Post #2683893
2026-04-26 21:42 UTC
Replies (3)
-
@neon_parity@makervine.net 2026-04-26 21:46
I installed Ubuntu Server on the Pi4, and wrote a shell script that leverages SANE (Linux scanner drivers/utilities), scanbd (to detect when the scan button is pressed on the scanner), Tesseract (local OCR to add a text layer to the scanned PDF), and a small local Ollama instance running elsewhere in the house.
-
@neon_parity@makervine.net 2026-04-26 21:51
Now when you press “Scan", the Pi4 wakes up, scans the document pages to TIFFs, combines them, runs them thru OCR, passes the first 2 pages of text to the local Ollama server which uses an LLM to try and generate a meaningful filename, passes *that* back to the script, which then renames the resultant PDF file and drops it in a “Scans" folder on my local NAS. Nothing leaves my network, and there’s no need to use Fujitsu/Ricoh's crappy software.
-
@neon_parity@makervine.net 2026-04-26 21:53
About the only part of this that I'm not fully happy with is the LLM-use. I'm not a huge fan of most LLMs for many practical and ethical reasons. This particular application is a good fit for them though, doesn't require a big model, and is a good opportunity to experiment with a few of the ethically-trained models out there.