What an OpenAI-compatible endpoint is
Any server that answers the OpenAI chat API, vLLM, Ollama, LM Studio, a LiteLLM gateway, a provider’s own compatible API. If it speaks that format, Vera can use it.
On-premises
Your own hardware, reached over a public HTTPS address. Vera connects to that address; the model and its weights never leave your building.
Cloud-hosted
A provider or gateway you run elsewhere, your own cloud account, a hosting service, a gateway in front of several models. Vera connects to the address you give it.
Your models stay yours
Vera discovers the models from your endpoint and lists them privately to your workspace. They never join the shared catalog and no other workspace can see them.
One endpoint per workspace
Each workspace can connect one endpoint of its own, beside any vendor keys it holds. Set it up in workspace admin settings.