I barely know what I’m doing (I’m an old IT infrastructure guy who doesn’t code) but today I ran across this python library. It seems like it’s handling the same limitations that we all deal with in context and that RAG and other solutions are trying to solve.
Could osmeone who understand this better than me have a look at this and see if it’s really a different approach or worth looking at?
" Memoripy is a Python library designed to manage and retrieve context-aware memory interactions using both short-term and long-term storage. It supports AI-driven applications requiring memory management, with compatibility for OpenAI and Ollama APIs. Features include contextual memory retrieval, memory decay and reinforcement, hierarchical clustering, and graph-based associations."
Thanks and I’m looking forward to anyone who understands this to respond.
I’ve been comparing this approach to other similar efforts people are talking about. Here’s another approach that sounds really exceptional but it involves far more changes to implement as it touches on the architecture of LLMs themselves: [2410.13166] An Evolved Universal Transformer Memory
I’d like to work on the latter but I expect I’ll use memoripy or a similar approach more immediately.
Thank you for sharing. Have you explored it further? I’m currently storing chat conversations as “memory” in a vector embedding and running similar queries to enhance my prompts with some contextual memory. I recently migrated the code from TypeScript/JavaScript to Python, and so far, the fundamental aspects are working well in Python as well. I’m using a method I found, and I haven’t looked into its limitations yet since I’m still in the early stages of app development; the essential agent memory is functioning as expected for now.
I have a similar background in IT, having been in the field for 25 years. I just started coding with AI three months ago. Welcome to the club!