Open WebUI getting started

Open WebUI is a self-hosted web interface that allows you to work with Large Language Models (LLMs) via your browser. Open WebUI is a 'ChatGPT-like' interface: the model itself runs at an external provider such as OpenAI or Anthropic, or locally via e.g. Ollama. 

If you ordered a pre-installed VPS with Open WebUI, you no longer have to install Open WebUI yourself. In this manual you continue from the moment your VPS has been delivered: you log in for the first time, create your administrator account and then link a model provider so that you can start immediately.

Logging in to Open WebUI for the first time

Step 1

Please refer before proceeding a (sub)domain to the IP address of your VPS. Do you host your domain with TransIP and create a new DNS record for Open WebUI? This change will then be processed within a few minutes. 

Step 2

Connect to your VPS via the VPS console in your control panel or via SSH. Have you chosen a one-time password as the installation method? First change the password after logging in. 

You will now see (approximately) the following screen:

Enter the (sudo) password of your user account: the necessary ports will now be opened automatically in your firewall. 

Step 3

Open the file /opt/open-webui/.env:

sudo nano /opt/open-webui/.env

Adjust the following values:

• DOMAIN_NAME: the domain name that you link to Open WebUI, for example example.nl. Even if you want to use a subdomain, only enter the 'fully qualified domain name' (FQDN) here.
• SUBDOMAIN: the subdomain name without the FQDN, for example webui.
• SSL_EMAIL: the email address at which you want to receive emails regarding the renewal of your SSL certificates.

WEBUI_SECRET_KEY=e1d5e3d982ec0aea3b88e249681a5ca8
DOMAIN_NAME=voorbeeld.nl
SUBDOMAIN=webui
SSL_EMAIL=demo@voorbeeld.nl

Save the changes and close the file (ctrl + X > y > enter). 

Step 4

Reload Open WebUI to process the changes: 

sudo systemctl restart open-webui

Now give your server a minute or two to process the changes and associate an SSL certificate before moving on to step 5.

Step 5

In your browser, open the URL of your Open WebUI installation. To do this, use the IP address that you received with your VPS or the (sub)domain linked in steps 1 to 5. You will now see a screen like the one below. Click on 'Get started’.

Step 6

First create your first account. This first account is your admin account and has access to Open WebUI's global settings, including adding model providers. Enter your name, email address and password and click 'Create Admin Account’.

Step 7

You will now see the release notes of the installed version. Click on 'Okay, Let’s Go!' to actually start with Open WebUI.

Step 8

After logging in, click on 'Select a model’ > ‘Manage Connections'. Here you add the LLM providers that Open WebUI is allowed to use.

Step 9

You will now configure a LLM to use with Open WebUI. Simply put, there are two options:

• A local model use via Ollama on your VPS (pre-installed). 
  • Advantage: Your data remains local. You are 100% free to choose which model you use. For example, a good, optimized choice is Gemma 4 (or newer).
  • Disadvantage: Usually less good performance than an external model. There are exceptions, but that is use case specific and then you will quickly need approximately 256GB RAM and 24GB GPU to host it locally, which increases the costs compared to a ChatGPT subscription.
     
• An external model use, for example Claude, GPT, or external (own) Ollama server.
  • Advantage: Access to state-of-the-art models. 
  • Disadvantage: Your data goes to an external party. Even if that party has the best intentions, geopolitical factors can influence whether your data remains private or not.

Using a local model

Look at here the available models. For each model you will find under 'size' how much RAM you need to load the model into memory.

 

 

Step 1

Click on the 'Manage' icon to the right of http://ollama:11434. 

---

 

Step 2

Enter the name of a model of your choice (1) and click on 'Pull Model' (2). For example, on a small VPS (4GB or less), gemma4:e2b is a good choice. Then close this window via the 'X' top right.

Optionally update the available models via 'Update All Models' (3) and view the overview of all available models via 'click here’ (4).

 

 

 

Using an external model

Click on one of the following:

1. + behind 'Manage OpenAI API Connections' to create a OpenAI compatible endpoint API connection
2. + behind 'Manage Ollama API Connections' to connect to a local or remote Ollama server.
3. Toggle optional 'Direct Connections' in order users to allow connections to their own OpenAI compatible API endpoint.

