Best Design at PennApps XXVI🏆

A project built to collect, duplicate, cluster, and summarize breaking news stories from across the web to mitigate biases that different news sources give.

DEMO - https://www.youtube.com/watch?v=rZpY3STlQxs

noogie automatically fetches articles from major outlets, groups similar stories, and generates concise AI-powered headlines. It helps users see the big picture of what’s happening, filtering the opinions of different news sources.

• Multi-Source Collection – Pulls stories via RSS and scraping from a variety of news sources
• Clean Parsing – Uses newspaper4k to extract titles and content, filtering out empty or paywalled pages.
• Smart Deduplication – Computes sentence embeddings (sentence-transformers) and stores them in a FAISS vector index
• Clustering & Summarization – Groups similar articles and calls OpenAI/Gemini to generate short, human-readable cluster headlines
• Scalable Design – Supports multithreading (planned) so multiple feeds can be fetched concurrently.

• TypeScript + React with d3.js for node visualization
• Python newspaper4k to extract news sources & flask for the servers
• OpenAI's GPT 4o-mini for summaries
• Supabase for storage

cd front-end
npm install
npm run dev

cd server
python -m venv
pip install -r requirements.txt
python server.py
python main.py

• GET api/clusters
• GET api/articles
• GET api/clusters/:clusters_id/articles
• GET api/clusters/:cluster_id
• GET api/articles/:article_id
• POST api/clusters/batch
• POST api/data/bulk
• POST api/clusters/:cluster_id/articles/batch

A special thanks to PennApps 2025.

Copyright (c) 2025 Jatin Punjabi, Andres Lopez, Ruslan Akmyradov