Skip to content

Introduction to UglyFeed

UglyFeed is an open-source intelligence and news aggregation engine that leverages Large Language Models (LLMs) to retrieve, filter, evaluate, rewrite, and serve clean, enriched RSS feeds.

GitHub starsLicense: MIT


Why UglyFeed?

Traditional RSS readers suffer from information overload:

  • Multiple feeds cover the same breaking stories repeatedly.
  • Articles are filled with clickbait headlines and low-signal summaries.
  • Manual curation is time-consuming and difficult to automate.

UglyFeed solves this by inserting an AI intelligence layer between the raw RSS sources and your feed consumer:

mermaid
flowchart LR
    A[Raw RSS Feeds] --> B[RSS Reader Engine]
    B --> C[Semantic Deduplication]
    C --> D[LLM Rewriting & Enrichment]
    D --> E[Quality & Safety Checks]
    E --> F[Published RSS 2.0 / CDN]

Key Highlights

  1. Multiple LLM Providers:

    • Google Gemini: Full support including Gemini 2.0 Flash and Gemma 3 on free tier.
    • OpenAI: GPT-4o, GPT-4, and GPT-3.5 Turbo.
    • Groq: Ultra-fast Llama 3 8B/70B and Mixtral inference.
    • Ollama: 100% private local model execution on your own hardware.
  2. Semantic Similarity & Vectorization:

    • Compares incoming articles against recently published items using vector embeddings.
    • Suppresses near-duplicate stories across completely different publishers.
  3. Autonomous Daily Delivery:

    • Run daily workflows in GitHub Actions without managing servers.
    • Publish output XML to any target Git repository acting as a zero-cost CDN.
  4. Web UI & CLI:

    • Full Streamlit graphical user interface for inspecting feeds, testing prompts, and configuring pipelines.
    • Script runner and CLI mode for headless server environments.

Next Steps

Released under the MIT License.