How to Deploy Open WebUI on a VPS: Self-Hosted ChatGPT Interface for Ollama and OpenAI

By VPS.DO
August 20, 2026

Open WebUI is a feature-rich, self-hosted web interface for large language models — it works with Ollama (local models) and the OpenAI API simultaneously, supports multi-user access with role-based permissions, enables RAG (Retrieval Augmented Generation) with document uploads, integrates web search, and provides conversation history, model management, and system prompts. Deploy it on a VPS to give your team a private ChatGPT-like experience backed by local or cloud models.

What Open WebUI Provides

Multi-model interface: Switch between Ollama local models and OpenAI/Claude/Gemini APIs in one UI
Multi-user: User registration, role-based access (admin, user), per-user conversation history
RAG: Upload PDFs, Word docs, or URLs — the model answers questions based on your documents
Web search: Integrate SearXNG or web search APIs — models answer with current information
Image generation: Connect to DALL-E, Stable Diffusion, or ComfyUI
Tools/Functions: Extend models with custom Python functions called as tools
Prompt templates: Save and share system prompts for consistent model behavior
API: OpenAI-compatible API — use Open WebUI as a gateway for your applications

Prerequisites

Ollama installed on the VPS (see the Ollama guide — Post 114), or an OpenAI API key
Docker and Docker Compose
2+ GB RAM for the WebUI itself (plus model RAM if using Ollama)

Step 1: Docker Compose Setup

<code">mkdir -p /opt/open-webui && cd /opt/open-webui
nano docker-compose.yml

<code">version: '3.8'

services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: open-webui
    restart: always
    ports:
      - "127.0.0.1:3000:8080"
    environment:
      # Connect to Ollama on the host machine
      OLLAMA_BASE_URL: http://host.docker.internal:11434

      # Optional: Add OpenAI API key for GPT-4, Claude, etc.
      OPENAI_API_KEY: ${OPENAI_API_KEY:-}

      # Enable user registration (set to false after creating admin)
      ENABLE_SIGNUP: "true"

      # Default admin email (first user to register with this email becomes admin)
      DEFAULT_USER_ROLE: pending   # Require admin approval for new users

      # RAG configuration
      ENABLE_RAG_WEB_SEARCH: "true"
      RAG_WEB_SEARCH_ENGINE: searxng
      SEARXNG_QUERY_URL: http://searxng:8080/search?q=&format=json

    volumes:
      - open_webui_data:/app/backend/data
    extra_hosts:
      - "host.docker.internal:host-gateway"  # Access host machine's Ollama

  # Optional: SearXNG for web search integration
  searxng:
    image: searxng/searxng:latest
    container_name: searxng
    restart: always
    ports:
      - "127.0.0.1:8888:8080"
    volumes:
      - searxng_data:/etc/searxng
    environment:
      SEARXNG_SECRET_KEY: your_searxng_secret

volumes:
  open_webui_data:
  searxng_data:

<code">nano .env

<code">OPENAI_API_KEY=sk-your-openai-api-key-here   # Optional

<code">chmod 600 .env
docker compose up -d
docker compose logs -f open-webui

Step 2: Nginx Reverse Proxy

<code">sudo nano /etc/nginx/sites-available/open-webui

<code">server {
    listen 80;
    server_name chat.yourdomain.com;
    return 301 https://$host$request_uri;
}

server {
    listen 443 ssl http2;
    server_name chat.yourdomain.com;

    ssl_certificate /etc/letsencrypt/live/chat.yourdomain.com/fullchain.pem;
    ssl_certificate_key /etc/letsencrypt/live/chat.yourdomain.com/privkey.pem;

    client_max_body_size 50M;   # Allow large document uploads for RAG

    location / {
        proxy_pass http://127.0.0.1:3000;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection 'upgrade';
        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;

        # Long timeout for model generation responses
        proxy_read_timeout 600s;
        proxy_buffering off;
    }
}

<code">sudo ln -s /etc/nginx/sites-available/open-webui /etc/nginx/sites-enabled/
sudo nginx -t && sudo systemctl reload nginx
sudo certbot --nginx -d chat.yourdomain.com

Step 3: Initial Admin Setup

Visit https://chat.yourdomain.com
Register the first account — this automatically becomes the admin
Admin Panel → Users → change pending users to “User” role to grant access
Admin Panel → Settings → disable “Enable New User Registration” if you want invite-only

Step 4: Add Model Connections

Ollama (Local Models)

Admin Panel → Connections → Ollama → verify URL: http://host.docker.internal:11434
Models list auto-populates from Ollama’s installed models

OpenAI API

Admin Panel → Connections → OpenAI → add API key
Now GPT-4o, GPT-4o-mini, o1 appear in the model selector

Other APIs (Anthropic Claude, Google Gemini)

Admin Panel → Connections → OpenAI-Compatible APIs
Add: Name: Claude, URL: https://api.anthropic.com/v1, Key: your Anthropic key
Model IDs: claude-3-5-sonnet-20241022

Step 5: Configure RAG with Document Upload

Start a new conversation → click the paperclip icon → upload PDF, Word, or TXT file
The document is chunked, embedded, and stored in a vector database
Ask questions: “What does section 3.2 say about refunds?” — model answers from the document

Configure embedding model (better RAG quality):

Admin Panel → Documents → Embedding Model
Select: nomic-embed-text (from Ollama) or text-embedding-3-small (from OpenAI)

Step 6: Web Search Integration

<code"># Configure SearXNG to allow JSON API
docker exec -it searxng nano /etc/searxng/settings.yml

<code"># Add to settings.yml:
server:
    secret_key: "your_searxng_secret"

search:
    formats:
        - html
        - json   # Enable JSON format for Open WebUI integration

<code">docker compose restart searxng

In a conversation, toggle the web search icon to enable real-time web search augmentation.

Step 7: System Prompts and Model Templates

<code"># Example system prompts to save as templates:

# Customer support agent:
"You are a helpful customer support agent for Acme Corp.
Be concise, professional, and empathetic.
If you cannot help, say 'I'll escalate this to our team.'
Never share pricing without first checking the current catalog."

# Code review assistant:
"You are a senior software engineer reviewing code.
Focus on: security vulnerabilities, performance issues, readability, and test coverage.
Provide specific, actionable feedback with examples."

Getting Started

Open WebUI needs 500 MB–1 GB RAM on its own. Combined with Ollama running a 7B model (6 GB RAM), a VPS at VPS.DO with 8 GB RAM handles the full stack. For teams using OpenAI or Anthropic APIs (no local model), a 2 GB VPS is sufficient — Open WebUI is just a frontend that proxies API calls. The NVMe storage speeds up document embedding for RAG use cases.

Conclusion

Open WebUI provides a complete, self-hosted AI workspace: a polished chat interface, multi-model access (local Ollama models + OpenAI/Anthropic APIs), RAG with document uploads, web search integration, multi-user access control, and conversation history — all on your VPS with no data sent to third-party AI dashboard services. For teams that want a ChatGPT Team-like experience with their own models and data policies, Open WebUI on a VPS is the definitive self-hosted solution.

How to Deploy Open WebUI on a VPS: Self-Hosted ChatGPT Interface for Ollama and OpenAI