Kubernetes, first-class
A Helm chart with sane values, replicas, persistence size, license secret, probes wired to /health, so the path Enterprise teams already take becomes a one-line install.
Roadmap
The direction of travel, and the lines we've committed never to cross.
A Helm chart with sane values, replicas, persistence size, license secret, probes wired to /health, so the path Enterprise teams already take becomes a one-line install.
OpenTelemetry trace export for per-query timelines, and the Reasoning DAG inspector so operators can see exactly which reasoning skeleton served a hit.
Replicated clusters with the same trust guarantees, so a warm cache in one region warms the next one.
Audio caching, agent checkpointing, request coalescence, and RAG-specialized paths, extending the same five-veto reuse discipline to more of the LLM stack.
No phone-home, no usage metering, no cloud-account requirement, and no feature that requires trusting us with your traffic. The offline, self-hosted contract is permanent.