In-cluster inference and the Klyo API companion on Kubernetes
Kubernetes installs now reach parity with Docker Compose. The chart deploys an in-cluster Ollama by default — a models volume and an idempotent pull job (default llama3.2) — so chat works out of the box without wiring an external model server. The Klyo API companion service is enabled by default as well, with optional keyed authentication.
Uploads now use a shared PersistentVolumeClaim by default, so files survive pod deletion and stay reachable across replicas instead of returning 404 on a pod that did not receive them.
No action required for managed tenants.