This has been my first time developing a browser extension.

In a perfect world with semantic clean markup everywhere the process of reformatting a website into a reader interface would be quite easy, but real world HTML can be really messy. Simply passing the DOM to @mozilla/readability (the engine Firefox uses for its reader view) was not sufficient. While it does a good job at identifying the main article, normalizing HTML and removing junk like widgets within it needs additional heuristics. So I built a custom pipeline of pre- and postprocessing steps around it and tested on as many websites as possible.

An interesting challenge has been how to get around CORS and CSP limitations by the host page. Some of the solutions I came up with are outlined on my blog post linked below.

Features I'm considering to add in a future release:

- one-click send to device for Kindle/Kobo etc.

- High quality Text-to-Speech

- Firefox/Edge/Safari release

github: https://github.com/fachkamera/booklike

blog post: https://mayermarkus.net/building-booklike