How to Install Open WebUI - An Opensource Web Interface for Running LLMs

Introduction

Open WebUI is a self-hosted, feature-rich, and user-friendly open-source web interface that lets you run large language models (LLMs) using Ollama and OpenAI-compatible APIs. You can extend Open WebUI with AI models and plugins to suit your needs.

This article explains how to install Open WebUI and run large language models (LLMs) using Ollama on a Vultr Cloud GPU instance. You will also configure Open WebUI with secure SSL certificates to expose the web interface using a custom domain.

Prerequisites

Before you begin:

• Deploy a Vultr Cloud GPU instance with at least 8 GB RAM to run LLMs. Choose:

  • 8 GB for 7B models
  • 16 GB for 13B models
  • 32 GB for 33B models.

• Set up a domain A recording pointing to the instance's public IP address. For example, openwebui.example.com.

• Access the instance using SSH as a non-root sudo user.

Install Ollama

Ollama is a lightweight, extensible framework for running open-source LLMs such as Llama, Code Llama, Mistral, and Gemma. Ollama integrates with application frameworks using APIs to create, manage, and customize models. Open WebUI requires Ollama to run and manage LLMs. Follow the steps below to install Ollama and download a sample model to run using Open WebUI.

1. Download the Ollama installation script.

   CONSOLE

   Copy

   $ wget https://ollama.ai/install.sh

   Explain Code

   Powered by Vultr AgentBeta

2. Grant execute permissions to the script.

   CONSOLE

   Copy

   $ sudo chmod +x install.sh

   Explain Code

   Powered by Vultr AgentBeta

3. Run the script to install Ollama.

   CONSOLE

   Copy

   $ sudo ./install.sh

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   >>> Installing ollama to /usr/local
   >>> Downloading Linux amd64 bundle
   ############################################################################################################### 100.0%>>> Creating ollama user...
   >>> Adding ollama user to render group...
   >>> Adding ollama user to video group...
   >>> Adding current user to ollama group...
   >>> Creating ollama systemd service...
   >>> Enabling and starting ollama service...
   Created symlink /etc/systemd/system/default.target.wants/ollama.service → /etc/systemd/system/ollama.service.
   >>> NVIDIA GPU installed.

4. Enable the Ollama system service to start at boot.

   CONSOLE

   Copy

   $ sudo systemctl enable ollama

   Explain Code

   Powered by Vultr AgentBeta

5. Start the Ollama service.

   CONSOLE

   Copy

   $ sudo systemctl start ollama

   Explain Code

   Powered by Vultr AgentBeta

6. View the Ollama service status and verify it's running.

   CONSOLE

   Copy

   $ sudo systemctl status ollama

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   ● ollama.service - Ollama Service
        Loaded: loaded (/etc/systemd/system/ollama.service; enabled; vendor preset: enabled)
        Active: active (running) since Sun 2024-11-17 11:51:45 UTC; 28s ago
      Main PID: 18515 (ollama)
         Tasks: 11 (limit: 9388)
        Memory: 70.2M
           CPU: 699ms
        CGroup: /system.slice/ollama.service
                └─18515 /usr/local/bin/ollama serve

7. Use Ollama to download a sample model such as llama3:8b.

   CONSOLE

   Copy

   $ sudo ollama pull llama3:8b

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   pulling manifest 
   pulling 6a0746a1ec1a... 100% ▕██████████████████████████████████████████████████████▏ 4.7 GB                         
   pulling 4fa551d4f938... 100% ▕██████████████████████████████████████████████████████▏  12 KB                         
   pulling 8ab4849b038c... 100% ▕██████████████████████████████████████████████████████▏  254 B                         
   pulling 577073ffcc6c... 100% ▕██████████████████████████████████████████████████████▏  110 B                         
   pulling 3f8eb4da87fa... 100% ▕██████████████████████████████████████████████████████▏  485 B                         
   verifying sha256 digest 
   writing manifest 
   success 

Install Open WebUI

You can install Open WebUI using Python Pip or Docker. The Pip method requires the Python version 3.11 while Docker uses a prebuilt Open WebUI container image. Follow the steps below to install Open WebUI using both methods.

