uglypy 0.0.14

A Python package for aggregating and processing RSS feeds with LLM-enhanced content rewriting.

UglyFeed

UglyFeed is a simple application designed to retrieve, aggregate, filter, rewrite, evaluate and serve content (RSS feeds) written by a large language model. This repository provides the code, the documentation, a FAQ page and some optional scripts to evaluate the generated content.

Features

• 📡 Retrieve RSS feeds
• 🧮 Aggregate feeds items by similarity
• 🤖 Rewrite content using LLM API
• 💾 Save rewritten feeds to JSON files
• 🔁 Convert JSON to valid RSS feed
• 🌐 Serve XML feed via HTTP server
• 🌎 Deploy XML feed to GitHub or GitLab
• 📈 Evaluate generated content
• 🖥️ Web UI based on Streamlit
• 🐳 Docker installable package
• 🐍 Pip installable package
• ⚙️ Github action workflow available
• 📰 RSS test feeds available

Requirements

• 🌎 Internet connection
• 📲 an RSS reader (to retrieve the final feed I use FluentReader on laptop and NetNewsWire on mobile)
• 🤖 a large language model via API

Supported API and models

• OpenAI API (gpt-3.5-turbo, gpt4, gpt4o)
• Ollama API (all models like llama3, phi3, qwen2)
• Groq API (llama3-8b-8192, llama3-70b-8192, gemma-7b-it, mixtral-8x7b-32768)
• Anthropic API (claude-3-haiku-20240307, claude-3-sonnet-20240229, claude-3-opus-20240229)

Quick start

Prerequisites

• Docker: Ensure you have Docker installed on your system. You can download and install it from Docker's official site.
• Ollama to run local models or an OpenAI or Groq API key.

Running the Container

To start the UglyFeed app, use the following docker run command:

docker run -p 8001:8001 -p 8501:8501 -v /path/to/local/feeds.txt:/app/input/feeds.txt -v /path/to/local/config.yaml:/app/config.yaml fabriziosalmi/uglyfeed:latest

Configure the application

In the Configuration page (or by manually editing the config.yaml file) you will find aggregation similarity, LLM API, LLM model, retention, scheduler and deploy options.

Execute the application scripts

Execute all scripts in the Run scripts page easily by clicking on the button Run main.py, llm_processor.py, json2rss.py sequentially. You can check for logs, errors and informational messages.

Serve the final rewritten XML feed via HTTP

Once all scripts completed go to the View and Serve XML page where you can view and download the generated XML feed. If you start the HTTP server you can access to the XML url at http://container_ip:8001/uglyfeed.xml

Deploy the final rewritten XML feed to GitHub/GitLab

Once all scripts completed go to the Deploy page where you can push the final rewritten XML file to the configured GitHub/GitLab repository, the public XML URL to use by RSS readers is returned for each enabled platform.

Documentation

Please refer to the extended documentation to better understand how to get the best from this application.

Use cases

The project can be easily customized to fit several use cases:

• Smart Content Curation: Create bespoke newsfeeds tailored to niche interests, blending articles from diverse sources into a captivating, engaging narrative.
• Dynamic Blog Generation: Automate blog post creation by rewriting and enhancing existing articles, optimizing them for readability and SEO.
• Interactive Educational Tools: Develop AI-driven study aids that summarize and rephrase academic papers or textbooks, making complex topics more accessible and fun.
• Personalized Reading Experiences: Craft custom reading lists that adapt to user preferences, offering fresh perspectives on favorite topics.
• Brand Monitoring: Aggregate and summarize brand mentions across the web, providing concise, actionable insights for marketing teams.
• Multilingual Content Delivery: Automatically translate and rewrite content from international sources, broadening the scope of accessible information.
• Enhanced RSS Feeds: Offer enriched RSS feeds that summarize, evaluate, and filter content, providing users with high-quality, relevant updates.
• Creative Writing Assistance: Assist writers by generating rewritten drafts of their work, helping overcome writer's block and sparking new ideas.
• Content Repurposing: Transform long-form content into shorter, more digestible formats like infographics, slideshows, and social media snippets.
• Fake News Detection Datasets: Generate datasets by rewriting news articles for use in training models to recognize and combat fake news.

Contribution

Feel free to open issues or submit pull requests. Any contributions are welcome!

Roadmap

I started this project a month ago to experiment, get fun, learn and contribute to the open source community on my free time. I am so grateful to those who already made me empowering this pathway in a so short timeframe 🙏

Here some improvements I am still working on:

• overall code improvements and tests
• extend to generate HTML/media from rewritten JSON with themes/styles (tentatives with PiperTTS and others)
• here something i forgot 😅

Disclaimer

It is crucial to acknowledge the potential misuse of AI language models by this tool. The use of adversarial prompts and models can easily lead to the creation of misleading content. This application should not be used with the intent to deceive or mislead others. Be a responsible user and prioritize ethical practices when utilizing language models and AI technologies.

License

This project is licensed under the AGPL3 License.