First I used manual approach: - add .json to URL to get metadata (fetches ALL comments) - pass json to AI to get the summary
It works, you can use it, but metadata are "dirty", containing lots of unnecessary fields that bloats up the context for AI. And it's not that fast, I wanted single click solution.
It wasn't hard to automate it in chrome extension. Supported pages are Threads, Subreddits, Search results.
It’s built with Plasmo and React. The backend is a serverless proxy (Vercel) to handle rate limiting, key security and calls to OpenRouter API.
Demo: https://youtu.be/9MGwQElbBXE Chrome extension: https://chromewebstore.google.com/detail/jccgjnbkibbigimcfoh...
Feedback is welcome!
SilverElfin•1mo ago
MartinMachava•1mo ago
Right now when user analyzes: - Thread, it fetches all metadata, 1 req - Subreddit, fetches 100 posts, 1 req - Search result, 100 posts, 1 req
Subreddit posts and Search result posts are just posts. Extension doesn't open the posts to get all comments of each. But I thought of it. I don't remember exact numbers of reddit's rate limits but it should be possible to implement in the future. I'm thinking of adding "overheat bar" (like in games) to indicate to the user how much requests is remaining per minute and other safety checks.
Fetching metadata is done on user's behave, not reddit API. It's not scraping. Metadata are accessible to anyone by just adding .json to URL