MemGPT-inspired tiered memory system with working memory, archival storage, and automated context paging
-
Updated
Sep 29, 2026 - Python
MemGPT-inspired tiered memory system with working memory, archival storage, and automated context paging
MemGPT-inspired tiered memory system with working memory, archival storage, and automated context paging
COMEM : force the use of memory reclaim by COnstraining MEMory.
Extend RAM with NVMe/SSD capacity. Cross-platform tiered memory for GPU, AI and gaming workloads, written in Rust with a C API.
Tiered memory layout for Hermes: hot (always-injected), warm (lazy-loaded by topic), cold (grep-only). Drops in for any Hermes install.
Can a commodity PC with no CXL support benefit from a large, slow DRAM tier attached over PCIe, managed by unmodified Linux memory management?
To associate your repository with the tiered-memory topic, visit your repo's landing page and select "manage topics."