← Return to Log

🏠 Getting Started with Local LLMs [04] Installing Open WebUI and Using It Across a LAN

• P-chan •
LLM DevLog PochomLab

Local LLM Setup Series

👀 Table of Contents

🛠️ Why Install Open WebUI

A local LLM can already run with Ollama alone, but it becomes much easier to use once it is available
through a browser on another PC, such as a working laptop.

That is why I installed Open WebUI this time.

The goals were simple:

  • to make the local LLM usable from a browser
  • to access it from another PC on the same LAN
  • to make model switching and chat testing easier

🐍 Preparing Python 3.11

One thing that needed a little attention during setup was the Python version.

The current version, open-webui 0.8.12, supports Python 3.11 or newer, but below 3.13,
so Python 3.13 was not supported.

Because of that, I removed my existing Python 3.13.1 installation,
uninstalled it from the Windows Settings screen, and then prepared Python 3.11 using Python Install Manager.

✋ Installing Python Install Manager

I installed it from the Microsoft Store.

When it starts, it asks whether you want to add it to PATH, so I proceeded with Yes.

After that, I checked things in PowerShell or Command Prompt.

py list
py list --online
py install 3.11

Python 3.11 installation screen

At this point, Python 3.11 was ready.

📦 Installing Open WebUI

First, I created a working folder for the installation.

mkdir D:\AI\open-webui
cd D:\AI\open-webui

Next, I created a virtual environment using Python 3.11.

py -3.11 -m venv .venv

Then I activated it in PowerShell.

.\.venv\Scripts\Activate.ps1

Once activated, I installed Open WebUI.

python -m pip install --upgrade pip
pip install open-webui

To start it:

open-webui serve

Open WebUI running

Then I accessed it from a browser.

http://localhost:8080

Accessed from the browser

Connecting It to Ollama

After launching Open WebUI, I first created an admin account.

Admin account creation screen

At first, I confirmed that it was not yet connected from the model selection screen.

Model selection screen before connection

Then I changed the connection target from:

Settings > Manage Ollama API Connections

http://localhost:11434 → http://<host-pc-name>:11434

Ollama API connection settings

After that change, Open WebUI was able to connect to Ollama.

I then confirmed that qwen2.5:3b could be selected from the model list.

Model available after connection

I also entered a chat prompt and confirmed that it returned a response.

Chat response screen

🌐 Using It from Another PC on the Same LAN

Once everything above was working, the local LLM setup became available from another PC on the same LAN through a browser.

That was exactly what I wanted to achieve:

  • run Ollama on the main PC
  • use Open WebUI as the browser-based interface
  • access it from a working laptop or another machine

This made the environment much easier to use for experiments, because inference runs locally while the operation can happen from another device.

💾 Creating a Startup Batch File

Just like I do for Stable Diffusion and kohya_ss, I turned Open WebUI into a batch file so that it would be easier to launch from the VS Code terminal.

It can also be launched from PowerShell, but that sometimes involves extra permission-related setup, so using a .bat file felt simpler this time.

@echo off
REM ===== Open WebUI startup batch =====

REM Move to D drive
d:

REM Move to the Open WebUI folder
cd D:\AI\open-webui

REM Create .venv if it does not exist (only needed the first time)
if not exist .venv (
    py -3.11 -m venv .venv
)

REM Activate the virtual environment
call .venv\Scripts\activate.bat

REM Start Open WebUI
call .venv\Scripts\open-webui serve

pause

✍️ Notes This Time

After actually installing Open WebUI, I noticed that it has quite a lot of configuration options. It gave me a small glimpse of how many tuning points may also exist behind cloud-based generative AI services.

It also looks like Open WebUI can be configured to work together with Stable Diffusion, so that is something I would like to explore little by little in the future.