That Obsidian sync is the real prize. I've got a similar pipeline dumping highlights into my daily notes, but I'm sending them through a small Lambda that strips out any auto-generated "key points" boilerplate before they land. Makes the review flow way cleaner.
Watch out for Scholarcy's DOM changes though, they've broken my scrapers three times this year. 😅 Good call putting it on GitHub, community patches are the only thing keeping my own hacky connectors alive.
The Obsidian sync is definitely the killer feature you unlocked. I've seen teams try to cobble this together with Zapier and it's always a mess of rate limits and formatting drift.
My main caveat would be about the direct DOM scraping. You're going to be in a constant maintenance race with Scholarcy's frontend team. I had a similar setup for exporting Anki cards, and it broke with every other minor UI refresh. I ended up forking the project to add a proper API endpoint because maintaining the scraper became a part-time job.
Consider adding a simple validation step that checks for the presence of expected DOM elements before attempting to send. At least then the extension fails visibly instead of silently pushing empty highlights.
Yep, the Zapier comparison hits home. It's amazing how quickly a "no-code" solution becomes a full-time job of troubleshooting.
Your point about visible failure is crucial. I've found that a simple validation flag - like checking if the 'key points' section contains actual text, not just placeholder divs - can save users from a whole batch of empty pushes. It turns a silent failure into a clear "can't find the content" message, which at least tells them to refresh the page or check for a UI update.
Stay curious, stay skeptical.