Emmimal/memory-decay-engine ? reverse-engineered prompt

Build me a small Python library that helps an AI agent remember important things over long chats instead of just forgetting anything old. I want it to keep useful facts alive when they get recalled, and let unimportant noise fade out naturally over time.

It should include two modes, one that uses a forgetting curve with reinforcement from repeated use, and a simple baseline that just forgets things after a fixed number of turns so I can compare them. Please add a tiny demo, a session simulator, a few metrics to judge how well it keeps important facts, and a sanity check script so I can verify the logic before trusting any results.

Keep it pure Python with no third party dependencies, and make it easy to run from the command line. If anything about the retention logic or benchmarking is unclear, look up current docs online if you need to, but keep the finished project simple, readable, and reproducible.