Option 1: Install Open WebUI Using Pip

1. Install Python 3.11 version.

   CONSOLE

   Copy

   $ sudo apt install python3.11

   Explain Code

   Powered by Vultr AgentBeta

2. Install Open WebUI.

   CONSOLE

   Copy

   $ sudo python3.11 -m pip install open-webui

   Explain Code

   Powered by Vultr AgentBeta

3. Upgrade pillow and pyopenssl modules.

   CONSOLE

   Copy

   $ sudo python3.11 -m pip install -U Pillow pyopenssl

   Explain Code

   Powered by Vultr AgentBeta

4. Run Open WebUI and verify it does not return errors.

   CONSOLE

   Copy

   $ sudo open-webui serve

   Explain Code

   Powered by Vultr AgentBeta

   Output:

     / _ \ _ __   ___ _ __   \ \      / /__| |__ | | | |_ _|
    | | | | '_ \ / _ \ '_ \   \ \ /\ / / _ \ '_ \| | | || |
    | |_| | |_) |  __/ | | |   \ V  V /  __/ |_) | |_| || |
     \___/| .__/ \___|_| |_|    \_/\_/ \___|_.__/ \___/|___|
          |_|
   
   
   v0.4.6 - building the best open-source AI user interface.
   
   https://github.com/open-webui/open-webui
   
   INFO:     Started server process [16779]
   INFO:     Waiting for application startup.
   INFO:     Application startup complete.
   INFO:     Uvicorn running on http://0.0.0.0:8080 (Press CTRL+C to quit)

   Press Ctrl + C to stop the application.

Create a System Service to Run Open WebUI

If you install Open WebUI using Pip, the application runs using the open-webui serve command. You should create a system service to manage the Open WebUI processes without running the open-webui serve command directly. Follow the steps below.

1. Create a new /usr/lib/systemd/system/openwebui.service system service file using a text editor like vim.

   CONSOLE

   Copy

   $ sudo vim /usr/lib/systemd/system/openwebui.service

   Explain Code

   Powered by Vultr AgentBeta

2. Add the following service configurations to the /usr/lib/systemd/system/openwebui.service file.

   INI

   Copy

   [Unit]
   Description=Open WebUI Service
   After=network.target
   
   [Service]
   Type=simple
   ExecStart=open-webui serve
   ExecStop=/bin/kill -HUP $MAINPID
   
   [Install]
   WantedBy=multi-user.target

   Explain Code

   Powered by Vultr AgentBeta

   Save and close the file.

   In the above system service configuration, starting the Open WebUI service automatically runs the open-webui serve command on your server.

3. Reload systemd to apply the new service configuration.

   CONSOLE

   Copy

   $ sudo systemctl daemon-reload

   Explain Code

   Powered by Vultr AgentBeta

4. Enable the Open WebUI system service to start at boot.

   CONSOLE

   Copy

   $ sudo systemctl enable openwebui.service

   Explain Code

   Powered by Vultr AgentBeta

5. Start the Open WebUI service.

   CONSOLE

   Copy

   $ sudo systemctl start openwebui

   Explain Code

   Powered by Vultr AgentBeta

6. Test the Open WebUI service status and confirm it's running.

   CONSOLE

   Copy

   $ sudo systemctl status openwebui

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   ● openwebui.service - Open WebUI Service
        Loaded: loaded (/lib/systemd/system/openwebui.service; enabled; vendor preset: enabled)
        Active: active (running) since Wed 2024-12-04 20:20:29 UTC; 6s ago
      Main PID: 3007 (open-webui)
         Tasks: 9 (limit: 14212)
        Memory: 522.7M
           CPU: 6.456s
        CGroup: /system.slice/openwebui.service
                └─3007 /usr/bin/python3.11 /usr/local/bin/ open-webui serve

Option 2: Install Open WebUI Using Docker

1. Verify that Docker is running on your server.

   CONSOLE

   Copy

   $ sudo docker ps

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   CONTAINER ID   IMAGE     COMMAND   CREATED   STATUS    PORTS     NAMES

   Run sudo systemctl start docker to start Docker in case you receive the following error.

   Cannot connect to the Docker daemon at unix:///var/run/docker.sock. Is the docker daemon running?

2. Verify that a GPU device is available on your server.

   CONSOLE

   Copy

   $ nvidia-smi

   Explain Code

   Powered by Vultr AgentBeta

3. Pull the Open WebUI Docker image for Nvidia GPUs to your server.

   CONSOLE

   Copy

   $ sudo docker pull ghcr.io/open-webui/open-webui:cuda

   Explain Code

   Powered by Vultr AgentBeta

   To use Open WebUI on a non-GPU server, download the main Docker image instead.

