Stirling-PDF OCR clean-up yields no visible UI (with fix)

[stiggy87]

I was trying to clean-up various PDF files to be used with a local LLM. When the nordic encoder is used it reads a lot of data as garbled mess. To fix this, I went to my local Stirling-PDF install to try and clean up the OCR, but when I tried, it didn't show anything!

Digging some more, I found that the tesseract-data was not installed correctly. A couple things I found out:

1. The tesseract-data was installed in /usr/share/tesseract-ocr/5 when Stirling-PDF is looking for /usr/share/tessdata by default
2. The /opt/Stirling-PDF/.env does not contain the TESSDATA_PREFIX which is the path to the data location. (You could use the existing path, but the Stirling-PDF docs suggest the tessdata directory).

I fixed these items and restarted the service and everything looks to work!

I think the Stirling-PDF install/update script should be adding this info to make it work out of the box.

Anyone else have this issue? Or is it my install being bad? I did make this LXC before the refactor.

[tremor021]

Tesseract is installing with a simple apt-get install -y 'tesseract-ocr-*'. We don't have controll over how tesseract deb packages install

[stiggy87]

I understand there's no control over the installation of the tesseract-ocr packages.

What I'm bringing up is, the installing/update script might need to be updated following StirlingPDF documentation related to where the tesseract-ocr is placed. https://docs.stirlingpdf.com/Advanced%20Configuration/OCR/

What I mentioned is a good workaround if people want to use the OCR feature and run into the problem I did.

[MickLesk]

Isn't it much easier to just create a symlink?