My pleasure. I'm also unaware of a software or service since it seems a bit of a too specific usecase.
I'm doing all this (with no cost) using an AWS Lambda function written in Java which is scheduled to be called daily at 7 am. It basically parses some RSS feeds[1] (or parses html[2] when RSS is non-existent), extracts each article content using readability and packages them all in a daily epub[3] (with a TOC) which is stored in a S3 folder. Then a separate endpoint (the one I use in my X4) is taking most recent N S3 objects from that S3 folder and exposes the OPDS. The only downside of this is that the response time takes about 3 extra seconds since it uses a cold start, but response speed is not important to me.
Using 'OPDS as a news library' works incredibly well with my X4, it takes me under 10 seconds to download and open the epub and should work with any e-ink device that can install KOReader (or any other software with OPDS support). And I love that all articles are inside a single epub and I can easily go to another one from the TOC.
It's probably my most useful personal project and I like that by now it follows the 'set it and forget it' principle but in need, since it's written by hand, it's very easy to customize it: filter out AI/war/finance/local-gibberish/whatever articles, add new 'bridges' to websites without RSS, some articles don't respect html guidelines and need some tinkering, etc.
