Open WebUI role

Deploy Open WebUI, a web interface for interacting with AI models via Ollama, with Traefik reverse proxy integration.

Usage

Include the role in your playbook:

# playbooks/proxmox/deploy_docker_services.yml
- hosts: docker_services
  roles:
    - role: openwebui

The role provides these tags: - openwebui-install - Deploy Open WebUI container - openwebui-uninstall - Remove Open WebUI container

Docs

Deploy Open WebUI to your Docker host:

ansible-playbook -i inventory/hosts.yml playbooks/proxmox/deploy_docker_services.yml --tags openwebui-install

Remove Open WebUI installation:

ansible-playbook -i inventory/hosts.yml playbooks/proxmox/deploy_docker_services.yml --tags openwebui-uninstall

Configuration Details

The role deploys Open WebUI with: - Official Image: ghcr.io/open-webui/open-webui:main (latest) - Traefik Integration: Automatic HTTPS via Traefik reverse proxy - Persistent Storage: Application data and settings - Ollama Backend: Connects to remote Ollama instance via Tailscale - Network: Connects to external proxy network

Accessing Open WebUI

After deployment, Open WebUI is available at: - URL: https://chatgpt.rwcloud.ch - Port: 8080 (internal, accessed via Traefik)

Ollama Integration

The deployment is configured to connect to a remote Ollama instance: - Ollama URL: http://macbookpro01.tail8c2418.ts.net (via Tailscale)

This allows Open WebUI to access AI models hosted on your Ollama server without running models locally in the container.

Storage

One volume is created: - data - Contains Open WebUI configuration, chat history, and settings

Features

Open WebUI provides: - ChatGPT-like interface for local AI models - Support for multiple Ollama models - Conversation history and management - Model switching - Markdown rendering - Code syntax highlighting - Document upload and processing - User authentication and multi-user support

First-Time Setup

On first access: 1. Create an admin account 2. Configure Ollama connection (pre-configured to Tailscale endpoint) 3. Select available models from your Ollama instance 4. Start chatting!

Available Models

The models available depend on what's installed on your Ollama server. Common models include: - llama2 - codellama - mistral - mixtral - phi

Pull models on your Ollama server:

ollama pull llama2
ollama pull codellama

Customization

To change the Ollama backend URL, update the OLLAMA_BASE_URL environment variable in the role's install task.