Open WebUI - AI Chat Interface

Overview

The openwebui command manages the Open WebUI service using Podman Quadlet containers. It provides a web-based chat interface for interacting with Ollama LLM models.

Key Concept: Open WebUI connects to Ollama via the bazzite-ai network using DNS (http://ollama:11434). Ensure Ollama is running before using Open WebUI.

Quick Reference

Named Parameters

All parameters use named syntax (e.g., PORT=3001):

Configuration

# Default configuration (port 3000, localhost only)
ujust openwebui config

# Custom port
ujust openwebui config PORT=3001

# Network-wide access (0.0.0.0)
ujust openwebui config BIND=0.0.0.0

# Combine parameters
ujust openwebui config PORT=3001 BIND=0.0.0.0

# GPU-optimized image
ujust openwebui config IMAGE=ghcr.io/open-webui/open-webui:cuda

Update Existing Configuration

Running config when already configured updates the existing settings:

# Change only the bind address
ujust openwebui config BIND=0.0.0.0

# Update port without affecting other settings
ujust openwebui config PORT=3002

Container Images

Note: GPU is auto-detected and attached regardless of image choice.

Lifecycle Management

# Start Open WebUI
ujust openwebui start

# Stop service
ujust openwebui stop

# Restart (apply config changes)
ujust openwebui restart

# View logs (default 50 lines)
ujust openwebui logs

# View more logs
ujust openwebui logs LINES=200

# Check status
ujust openwebui status

# Show access URL
ujust openwebui url

Multi-Instance Support

# Start all instances
ujust openwebui start INSTANCE=all

# Stop specific instance
ujust openwebui stop INSTANCE=2

# Delete all instances
ujust openwebui delete INSTANCE=all

Shell Access

# Interactive shell
ujust openwebui shell

# Run specific command
ujust openwebui shell CMD="ls -la /app/backend/data"
ujust openwebui shell CMD="cat /app/backend/data/config.json"

Network Architecture

Open WebUI uses the bazzite-ai bridge network for cross-container DNS:

+-------------------+     DNS      +-------------------+
|   Open WebUI      | -----------> |      Ollama       |
|   (openwebui)     |              |    (ollama)       |
|   Port 3000       |              |   Port 11434      |
+-------------------+              +-------------------+
         |                                  |
         +------ bazzite-ai network --------+

Environment Variables (injected automatically):

OLLAMA_BASE_URL=http://ollama:11434
OLLAMA_HOST=http://ollama:11434
JUPYTER_HOST=http://jupyter:8888
COMFYUI_HOST=http://comfyui:8188

Network Binding

Security Note: Using BIND=0.0.0.0 exposes the service to your network. Consider using Tailscale for secure remote access:

# Expose via Tailscale (secure)
ujust tailscale serve openwebui

Data Persistence

Data persists across container restarts. Each instance has isolated data.

Common Workflows

Initial Setup

# 1. Ensure Ollama is running
ujust ollama start

# 2. Configure Open WebUI
ujust openwebui config

# 3. Start the service
ujust openwebui start

# 4. Access the web UI
ujust openwebui url
# Output: http://127.0.0.1:3000

Remote Access Setup

# Configure for network access
ujust openwebui config BIND=0.0.0.0

# Start the service
ujust openwebui start

# Or use Tailscale for secure access
ujust tailscale serve openwebui

Upgrade Container Image

# Stop service
ujust openwebui stop

# Update to new image
ujust openwebui config IMAGE=ghcr.io/open-webui/open-webui:main

# Restart
ujust openwebui start

GPU Support

GPU is automatically detected and attached:

Check GPU status:

ujust openwebui shell CMD="nvidia-smi"

Troubleshooting

Service Won't Start

# Check status
ujust openwebui status

# View logs
ujust openwebui logs LINES=100

# Check if Ollama is running
ujust ollama status

Common causes:

• Port 3000 already in use
• Ollama not running
• Container image not pulled

Can't Connect to Ollama

Symptom: "No models available" in web UI

Check:

# Verify Ollama is running
ujust ollama status

# Test Ollama connection from Open WebUI container
ujust openwebui shell CMD="curl http://ollama:11434/api/tags"

Fix:

# Start Ollama first
ujust ollama start

# Restart Open WebUI
ujust openwebui restart

Web UI Not Accessible

Symptom: Browser can't connect to http://localhost:3000

Check:

ujust openwebui status
ujust openwebui url

Fix:

# If using wrong bind address
ujust openwebui config BIND=127.0.0.1
ujust openwebui restart

Clear Data and Start Fresh

# Delete everything
ujust openwebui delete INSTANCE=all

# Reconfigure
ujust openwebui config
ujust openwebui start

Cross-References

• Required: ollama (Ollama must be running for models)
• Related: jupyter (ML development), comfyui (image generation)
• Network: Uses bazzite-ai network (shared with ollama, jupyter, comfyui)
• Docs: Open WebUI GitHub

When to Use This Skill

Use when the user asks about:

• "install open webui", "setup chat interface", "web ui for ollama"
• "configure openwebui", "change port", "network access"
• "open webui not working", "can't see models", "connection error"
• "open webui logs", "debug open webui"
• "delete open webui", "uninstall"