I use https://openpaper.work/en/ https://openpaper.work/en/ - the direct scanning feature doesn't support duplex (yet) so I additionally made a small wrapper sc

I use https://openpaper.work/en/ https://openpaper.work/en/ - the direct scanning feature doesn't support duplex (yet) so I additionally made a small wrapper script around SANE's 'scanimage' command to scan, which also directly pipes it into Tesseract 4 (OCR). The result is a pdf file with an invisible text layer. I then import those pdfs into paperwork in bulk. It has a label system and learns to apply labels correctly, which isn't 100% accurate but rather helpful. It also has rather quick fulltext search, which is nice. All in all, paperwork isn't perfect but it does the job well enough for me. As a scanner, I got an Epson ES-200 (DS-310 in Europe) from Amazon Warehouse Deals (the packaging was dinged, which apparently justified a 35% discount). I keep around the originals in a big chronological pile, rotated yearly. If I ever need an original document, I can search for it in my digital archive and find out its date. Then I can binary search for it in the pile. The annoying part is scanning all your old stuff. Even with a fast feed scanner (mine is about as fast as Tesseract can process it in simplex, and twice as fast as the OCR in duplex, making my CPU -- and not the scanner -- the limiting factor), it just takes a lot of time.