FERNme – agent memory that updates with ~zero LLM calls

FERNme is an AI/ML project that implements a brain-inspired, graph-based persistent memory system for agents, using fuzzy edges and a Hebbian co-occurrence rule to create memory tags with near-zero LLM calls. It is designed for developers building autonomous agents who need efficient, token-saving memory updates without constant reliance on expensive language models. The project is interesting because it dramatically reduces LLM token costs while mimicking biological memory processes, offering a scalable alternative to traditional context-window-based memory.