Reddy2Read

• 

  Home page

• 

  Analyzing a post given the url

• 

  Post Summary

• 

  Sentiment Analysis & Main Arguments

• 

  Notable & Insightful Comments

Inspiration

While browsing Reddit, we often found valuable discussions hidden beneath long threads. We wanted a tool that could automatically surface the key ideas and sentiments of any Reddit post, especially for debates. This inspired us to build Reddy, an app that transforms Reddit threads into concise, structured summaries using AI.

What it does

Reddy2Read is a full-stack web application that takes a Reddit thread URL and returns:

• A short topic summary
• Major arguments for and against
• Sentiment analysis of the comments
• A list of notable, high-upvoted comments

How we built it

The frontend is built using React and styled with Tailwind CSS. The backend is powered by FastAPI, using PRAW to scrape Reddit data and Perplexity Sonar API to perform natural language analysis.

Challenges we ran into

• Designing an effective and flexible prompt that works for various Reddit post types (debates, questions, advice)
• Handling non-debate posts where “arguments for/against” don’t apply
• Managing CORS between the frontend (Vercel) and backend (Render)
• Ensuring the frontend handled loading, errors, and empty responses gracefully

Accomplishments that we're proud of

What we’re most proud of isn’t just the app itself, but the journey that brought it to life. We started late because we went through multiple project ideas that ultimately weren’t feasible—Sonar couldn’t handle some of the tasks we initially envisioned. After testing and reworking several times, we had to make the tough decision to completely change our app idea just a few days before the deadline. It was stressful, but it also gave us clarity. Because we had already spent so much time on prompt engineering during those failed attempts, we had a deep understanding of Sonar’s strengths and limitations. That hard-earned insight allowed us to move fast when we pivoted. We knew exactly how to craft prompts, how to structure the responses, and how to design around the model’s boundaries. Despite the rush, we managed to build something complete, useful, and even polished, and that’s what makes us proud.

What we learned

The most important thing we learned was prompt engineering. We had to carefully guide the language model to return structured and meaningful results, which took several iterations. We also realized that the Sonar API couldn’t scrape content from URLs — so we had to integrate Reddit’s API myself using PRAW to collect the thread data before sending it to Sonar.

What's next for Reddy2Read

• Add discussion clustering to group similar viewpoints in large threads
• Integrate comment threading visualizations for deeper insight into how conversations evolve
• Allow users to submit any subreddit or keyword and explore trending debates
• Support other platforms like Twitter or Hacker News for broader analysis
• Add user authentication so people can bookmark or save insightful debates

Built With

• fastapi
• perplexity-sonar-api
• python
• react.js
• reddit
• render
• tailwind-css
• vercel
• vite