Best-effort high availability for single-replica Kubernetes workloads: no downtime from announced node disruptions (drains, upgrades, consolidation, spot reclaims) without a standing second replica.
Least recently used (LRU) eviction of Docker images. 🗑️
Code for the paper "Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning" https://arxiv.org/abs/2609.03430