and let’s uncover the patterns that make LLMs tick.