Thoughts on Optimizing Gemini CLI's Memory Mechanism: Will a Cursor-Style LLM Evaluation System Be Introduced?

[boss-mao]

I noticed that the Gemini CLI uses a memory tool to handle user behavioral preferences. Are there any plans to add a memory evaluation mechanism in the future, similar to Cursor's memory, where another LLM evaluates whether the memory is worth retaining?

[jackwotherspoon]

Converting this to a discussion as it a question and not directly an issue per-say 👍

[Manamama]

I just tell any AI to discard irrelevant memories from .md like files. (Similar to its "context compression" but by hand, indeed). Safer this way actually, as the User decides what is irrelevant.