I've been using elixir to build an app that lets an administrator add new rss feeds, render the article titles and summaries to a single page, scroll through them and push the ones they like to a landing page as a collection to read later. Many of the sites I like don't have the conventional RSS "structure" so I have to modify my main parser and adapt it to the outliers. I'm curious, how do you adapt to feeds that don't fit the conventional RSS structure? I was thinking of just using an LLM to automate it as I keep using Claude AI to expedite the process.
Do you mean they technically have RSS, but they’re using slightly different fields? E.g. summary instead of description?
I’m using FastRSS[^0] with some lightweight pattern matching to convert them to an internal model. I get error notifications for mismatches, and just push a new pattern match to handle the outlier.
Longer term it could be interesting to get an LLM to write some Lua to parse JIT.
Performance is rock solid, and it’s almost ready to release, I just need to tweak a few things (like free trial with no CC).
I have a very long to do list, and ultimately want to extend it with “change detection”, e.g. notify when an HTML element on a website changes.
All feedback is welcome
https://feedsync.net