Recursive Language Models (escape context limits)

Anyone know if there is some addon framework implementing RLMs I can add to my local LLaMa.cpp inference pipeline? This looks like a way to truly escape the confines of very limited local context windows of retail vid cards. If nothing exists, I could start with the rlm-minimal repository from the original paper: [rlm-minimal](https://github.com/alexzhang13/rlm-minimal), and modify it to use llama-cpp-python instead of API calls. >[Recursive Language Models](https://discuss.google.dev/t/recursive-language-models-in-adk/323523)