LiteLLM Dashboard
Create and revoke virtual keys, edit model allowlists, set limits, and review usage.
Operator launchpad / 01
Use the links below to reach the AI gateway, its management console, and LAN-only infrastructure without exposing the control plane.
Create and revoke virtual keys, edit model allowlists, set limits, and review usage.
Use this base URL in OpenAI-compatible clients. Choose a canonical model name from the registry below.
Open the authenticated Odysseus workspace through its dedicated HTTPS host. The application port is never exposed publicly.
Uptime Kuma remains private on the LAN; complete its first-run setup from inside the network.
Proxmox, OPNsense, and infrastructure stay off the public edge. Add WireGuard or Tailscale later; no new WAN admin forwards are needed.
| Public model ID | Node | Workload | Legacy alias |
|---|---|---|---|
| qwen38-27b-ada01-quality | node-ada01 | Qwen 3.8 27B / quality lane | local-quality |
| qwen38-27b-ada01-primary | node-ada01 | Qwen 3.8 27B / primary lane | vm2061-primary |
| qwen38-27b-ada01-benchmark | node-ada01 | Qwen 3.8 27B / benchmark lane | qwen38-bench |
| qwen38-27b-ada02-capable | node-ada02 | Qwen 3.8 27B / capable lane | qwen38-capable |
| ornith-9b-ada03-swarm | node-ada03 | Ornith 9B / swarm lane | fleet-ornith-swarm |
| utility-quadro01-fast | node-quadro01 | Utility / fast lane | fleet-utility-fast |
| utility-quadro01-small | node-quadro01 | Utility / small lane | fleet-utility-small |
| mtp-v10001-q6 | node-v10001 | MTP / Q6 lane | MTP-Q6_K.gguf |
Set the base URL once, then use a virtual key and one of the model IDs above.
Velocity is the only public Minecraft listener; backend servers remain private.