Aggregated Kubernetes apiserver and e2b-compatible data plane for warm-pooled agent sandboxes on microVMs. Sandboxes are served from per-node inventory with no per-sandbox etcd object and claimed from node-local sandboxd warm pools; the e2b SDK reaches them through an edge proxy.
Karpenter provider that autoscales cluster membership of pre-existing SSH-reachable hosts — for fleets where SSH is all you get.
Zero-downtime deployments on AWS using Blue-Green Deployments, Rolling Deployments, Auto Scaling, Application Load Balancer, Lifecycle Hooks, and Warm Pools.
Phase-by-phase benchmark of LLM instance cold start on RTX 2070: weight loading, PCIe transfer, CUDA init, KV cache allocation, and JIT warmup — with pre-warming savings and extrapolation to 7B+ models.
Isolated agent sandboxes on Kubernetes with injected runtimes, LLM integration, and active-standby orchestration
To associate your repository with the warm-pool topic, visit your repo's landing page and select "manage topics."