OpenAI compatible endpoint API connection

1. Enter the address as URL https://api.openai.com/v1/models or the OpenAI compatible endpoint address of your chosen LLM provider.
2. Enter your API key after 'Bearer’.
3. Click on 'Save' to save your changes.

By not specifying a model you get access to all available OpenAI models.

 

 

Ollama API connection

1. Enter the IP address of your external Ollama server as URL with the addition :11434, see the screenshot.
2. Click on 'Save' to save your changes.

Not specifying a model gives you access to all available models on your Ollama server.

 

 

 

 

After saving, the available models will appear in the model selector of Open WebUI. Select the model you want to work with.

Start your first chat

After adding a LLM you can get started right away and chat with one of your added models.

Step 1

Choose the chat interface at the top left and select the model you want to use.

Step 2

Start a new chat and ask a simple test question, for example: 

Open WebUI is now ready to use. You can then continue adding users, documents, tools, and other integrations.

(admin) adjust settings

Adjust user settings

Step 1

You can adjust the general settings of Open WebUI by clicking on the icon with the letters of your first and last name at the bottom left and 'Settings' to select.

Step 2

Adjust the settings in this screen and the other tabs as desired. These settings only apply to your own account and not to other users of your Open WebUI environment. Some useful settings to adjust include the System Prompt ("You are a helpful assistant") and the Audio settings (speech to text and text to speech).

Adjust admin settings

The admin settings are hidden in the normal settings menu. 

Click on the icon at the bottom left with the letters of your first and last name and select 'Admin Panel’.

Then open the tab 'Settings' at the top and adjust the settings as desired. From top to bottom you can manage the following options:

• General: Contains the general platform settings. Here you configure basic behavior of the environment, authentication settings, default roles and groups, registration of new users, contact information for administration, API key usage and session or token settings, among other things. 
• Connections: Manage connections to external services and LLM providers. 
• Models: Set which models are available, how they are presented to users, and what default or access settings apply per model. 
• Evaluations: Configure assessments, manage test scenarios and adjust settings for quality control or benchmarking. 
• Integrations: Manages integrations with other tools, platforms or services that add additional functionality to Open WebUI. 
• Documents: Contains settings for configuring RAG tools from Open WebUI. Here you determine how documents are added, processed, indexed or made available for search and AI functionality. Web Search Controls the settings for web search functionality. Here an admin configures whether web search is available, which search provider is used and which settings apply to online searches. 
• Code Execution: Contains settings for executing code within the platform (by default this is on and Python is used). Here, an admin determines whether code execution is allowed and how that environment is set up, for example in terms of security, access and execution behavior. 
• Interface: Manage the visual components of the UI, such as display settings, branding, theming, and general interface preferences. 
• Audio: Contains settings for audio-related functions: speech-to-text, text-to-speech and any associated audio providers. 
• Images: Adjust settings for generating and editing images, for example what system you use for this (for example OpenAI).
• Pipelines: Here an admin manages pipelines that combine multiple steps or services, for example for processing, routing or automation.
• Database: Back up your database, conversations, or import your database.

Use audio

In the audio menu of both the user and admin you will find settings for speech-to-text and text-to-speech. Simply put, the admin sets which options are available in addition to the default options and the user chooses which of the available options he/she wants to use.

User settings

Simply adjust which STT and TTS tools you use. Are you using Kokoro.js (provided by your browser)? Then select fp16 as 'Dtype' for a balance between performance and accuracy. You will get the best results with an external model, for example Elevenlabs for TTS.

Admin settings

Customize which STT and TTS tools you make available to your end users. The results with the built-in Whisper (Local) and Web API options are quite mediocre compared to external options, especially on a server without GPU. A solution to this is to create a OpenAI compatible endpoint for STT itself (for example on the same server) and then use it as 'Speech to Text Engine' OpenAI and then point to your server. For TTS, for example, you can use Kokoro yourself instead of using the browser version, see for example this manual. 

Troubleshooting

• You don't see any models in the list: check if the connection is saved and enabled.
• OpenAI or Anthropic gives an error message: check whether your API key is correct and whether your account or budget is active.
• Ollama is not reachable: check that the Ollama service is running and that Open WebUI is using the correct address and port number.
• The correct models are missing: if necessary, add a model manually or check for a filter Model IDs is active.

This brings us to the end of our guide to Open WebUI. For a complete overview of all the ins and outs and features of Open WebUI, take a look at Open WebUI's own documentation.