グローバル共有メモ
jrf> 英文告知。 I conducted experiments on "Bear-Sword Maze Problem with LangChain 1.x" and "Experimental RAG Implementation using MemoryBanditWorkflow (LangChain 1.x)." 《langchain_maze_en_0_0_15.ipynb - JRF-2018/langchain_maze》 https://github.com/JRF-2018/langchain_maze/blob/master/langchain_maze_en_0_0_15.ipynb 《experimental_rag_en_0_0_16.ipynb - JRF-2018/langchain_maze》 https://github.com/JRF-2018/langchain_maze/blob/master/experimental_rag_en_0_0_16.ipynb 《JRF-2018/langchain_maze: Bear-Sword Maze Problem Revisited (with MemoryBanditWorkflow)》 https://github.com/JRF-2018/langchain_maze `MemoryBanditWorkflow` is an agent framework created based on experiences with the Bear-Sword Maze problem. It has now been updated to support LangChain 1.x. Furthermore, I have introduced `subtool_do`, which serves as an equivalent to features like "skills" or "toolboxes" in other frameworks. Although it originated as part of a maze-solving project, I aimed to design it as a versatile framework. This project represents an attempt to build a RAG (Retrieval-Augmented Generation) agent to demonstrate that the framework is indeed capable of general-purpose tasks. Unlike many complex multi-agent frameworks, `MemoryBanditWorkflow` focuses on a clear, traceable execution path.
---/2000
ログ