← sources

forget compact we need offload to memory.md and seamless context management

source: reddit (reddit:LocalLLaMA)published 2026-09-25 22:40ingested 2026-09-26 02:38

retained text

while /compact is great for reducing context we need an unload to "memory.md". We need a version were stuff in context is put into a memory file.  Then every request first does a quick check of the request, current context and if we need to look into memory.md to reload old context. Manually having to jump around session is crap, we need a new AI layer to do session management and rebuild context from every session, with only the most reliant info to build the new context for a prompt.

maybe this already exist if so I would really love to know about it.

document fetched 2026-09-26 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1wqa5da/forget_compact_we_need_offload_to_memorymd_and