CoRead - Supercharged Ctrl F

I mean, who doesn't Ctrl+F on a regular basis? Imagine reading full text articles, couldn't be me. This project aims to expand the power of Ctrl+F using vector queries and generative AI

Comment

Redefining Text Search

Inspiration

🌟 We found ourselves using Ctrl+F way too often while searching for information online.

What it does

🚀 CoRead is a game-changer for text searching. It analyzes webpages, making it a breeze to find specific information while also providing citations and confidence levels. Imagine effortlessly parsing through lengthy academic papers. For a simplified view of the process, check out this graphic: CoRead Pipeline

How we built it

🔧 Building CoRead was a journey of learning. We leveraged pre-made AI models due to our limited team size and inexperience with AI. It was also our first time coding a browser extension. Our toolbox included:

  • Cohere API
  • HuggingFace Hub
  • PineCone DB
  • Flask
  • NodeJS
  • ngrok

For detailed citations and attributions, visit our GitHub page.

Challenges we ran into

💡 To put it succinctly, we faced challenges as beginners in both AI and extension development. Additionally, we had half of our team depart within the first two hours.

Accomplishments that we're proud of

🏆 Despite the hurdles, our backend works, and we proudly present an AI-powered API. Performance-wise, it can analyze Wikipedia in just 20 seconds (on my modest CPU) and provides answers within 2 seconds.

What we learned

🧠 CoRead opened our eyes to the potential of AI in improving everyday tasks. People use Ctrl+F daily, and once we fine-tune it, we'll integrate it into our personal lives.

What's next for CoRead

🌐 We primarily tested this on Wikipedia, especially the Genghis Khan page, which we humorously call "Ctrl+Khan." To make it universally useful, we're exploring open-source alternatives.

Acknowledgments:

This README was co-written with assistance from ChatGPT.

Built With

Share this project:

Updates