   CONSOLE

   Copy

   $ sudo docker pull ghcr.io/open-webui/open-webui:main

   Explain Code

   Powered by Vultr AgentBeta

4. Run Open WebUI and use all Nvidia GPU devices on the host server.

   CONSOLE

   Copy

   $ sudo docker run -d -p 8080:8080 --gpus all --add-host=host.docker.internal:host-gateway -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:cuda

   Explain Code

   Powered by Vultr AgentBeta

   In the above command, Docker runs Open WebUI using the open-webui:cuda container image you downloaded earlier with the following options:

   • -p 8080:8080: Maps the host port 8080 to the Open WebUI container port 8080 for access on your server.
   • --gpus all: Enables Open WebUI to access all GPUs available on your server.
   • --add-host=host.docker.internal:host-gateway: Enables the Open WebUI container to communicate with services outside Docker that are installed on the server.
   • -v open-webui:/app/backend/data: Creates a new open-webui volume and mounts it to /app/backend/data inside the Open WebUI container.
   • --name open-webui-new: Sets the Open WebUI container name for visibility and management.
   • --restart always: Enables the Open WebUI container to automatically restart in case the container stops unexpectedly.
   • ghcr.io/open-webui/open-webui:cuda: Specifies the Open WebUI docker image to run.

5. Ensure that the Open WebUI Docker container is running.

   CONSOLE

   Copy

   $ sudo docker ps

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   CONTAINER ID   IMAGE                                COMMAND           CREATED          STATUS                             PORTS                                       NAMES
   d7957cd20569   ghcr.io/open-webui/open-webui:cuda   "bash start.sh"   12 seconds ago   Up 12 seconds (health: starting)   0.0.0.0:8080->8080/tcp, :::8080->8080/tcp   open-webui

Configure Nginx as a Reverse Proxy to Run Open WebUI

You can access the Open WebUI on the default port 8080 using your server's IP address or linked domain. Exposing the Open WebUI port in a production environment is insecure. You can use Nginx as a reverse proxy to handle secure connections to the Open WebUI's backend port using HTTP or HTTPS. In the following steps, install Nginx and create a new virtual host configuration to access Open WebUI on your server.

1. Install Nginx.

   CONSOLE

   Copy

   $ sudo apt install nginx -y

   Explain Code

   Powered by Vultr AgentBeta

2. Start the Nginx system service.

   CONSOLE

   Copy

   $ sudo systemctl start nginx

   Explain Code

   Powered by Vultr AgentBeta

   If you receive an error, stop Apache or any other application using the HTTP port 80 on your server to enable Nginx to run.

3. Create a new /etc/nginx/sites-available/openwebui.conf virtual host for Open WebUI.

   CONSOLE

   Copy

   $ sudo vim /etc/nginx/sites-available/openwebui.conf 

   Explain Code

   Powered by Vultr AgentBeta

4. Add the following Nginx configurations to the /etc/nginx/sites-available/openwebui.conf file.

   NGINX

   Copy

   server {
       listen 80;
       server_name openwebui.example.com;
   
       access_log /var/log/nginx/openwebui_access.log;
       error_log /var/log/nginx/openwebui_error.log;
   
       location / {
           proxy_pass http://127.0.0.1:8080; 
           proxy_set_header Host $host;
           proxy_set_header X-Real-IP $remote_addr;
           proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
           proxy_set_header X-Forwarded-Proto $scheme;
       }
   }

   Explain Code

   Powered by Vultr AgentBeta

   Save the file.

5. Link the openwebui.conf configuration to the sites-enabled directory to enable it.

   CONSOLE

   Copy

   $ sudo ln -s /etc/nginx/sites-available/openwebui.conf /etc/nginx/sites-enabled/openwebui.conf 

   Explain Code

   Powered by Vultr AgentBeta

6. Test Nginx for configuration errors.

   CONSOLE

   Copy

   $ sudo nginx -t

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   nginx: the configuration file /etc/nginx/nginx.conf syntax is ok
   nginx: configuration file /etc/nginx/nginx.conf test is successful

7. Restart Nginx to apply the Open WebUI virtual host configuration changes.

   CONSOLE

   Copy

   $ sudo systemctl restart nginx

   Explain Code

   Powered by Vultr AgentBeta

8. Allow HTTP connections through the firewall.

   CONSOLE

   Copy

   $ sudo ufw allow http

   Explain Code

   Powered by Vultr AgentBeta

Generate SSL Certificates to Secure Open WebUI

SSL certificates encrypt the connection between a client's web browser and your server. Use a trusted certificate authority (CA) like Let's Encrypt to generate SSL certificates and secure connections to the Open WebUI using your Nginx configuration. In the following steps, install the Certbot Let's Encrypt client and generate SSL certificates to access Open WebUI using your domain.

