Services

ai / running

Open WebUI

Browser chat UI for local LLMs on Zenbook/Jarvis. Plugs into LM Studio as the main runtime and Ollama as legacy fallback.

What it is

Open WebUI is a ChatGPT-style browser UI for local LLMs. Multiple providers, model switcher in the corner, conversation history, file uploads, RAG against my own documents if I set that up. It's the user-facing piece on top of LM Studio and Ollama — they expose APIs, this is the chat interface.

Why I run it

LM Studio has its own chat UI. The reason I run Open WebUI on top of it anyway:

LM Studio's own UI is fine on the Zenbook. Open WebUI is what I use from everywhere else, and it is the internal browser face of the Zenbook AI host for the rest of the homelab.

How I use it

Two providers are configured: LM Studio (current daily driver) and Ollama (legacy fallback). The model picker shows them both — LM Studio models grouped under "External," Ollama under "Local." I pick whichever model fits the task: a 7B Qwen instruct for general chat, the reasoning-tuned DeepSeek when I want a longer chain of thought, Gemma multimodal when I need vision.

The connection to LM Studio uses host.docker.internal:1234/v1 because Open WebUI runs in Docker and localhost from inside the container resolves to the container itself, not the Zenbook's host. That confused me for a useful five minutes the first time.

Setup notes

Runbook