Hacker News Reader

Kudos! Loads superfast and well designed. I am curious about how to build a site like this? Could you shed some light on the stack used (frontend/backend)? I couldn't find anything immediately in the HTML like bundle.js or something similar.

View on HN

Sure - Heroku and Cloudflare for hosts. I have a lightweight JS server that handles ingestion of HN data and parses book references into a Postgres db. Site is built with Python using Starlette. There is quite a bit of caching with Redis and Cloudflare. There are a couple scheduled processes for ranking and digests. Spacy and vaderSentiment Python packages are used for text processing. There is no/minimal JS on the site except for hCaptcha. It's gone through a few iterations - the first version was just a one-off static site generation.

> I have a lightweight JS server that handles ingestion of HN data and parses book references into a Postgres db.

Do you parse every comment, and then look for book names?

Yes

First - thanks for building this!

I'm curious - how do you detect a book mention if, the first time it's mentioned, there's no link to Amazon? Do you use spaCy for NER and then use an API to check if the entity is a book name?

In case you were not aware, you can figure out the stack of a site pretty well using something like: https://builtwith.com/

Reply on news.ycombinator.com