[Paperless-NGX] OCR not working with additional languages

[Stooovie]

I need to add Czech (CES) to OCR because I'm getting MissingDependencyError: OCR engine does not have language data for the following requested languages: CES

PNGX documentation says:
If you run paperless on docker, paperless.conf is not used. Rather, configure paperless by copying necessary options to docker-compose.env

LXC is not Docker, so I'm editing /opt/paperless/paperless.conf

Adding PAPERLESS_OCR_LANGUAGE=eng ces makes zero difference.

anything else? thanks!

[Toxo666]

in lxc
apt install tesseract-ocr-ces

restart paperless service

[Stooovie]

Sorry but didn't help (package did install correctly). Even after reboot, same error message. ces is the correct code for Czech.

[Toxo666]

You can look at
tail -n50 /opt/paperless/data/log/paperless.log

or on the paperless website under protocol point.

[Toxo666]

look at the paperless.conf under PAPERLESS_OCR_LANGUAGE.
Look for : or =
In Paperless LXC you find config under /opt/paperless/paperless.conf
It must be an =
Test it with a single language first:
PAPERLESS_OCR_LANGUAGE=ces
After it works, test with two languages.
PAPERLESS_OCR_LANGUAGE=eng+ces

Link to Paperless documentation:
https://docs.paperless-ngx.com/configuration/#PAPERLESS_OCR_LANGUAGE

[Stooovie]

Sadly, not working. Still queued indefinitely (well, 1-page PDF for 20 minutes). I am not getting the error message anymore (MissingDependencyError: OCR engine does not have language data for the following requested languages: CES), but nothing's happening. I have restarted the LXC after each adjustment.

Thanks for your help but this is where I give up, I have already too many systems to babysit 🥲

[Toxo666]

its ok.
if you want to try it again, give the lxc the power to ocr.
ocr is cpu intense, if you have 2 languages, you need 2 times longer.
personally i have set 16 cores to paperless and 8gb ram ocr my documents.
the default setting from ttecks script is to low for my opinion. only for running the lxc without doing anything ;)

PAPERLESS_TASK_WORKERS=4
PAPERLESS_THREADS_PER_WORKER=4

https://docs.paperless-ngx.com/configuration/#PAPERLESS_TASK_WORKERS

greets toxo

[Stooovie]

I started from scratch, no dice.

With default settings, PNGX complains (even with notification in GUI) about MissingDependencyError: OCR engine does not have language data for the following requested languages: ces

Uncommenting and adding
PAPERLESS_OCR_LANGUAGE=ces

results in the file stuck in Queued (NO GUI warning)
with

Error occurred while consuming document mrdat.pdf: MissingDependencyError: OCR engine does not have language data for the following requested languages:
ces
Please install the appropriate language data for your OCR engine.

in log.

installing apt install tesseract-ocr-ces inside the LXC finishes fine, leads to the exact same behaviour.

I don't think this is me. CPU should not be an issue, the OCR process is not starting with Czech language at all.

[Stooovie]

It seems just installing the tesseract-ocr-ces package and not touching the conf file at all at the very least allows for the scanning to start and finish. The scanned document looks fine. Let's see what happens when I add english PDF.

[Snowii123]

hi there, how did this work out at the end please? thinkiong of installing paperless-nxg too but it working with Czech is essential for me... thank you!