1. Install Certbot for Nginx.

   CONSOLE

   Copy

   $ sudo apt install python3-certbot-nginx -y

   Explain Code

   Powered by Vultr AgentBeta

2. Generate a new SSL certificate for your domain. Replace openwebui.example.com with your domain and admin@example.com with an active email address.

   CONSOLE

   Copy

   $ sudo certbot --nginx -d openwebui.example.com -m admin@example.com --agree-tos

   Explain Code

   Powered by Vultr AgentBeta

   Output:

   Successfully received certificate.
   Certificate is saved at: /etc/letsencrypt/live/openwebui.example.com/fullchain.pem
   Key is saved at:         /etc/letsencrypt/live/openwebui.example.com/privkey.pem
   This certificate expires on 2025-02-25.
   These files will be updated when the certificate renews.
   Certbot has set up a scheduled task to automatically renew this certificate in the background.
   
   Deploying certificate
   Successfully deployed certificate for openwebui.example.com to /etc/nginx/sites-enabled/openwebui.conf
   Congratulations! You have successfully enabled HTTPS on https://openwebui.example.com

3. Restart Nginx to apply the SSL configuration changes.

   CONSOLE

   Copy

   $ sudo systemctl restart nginx

   Explain Code

   Powered by Vultr AgentBeta

4. Allow HTTPS connections through the firewall.

   CONSOLE

   Copy

   $ sudo ufw allow https

   Explain Code

   Powered by Vultr AgentBeta

Access Open WebUI

You can access Open WebUI using your domain based on the configuration steps you performed earlier. In the following steps, access Open WebUI and create a new administrator account to run LLMs on the server.

1. Access your Open WebUI's domain to access the web interface.

   https://openwebui.example.com

2. Click Get started to access Open WebUI.

   

3. Enter a new administrator username, email address, and password in the Open WebUI fields.

   

4. Click Create Admin Account to create the first Open WebUI administrator account.

5. Confirm that the Open WebUI page displays in your browser, click the Arena Model drop-down in the top menu, and select the Llama 8b model you installed earlier.

   

6. Enter a new prompt like Hello today! Write me a simple poem to generate a result using the default Llama 8b model you installed earlier.

   

Install and Run LLMs in Open WebUI

Open WebUI ships with some default AI models. You can also install new models directly using Ollama on the server or by uploading the model files through the Open WebUI interface. In the following steps, install new models using Ollama and run them using Open WebUI.

1. Visit the Ollama library and find new models.

2. Download a new model. For example, the mistral 7B model.

   CONSOLE

   Copy

   $ sudo ollama pull mistral

   Explain Code

   Powered by Vultr AgentBeta

3. Download another model, such as llama3:70b.

   CONSOLE

   Copy

   $ sudo ollama pull llama3:70b

   Explain Code

   Powered by Vultr AgentBeta

   The llama3:70b model is 40GB and requires more RAM to run on the server.

4. Restart Open WebUI to synchronize the new model changes.

   CONSOLE

   Copy

   $ sudo docker restart open-webui

   Explain Code

   Powered by Vultr AgentBeta

   Or, restart the Open WebUI system service.

   CONSOLE

   Copy

   $ sudo systemctl restart openwebui

   Explain Code

   Powered by Vultr AgentBeta

5. Access Open WebUI in a new web browser.

   https://openwebui.example.com

6. Click the models list on the top left drop-down list and verify the new models are available.

   

7. Select a model to use in Open WebUI. For example, select the Mistral model you installed.

8. Enter a new prompt in the input field, such as What is Vultr? and verify that the model processes new results.

   

Conclusion

Open WebUI is a powerful web interface for running ML LLMs on a server. In this guide, you installed Open WebUI on a Vultr Cloud GPU instance and secured it with SSL certificates. You can customize the Open WebUI interface and create multiple users to log in or sign up to use LLMs directly on your server. For additional guidelines, please refer to the Open WebUI documentation.