Private AI interface and inference

Connect Open WebUI to a customer-operated vLLM inference API. The interface and model server have separate storage, access, model-license and resource requirements.

Documented connection

Open WebUI documents adding the vLLM OpenAI-compatible endpoint under Manage OpenAI API Connections; configure bearer authentication when enabled on the model server. Official integration guide (reviewed 2026-11-16).

Components

Architecture boundaries

This pair does not provide a vector database, retrieval pipeline, observability, model entitlement or a complete air-gapped design. A cloud badge on one component does not establish deployment of the entire stack on that provider.

Compare component evidence · Start a shortlist with these components