[OutOfHere]

The stronger need for LLM apps is for persistent response caching and reuse. Once this is available at scale with well-defined cache expiration policies, printing logs selectively is easy. The solution ought to also support individual key invalidation as needed.

[idomi]

I think it really depends on your need, if you're working on your local and trying to get something to work, caching, scale etc, might be an overkill.