How to install Open WebUI on Windows with Ollama and uv

Set up a ChatGPT-style web app on your own Windows PC that runs local Ollama models and answers questions about your PDFs. It suits anyone who already has Ollama installed and wants a private chat interface without Docker or a cloud account. You need about 2 GB of free disk space.

Open WebUI Install: Your Own ChatGPT with Ollama, by Laxmi Kant | KGP Talkie on YouTube. 12 min 5 s.

18 chapters. Tap a step to jump to it.Chapters from the creator
Video length
12 min 5 s
Time to do it
20 minutes
Difficulty
Medium
Steps
10

What you need

Steps

  1. Open PowerShell and install uv with the official one-line install command from the uv docs. This also brings the Python 3.11 or 3.12 that Open WebUI needs, so you do not install Python yourself.

  2. Start Open WebUI with the uvx command, pointing it at a data folder on your disk. That folder keeps your chats and settings safe between restarts.

  3. Wait for the first start to finish. It downloads around 245 packages, about 1.7 GB, which takes roughly 1 minute 40 seconds on a normal connection.

  4. If the browser shows Cannot GET / on port 8080, another app is using that port. Start Open WebUI again on port 3000 and open localhost:3000 instead.

  5. Create the admin account on the sign-up screen. It is stored only on your computer, so pick a password you will remember.

  6. Check that Ollama connected by itself under Admin Panel, Settings, Connections. If it did not, add the Ollama address there manually.

  7. Switch off the OpenAI API connection in Settings to stop the red Missing bearer authentication errors, since you are not using OpenAI.

  8. Open the model picker, choose a model such as Qwen, and set it as the default. You can also download new Ollama models from the same picker.

  9. Send your first chat message to test the setup. On a strong graphics card you get fast replies; without one, use a small model like Qwen 3.5 4B.

  10. Upload a PDF in the chat and ask a question about it. The answer comes with sources you can check, and nothing leaves your computer.

Tips and common mistakes

Without a dedicated graphics card, pick a small model such as Qwen 3.5 4B so replies stay quick.

Keep your data folder in a fixed place, otherwise your chats and admin account disappear on the next start.

If port 8080 is busy, use 3000 from the start instead of hunting for the app that holds the port.

Update Open WebUI regularly by restarting the same uvx command, since the latest version fixes connection bugs.

Questions

Do I need Docker or Python installed first?

No. The uv route installs the right Python version for you, and Docker is only the alternative route in the official docs.

Can I use Open WebUI from my phone?

Yes, as long as the phone is on the same network. Open the PC's local IP address with the port in the phone browser.

Is anything sent to the cloud?

No. Models run through Ollama on your machine and chats are stored in your local data folder.

This guide was drafted from the video's details and general task knowledge. It has not been checked against the full video. Check the video and relevant product instructions before following a procedure. Report a correction.

More in Computers & Phones

See all