Description
You have heard of Reflexion, LATS, GraphRAG and MemGPT, but picking one for your task is guesswork. This project turns 35 agentic AI architectures into a runnable Python library, each with real model outputs.
It supports 9 LLM providers, ships 283 tests and a 17-task benchmark leaderboard comparing architectures head to head, and doubles as a textbook you can run.
35 architectures:Reflexion, LATS, GraphRAG, MemGPT, Voyager and more.
Many providers:Nine LLM vendors.
Leaderboard:17 tasks compared.
Runnable textbook:Explanations with code.
It supports 9 LLM providers, ships 283 tests and a 17-task benchmark leaderboard comparing architectures head to head, and doubles as a textbook you can run.
Features
35 architectures:Reflexion, LATS, GraphRAG, MemGPT, Voyager and more.
Many providers:Nine LLM vendors.
Leaderboard:17 tasks compared.
Runnable textbook:Explanations with code.
