THE INTERFACE

Open http://localhost:3000 and let's take a tour. If you've used ChatGPT, most of this will feel familiar.

The Chat Area

The center of the screen is your conversation. Type a message at the bottom, hit Enter, and the model responds above. Streaming is enabled by default, so you'll see the response appear word by word.

At the top of the chat area, there's a model selector dropdown. Click it to switch between any model you've pulled. You can change models mid-conversation if you want.

The Sidebar

On the left is your conversation history. Every chat gets saved automatically. You can:

Rename conversations by clicking the title
Organize them into folders
Pin important ones to the top
Search across all your conversations
Delete ones you don't need

The sidebar also has a "New Chat" button at the top. Use it to start fresh conversations. Each chat has its own context and history.

Starting a Conversation

Select a model from the dropdown and type something. Try:

What can you tell me about the city I live in?

The model will probably ask which city. That's fine. It shows the conversation works and the model is generating.

A few things to notice as it responds:

The response streams in real time
You can stop generation mid-stream with the stop button
Markdown formatting renders automatically (headers, code blocks,
lists, bold, etc.)
Code blocks get a copy button

The Message Actions

Hover over any message (yours or the model's) and you'll see small action icons appear:

Copy:           Copies the message text to clipboard
Edit:           Edit your message and regenerate the response
Regenerate:     Get a different response to the same prompt
Good/Bad:       Rate the response (useful for your own reference)

The edit feature is particularly handy. If the model misunderstood you, click edit on your message, rephrase it, and it regenerates from that point. You don't have to start over.

Model Dropdown Details

Click the model selector at the top. You'll see all available models listed. A few things to notice:

The currently selected model has a checkmark
You can search by typing in the dropdown
Custom models you create in Workspace appear here too
You can select multiple models to compare responses side by side

That last one is slick. Select two or three models and send a message. OpenWebUI shows their responses in parallel columns so you can compare quality, style, and speed. Great for figuring out which model you prefer for different tasks.

File Uploads

Click the attachment icon (paperclip or plus sign) next to the message input. You can upload:

Documents:    PDF, Word, text files
Images:       If using a vision-capable model
Code files:   Any text-based file

For documents, OpenWebUI extracts the text and includes it in the conversation context. The model can then answer questions about the document.

For images, you need a vision model selected (like Qwen3-VL or Mistral Small 3.2). Upload an image, ask a question about it, and the model will analyze what it sees. We'll demo this in detail later.

The @ Symbol

In the chat input, type @ to reference specific items:

@model_name      Switch to a different model mid-conversation
@knowledge_base  Pull in a specific knowledge collection
@file            Reference an uploaded file

This is useful in longer conversations when you want to bring in additional context without navigating away from the chat.

Keyboard Shortcuts

A few worth knowing:

Enter:              Send message
Shift+Enter:        New line without sending
Ctrl+Shift+O:       New chat
Ctrl+Shift+S:       Toggle sidebar
Up Arrow:           Edit your last message (when input is empty)

These speed things up once you're comfortable with the interface.

The Settings Menu

Click your profile icon in the bottom left to access settings. Worth poking around in:

General:       Theme (dark mode!), language, notification prefs
Interface:     Chat display options, code highlighting
Audio:         Text-to-speech voice and speed
Connections:   Ollama URL configuration
Models:        Pull new models directly from the UI

You can pull new Ollama models from right here in the browser instead of going to the terminal. Navigate to Admin Panel, then Settings, then Models, and type a model name to pull it.

Next Up

That covers the basic interface. Now let's get into the really powerful stuff: creating custom models in the Workspace.