Open WebUI gets filed as "the Ollama GUI," which undersells the useful half: it speaks to any OpenAI-compatible endpoint. Point it at Groq, OpenRouter, NIM, a LiteLLM proxy, vLLM on another box — each becomes a model in the same picker, side by side with your local Ollama models.
That makes it something more interesting than a chat skin: one self-hosted interface where local and hosted models coexist, with conversation history stored on your machine instead of five providers' dashboards. Ask the same question of a local q4 model and a hosted 120B and diff the answers in one place.
Setup is the settings page: add a connection with base URL + key. The practical win is switching mid-conversation — draft with the free fast hosted model, finish with the local one when the content is sensitive.
Install and multi-backend config: https://toolfreebie.com/open-webui-self-hosted-ai/
Top comments (0)