OpenAI

Any OpenAI-compatible endpoint works — OpenAI, vLLM, LM Studio, or your own proxy.

export OPENAI_API_KEY=your-key

contenox backend add openai --type openai --api-key-env OPENAI_API_KEY
contenox config set default-model gpt-5-mini
contenox config set default-provider openai

For a custom endpoint (vLLM, LM Studio, etc.) add --url:

contenox backend add local-vllm --type openai --url http://localhost:8000/v1

Local servers typically need no key; if yours does, add --api-key-env <VAR>.

Contenox also has a native vllm backend type (contenox backend add myvllm --type vllm --url http://localhost:8000); the --type openai route works for any OpenAI-compatible server. If you want a commercially supported vLLM for production, Red Hat ships one as Red Hat AI Inference Server.

EU data residency

OpenAI offers European data residency for eligible API customers. It is configured per project: create a new project with Europe as its region (existing projects cannot be switched), and requests for that project go through the regional endpoint eu.api.openai.com, where OpenAI states eligible traffic is handled in-region with zero data retention. Point contenox at it with --url:

contenox backend add openai-eu --type openai --url https://eu.api.openai.com/v1 --api-key-env OPENAI_API_KEY

Eligibility, covered endpoints, and mechanics are OpenAI’s to define — check OpenAI’s data residency documentation before relying on it. See AI sovereignty & the EU AI Act for how region pinning compares to self-hosted inference.

See also

• Configuration reference