TL;DR — Quick Summary

Deploy Open WebUI for a ChatGPT-like experience with Ollama, OpenAI, or any OpenAI-compatible API. Covers Docker setup, RAG, multi-user RBAC, web search, and admin configuration.

What Is Open WebUI?

Open WebUI (formerly Ollama WebUI) is a feature-rich, self-hosted web interface that provides a ChatGPT-like experience for interacting with AI language models. It works with Ollama for local models, OpenAI for cloud models, or any OpenAI-compatible API — all from a single, polished interface.

Key features:

  • ChatGPT-like UI — conversations, message editing, regeneration, branching
  • Multiple model backends — Ollama, OpenAI, Anthropic, any compatible API
  • RAG (Retrieval-Augmented Generation) — upload documents and chat with them
  • Web search — augment responses with live web results
  • Multi-user RBAC — admin/user roles, model access control, shared conversations
  • Model presets — save system prompts, temperature, and parameters as reusable profiles
  • Image generation — integrate with AUTOMATIC1111, ComfyUI, or OpenAI DALL-E
  • Voice input/output — speech-to-text and text-to-speech support
  • Markdown rendering — code blocks, tables, LaTeX math, Mermaid diagrams
  • API access — programmatic access to conversations and models

Prerequisites

  • Docker installed on your server or local machine.
  • Ollama running locally or on a remote server (see Ollama setup guide).
  • At least 1 GB RAM for Open WebUI itself (models use separate memory via Ollama).
  • Optional: OpenAI API key if you want to use GPT-4/GPT-4o alongside local models.

Installation

Standard (Ollama on Same Machine)

docker run -d -p 3000:8080 \
  --add-host=host.docker.internal:host-gateway \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

The --add-host flag allows the container to reach Ollama on the host machine via host.docker.internal.

Ollama on a Remote Server

If Ollama runs on a different machine (e.g., a GPU server):

docker run -d -p 3000:8080 \
  -e OLLAMA_BASE_URL=http://192.168.1.100:11434 \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Replace 192.168.1.100 with your Ollama server’s IP.

Docker Compose

services:
  open-webui:
    image: ghcr.io/open-webui/open-webui:main
    container_name: open-webui
    restart: always
    ports:
      - "3000:8080"
    extra_hosts:
      - "host.docker.internal:host-gateway"
    volumes:
      - open-webui:/app/backend/data
    environment:
      - OLLAMA_BASE_URL=http://host.docker.internal:11434
      # Optional: OpenAI integration
      # - OPENAI_API_BASE_URL=https://api.openai.com/v1
      # - OPENAI_API_KEY=sk-your-key-here

volumes:
  open-webui:

Initial Setup

First User = Admin

Open http://localhost:3000 and create an account. The first user automatically becomes the administrator. Subsequent users who sign up get the default “user” role.

Configure Connections

Navigate to Admin Panel → Settings → Connections:

BackendURLNotes
Ollamahttp://host.docker.internal:11434Auto-detected for local Ollama
OpenAIhttps://api.openai.com/v1Requires API key
CustomAny OpenAI-compatible URLe.g., Anthropic, Groq, Together AI

Manage User Registration

By default, anyone can create an account. To restrict registration:

  1. Go to Admin Panel → Settings → General.
  2. Toggle Enable New Sign Ups off.
  3. Now only the admin can create new accounts manually.

RAG — Chat with Your Documents

One of Open WebUI’s most powerful features is RAG (Retrieval-Augmented Generation). Upload documents and ask questions about them — the AI retrieves relevant passages and generates answers grounded in your data.

How to Use RAG

  1. In any conversation, click the paperclip icon (📎) or drag and drop a file.
  2. Supported formats: PDF, TXT, DOCX, XLSX, CSV, MD, HTML.
  3. Open WebUI automatically:
    • Chunks the document into segments
    • Creates embeddings using a local model (e.g., nomic-embed-text via Ollama)
    • Stores embeddings in its built-in vector database (ChromaDB)
  4. Ask questions about the document — the AI cites relevant passages.

Document Collections

For persistent document libraries:

  1. Go to Workspace → Documents.
  2. Upload files to create a collection.
  3. Reference the collection in any conversation with #collection-name.

