llm-d/llm-d-router v1.0.0-rc.2

on GitHub

Release v1.0.0-rc.2

Highlights

• Give the storage tier (vLLM's native OffloadingConnector disk-offload medium) a default weight of 0.6 in prefix scoring so disk-offloaded blocks count toward prefix affinity without explicit kvCacheBackendConfigs (#3463).

What's Changed

• fix(kvcache): default the storage tier weight to 0.6 by @mevince in #3463

Full Changelog: v1.0.0-rc.1...v1.0.0-rc.2