MyWebUI

My personal Open WebUI setup with custom actions and tools for news aggregation, web scraping, and AI workflow automation.

Overview

This repository contains my custom extensions for Open WebUI, a self-hosted web interface for running Large Language Models (LLMs) with Ollama. The extensions include deterministic action buttons, callable tools, and system prompts that enhance the base Open WebUI functionality.

Features

📰 News Aggregation

• RSS News Fetcher: Automatically pulls tech news from Kubernetes, ArgoCD, AWS, CNCF, The New Stack, and Hacker News
• Brutalist Report Scraper: Extracts curated tech headlines with real external links
• Smart Filtering: Filters news based on keywords (Kubernetes, DevOps, SRE, GitOps, Platform Engineering)
• Customizable Timeframes: Fetches articles from the last 48 hours

🧠 Memory Management

• Add to Memory Button: One-click action to save assistant responses to Open WebUI's memory system
• Status Indicators: Real-time feedback during memory operations
• Error Handling: Graceful error reporting with citations

🤖 Workflow Automation

• Deterministic Actions: Buttons that execute Python code without LLM involvement
• Custom Prompts: Pre-configured system instructions for specific workflows
• Event Emitters: Real-time status updates and message injection

What's Inside

mywebui/
├── actions/                    # Action buttons (deterministic Python code)
│   ├── add_to_memories_action_button.py
│   └── news.py
├── tools/                      # Callable tools for LLM
│   └── tool-brutalist_scraper-export-*.json
├── prompts/                    # System prompts for workflows
│   └── news.md
├── docker-compose.yaml         # Container configuration
├── .env                        # Environment variables (not committed)
└── CLAUDE.md                   # Developer documentation

Prerequisites

• Docker and Docker Compose
• Ollama running locally or on network
• Open WebUI compatible version: >= 0.5.0

Quick Start

1. Clone the Repository

git clone https://github.com/yourusername/mywebui.git
cd mywebui

2. Configure Environment

Create a .env file in the root directory:

# Tavily API Key for web search (optional)
TAVILY_API_KEY=your_tavily_api_key_here

3. Update Docker Compose

Edit docker-compose.yaml to point to your Ollama instance:

environment:
  - OLLAMA_BASE_URL=http://your-ollama-host:11434

4. Start Open WebUI

docker compose up -d

Access the interface at http://localhost:3000

5. Import Extensions

Option A: Manual Import (Recommended)

1. Go to Admin Panel → Actions
2. Upload action files from actions/ directory
3. Go to Admin Panel → Tools
4. Import tool files from tools/ directory

Option B: Volume Mount Mount the directories in your docker-compose.yaml:

volumes:
  - ./actions:/app/backend/data/actions
  - ./tools:/app/backend/data/tools
  - ./prompts:/app/backend/data/prompts

Usage

Fetching News

1. RSS News Action: Click the "Fetch Tech News" button in any chat to get the latest tech headlines
2. Brutalist Scraper Tool: Use the news prompt or ask the assistant to fetch news using the Brutalist tool
3. Custom Filtering: The LLM will automatically filter results based on your tech stack preferences

Saving to Memory

1. After receiving a useful assistant response, click the "Add to Memory" action button
2. The response will be saved to your personal memory store
3. Future conversations can reference this information

Configuration

Customizing News Sources

Edit actions/news.py to modify RSS feeds:

self.feeds = {
    "Your Source": "https://example.com/feed.xml",
    # Add more feeds here
}

Adjusting Keywords

Modify the keyword filter in actions/news.py:

self.keywords = [
    "kubernetes",
    "your-keyword",
    # Add more keywords
]

RAG Settings

Open WebUI is configured with:

• Embedding Model: mxbai-embed-large:latest
• Chunk Size: 1000 tokens
• Chunk Overlap: 200 tokens
• Top K Results: 5

Adjust these in docker-compose.yaml under the environment section.

Development

Adding Custom Actions

1. Create a new Python file in actions/ directory
2. Define an Action class with an async def action() method
3. Use __event_emitter__ to send status updates and messages
4. Import into Open WebUI admin panel

Example structure:

class Action:
    async def action(self, body: dict, __event_emitter__=None):
        await __event_emitter__({
            "type": "status",
            "data": {"description": "Working...", "done": False}
        })
        # Your code here
        return None

Creating Tools

Tools provide callable functions for the LLM. See tools/ directory for JSON export format.

Dependencies

Actions can auto-install Python packages:

try:
    import feedparser
except ImportError:
    subprocess.check_call([sys.executable, "-m", "pip", "install", "feedparser"])
    import feedparser

Tech Stack

• Backend: Open WebUI (Python/FastAPI)
• LLM Runtime: Ollama
• Models: Qwen 2.5 7B (task model), mxbai-embed-large (embeddings)
• Search: Tavily API
• Container: Docker

Environment Variables

Variable	Description	Default
OLLAMA_BASE_URL	Ollama API endpoint	http://192.168.15.13:11434
TAVILY_API_KEY	Tavily search API key	Required for web search
WEBUI_AUTH	Enable authentication	False
RAG_EMBEDDING_MODEL	Embedding model name	mxbai-embed-large:latest
TASK_MODEL	Task execution model	qwen2.5:7b-instruct-q4_K_M

Security Notes

• API Keys: Never commit .env files or hardcode API keys
• Authentication: Disabled by default (WEBUI_AUTH=False) - enable for production
• Network: Configured for local network access only
• Data: Persistent data stored in ./data directory (gitignored)

Contributing

This is a personal configuration repository, but feel free to:

• Fork and adapt for your own use
• Submit issues for bugs
• Share your own extensions

License

MIT License - Feel free to use and modify for your own projects.

Acknowledgments

• Open WebUI - The amazing self-hosted LLM interface
• Ollama - Local LLM runtime
• Brutalist Report - Curated tech news aggregator

Resources

• Open WebUI Documentation
• Open WebUI GitHub
• Ollama Models
• CLAUDE.md - Developer documentation for working with this repository

---

Note: This is a personal configuration. Customize settings, models, and endpoints to match your infrastructure.