Tip: For best RAG results, use nomic-embed-text as the embedding model: ollama pull nomic-embed-text. Configure it in Admin Panel → Settings → Documents → Embedding Model.

Web Search Integration

Open WebUI can augment model responses with live web search results:

  1. Go to Admin Panel → Settings → Web Search.
  2. Enable web search and configure a search provider:
    • SearXNG (self-hosted, recommended for privacy)
    • Google PSE (requires API key)
    • Brave Search (requires API key)
    • DuckDuckGo (no key required)
  3. In conversations, toggle the web search icon (🌐) to include web results.

Model Presets (Custom System Prompts)

Create reusable model configurations:

  1. Go to Workspace → Models.
  2. Click Create a Model.
  3. Configure:
    • Name: e.g., “DevOps Assistant”
    • Base Model: Select from available Ollama/OpenAI models
    • System Prompt: Custom instructions (e.g., “You are a senior DevOps engineer…”)
    • Parameters: Temperature, top-p, max tokens

These presets appear as selectable “models” in the chat dropdown, making it easy to switch personas without retyping system prompts.

Admin Features

FeatureLocationDescription
User ManagementAdmin → UsersCreate/delete users, assign roles
Model AccessAdmin → Users → PermissionsRestrict which models each user can see
Default ModelAdmin → Settings → InterfaceSet the default model for new conversations
System PromptAdmin → Settings → InterfaceSet a global system prompt for all users
Rate LimitingAdmin → Settings → InterfaceLimit messages per user per minute
Chat HistoryAdmin → Settings → GeneralEnable/disable chat saving
TelemetryAdmin → Settings → GeneralDisable usage analytics

Updating Open WebUI

# Pull the latest image
docker pull ghcr.io/open-webui/open-webui:main

# Stop and remove the old container (data is preserved in the volume)
docker stop open-webui && docker rm open-webui

# Re-run with the same command
docker run -d -p 3000:8080 \
  --add-host=host.docker.internal:host-gateway \
  -v open-webui:/app/backend/data \
  --name open-webui \
  --restart always \
  ghcr.io/open-webui/open-webui:main

Your conversations, users, and settings are preserved in the open-webui Docker volume.

Troubleshooting

”Could not connect to Ollama”

Cause: Open WebUI cannot reach the Ollama API.

Fix:

  1. Verify Ollama is running: curl http://localhost:11434/api/tags.
  2. If using Docker, ensure the --add-host=host.docker.internal:host-gateway flag is set.
  3. If Ollama is on a remote server, set OLLAMA_HOST=0.0.0.0 on the Ollama server and restart it.

Models Not Appearing in Dropdown

Cause: Ollama has no models downloaded, or the connection URL is wrong.

Fix:

  1. Download at least one model: ollama pull llama3.2.
  2. Check the Ollama URL in Admin → Settings → Connections.
  3. Click the refresh icon next to the Ollama connection.

RAG Returns Irrelevant Results

Cause: The embedding model is not configured or is too small.

Fix:

  1. Pull a good embedding model: ollama pull nomic-embed-text.
  2. In Admin → Settings → Documents, set Embedding Model to nomic-embed-text.
  3. Re-upload your documents to re-index them with the new model.

Open WebUI vs. Other LLM Interfaces

FeatureOpen WebUIChatGPTLM StudioText Gen WebUI
Self-Hosted✅❌❌ (desktop)✅
Multi-User✅ RBAC❌ (per-account)❌❌
RAG✅ Built-in✅ (paid)❌✅ (plugin)
Web Search✅✅ (paid)❌❌
Ollama Support✅❌❌❌
OpenAI Support✅✅ (native)✅✅
Model Presets✅✅ (GPTs)❌✅
Image Gen✅✅ (DALL-E)❌✅
Voice I/O✅✅❌✅
CostFree$20/mo+FreeFree

Summary

Open WebUI transforms your local Ollama setup (or any OpenAI-compatible API) into a full-featured, multi-user AI platform with RAG, web search, model presets, and admin controls. Deploy it in 30 seconds with Docker and give your team a private, cost-free ChatGPT alternative.