Service preview
EHX Launch
Production-ready Kubernetes deployment patterns.
- Opinionated cluster topology with HA, upgrades, and Gateway API edge defaults suited for Web3-adjacent and cloud-native workloads.
- LLM inference on Kubernetes (GPU, vLLM/TGI/Ollama, OCI Image Volume) via Project Builder and the `llm_inference` catalog recipe — targets K8s 1.36.
- Reproducible infrastructure packs: reproducible Terraform, Kubernetes manifests, and Helm values—not one-off snowflakes.
Why teams evaluate this line
K8s 1.36
Platform target
Starter packs and validation default to Kubernetes 1.36 — Gateway API edge, not retired Ingress NGINX.
LLM pack
GPU inference
vLLM, TGI, and Ollama GPU deployments plus OCI Image Volume model delivery in the catalog.
Git-first
Outputs
Terraform, Kubernetes YAML, and Helm values stay reviewable in version control.
Launch flow (preview)
Directional preview—not a deployment diagram for your environment.
Git repo
Your infra definitions as the source of truth.
Terraform
Cloud / bare-metal foundation.
Kubernetes + Helm
Workloads, ingress, policies.
Operate
Hand off to Monitor presets when you adopt both.
How teams evaluate EHX Launch
Illustrative scenarios aligned with roadmap personas—not customer logos or endorsements. Published case studies ship via the design partner program when we have written approval.
Solo DevOps engineer
Situation. A three-person product team needs production Kubernetes before hiring a platform org.
Outcome. Compare Launch defaults for ingress, upgrades, and Git-reviewed Helm against a one-off consultant quote.
AI startup team
Situation. The team needs vLLM (or TGI/Ollama) on Kubernetes with Gateway API exposure and Prometheus scrape before production traffic.
Outcome. Use Project Builder (LLM inference workload) or chat intent `llm_inference` to export a validated bundle — GPU deployments, Image Volume, and security baselines included.
Often evaluated together
Roadmap-aligned bundle hints—not separate products you must buy today.
EHX Monitor
Wire cluster metrics and alerts as soon as workloads land—Monitor presets match Launch outputs.
Open previewEHX Secure
Validate ingress, image, and policy posture before you promote to production traffic.
Open preview
Starter pack samples
Preview and copy on Infrastructure templates— no external repository access required.
- Start guided Launch setup
- Start guided Helm chart setup
- Start guided LLM inference setup
- Start guided full-stack setup (layer Include/Skip)
- Kubernetes namespace example (reference)
- Terraform foundation sketch
- Helm values example
- GPU vLLM Deployment (LLM pack)
- OCI Image Volume for models (K8s 1.36)
- Gateway API minimum
Commercial path & further reading
- Compare entitlements on Pricing and exercise limits on Plan & usage.
- Template libraries in the Infrastructure library; pay with USDT/USDC on Ethereum, Base, or Arbitrum One via Checkout. Docs at docs.ehxlabs.xyz.
What Launch covers
- Baseline cluster layout, node pools, and upgrade-safe defaults.
- Ingress, TLS edge, and platform integration patterns (cloud-specific wiring stays explicit).
- Stateful and stateless workloads: Deployments, StatefulSets, Services, Ingress, HPA, and PVCs as the building blocks.
Outputs
- Terraform direction across major clouds and bare-metal-friendly providers (Terraform, Kubernetes, Helm).
- Kubernetes YAML and Helm values that stay valid and reviewable in Git.
- Monitoring hooks that feed EHX Monitor presets when you adopt both bundles.
LLM inference (PLAT catalog)
- Project Builder workload: LLM inference on Kubernetes — GPU count, model URI, backend (vLLM, TGI, Ollama), and exposure mode.
- Starter pack files: namespace, GPU deployments, OCI Image Volume for model weights, Service, Gateway API exposure, scrape, and baseline NetworkPolicy.
- Chat intent `llm_inference` composes the bundle; validate and export on /generate with platform target Kubernetes 1.36.
Who it is for
- Teams that need production Kubernetes without standing up a full platform engineering org on day one.
- AI startup teams bringing up GPU inference without bespoke YAML snowflakes.
- Web3 infrastructure teams that want deployment patterns aligned with validators, RPC, and L2 operations reality.