Issue: Cannot run addon without Nix.

[tenpai-git]

Fri, January 24th, 2025, 10:15: Running 1 addon tasks for trigger Set(ExistingItem)
Fri, January 24th, 2025, 10:15: Preparing item data 'IMG_20250124_190042_612.jpg' for addon
Fri, January 24th, 2025, 10:15: Writing collected item data…
Fri, January 24th, 2025, 10:15: Storing all attachments…
Fri, January 24th, 2025, 10:15: Storing attachment /tmp/docspell-addons/addon-4862853053256892935/item/pdfs/9J7eAqcmcvb-apJPGeaTwh2-h4BqDwpK1wH-E8RxW3ZcTrd
Fri, January 24th, 2025, 10:15: Storing attachment /tmp/docspell-addons/addon-4862853053256892935/item/originals/9J7eAqcmcvb-apJPGeaTwh2-h4BqDwpK1wH-E8RxW3ZcTrd
Fri, January 24th, 2025, 10:15: Storing file metadata
Fri, January 24th, 2025, 10:15: About to run 1 addon(s) in /tmp/docspell-addons/addon-4862853053256892935
Fri, January 24th, 2025, 10:15: Extract 1 addons to /tmp/docspell-addons/addon-4862853053256892935/addons
Fri, January 24th, 2025, 10:15: Extracting AddonArchive(docspell-file:///1/addon/7omFKdA7cSanPHS3fJr17biQAHrc72T2xrxZwdzFXhDw,rotate-pdf-addon,0.2.0-pre)
Fri, January 24th, 2025, 10:15: Executing addon rotate-pdf-addon-0.2.0-pre
Fri, January 24th, 2025, 10:15: Storing user input into file
Fri, January 24th, 2025, 10:15: Attempt to run addon rotate-pdf-addon-0.2.0-pre with runner List(NixFlake)
Fri, January 24th, 2025, 10:15: Prepare addon rotate-pdf-addon-0.2.0-pre for executing via nix
Fri, January 24th, 2025, 10:15: Building with nix build
Fri, January 24th, 2025, 10:15: Running external command: nix build
Fri, January 24th, 2025, 10:15: Job 7XDy6toDK.../1/addon-existing-item/High execution failed. Retrying later.: Cannot run program "nix" (in directory "/tmp/docspell-addons/addon-4862853053256892935/addons/rotate-pdf-addon-0.2.0-pre"): error=2, No such file or directory

Hi @eikek

I'm having trouble getting this addon to run without installing nix. The issue I'm having is that images uploaded to Docspell become converted to PDFs for Tesseract, and those pdfs are always being rotated incorrectly automatically.

So I tried installing the rotation addon to fix the rotated pdf and reprocess the metadata. However, even though I manually installed qpdf and xpdf, I am unable running into the above output where I can't run the addon without nix.

I used the useage guide, but can you help me get the plugin running on a basic Debian install?

[eikek]

You need to remove the nix-runner from the executors in jour joex config. You can see https://docspell.org/docs/configure/defaults/#joex - search for executor-config. In there is a runner setting. you need to name the ones your joex machine supports

[tenpai-git]

Okay, thank you for that @eikek

So I wasn't able to get the trivial runner running (never actually heard of this project or runner and couldn't find information about it), so I installed docker, their repos, and set it up for the docspell user.

I still had another permissions error in the job queue but was able to resolve it with the answer in this thread for anyone looking for help on this in the future:
https://stackoverflow.com/questions/48957195/how-to-fix-docker-got-permission-denied-issue

So I did get it running correctly and I consider this issue "solved" - however it didn't help me because the root problem is really not being able to process the metadata on the file. Actually the converted PDF I think is more garbled because it appended the incorrect metadata after rotating it automatically to the file, too. It may not be saveable after rotation.

Consider this;

1. I upload a straight image, say from a camera a JPG or PNG
2. The auto convert automatically rotates it when converting to pdf by 90 degrees one way or another. so it can be tesseract'd.
3. The meta data (both labels and extracted metadata), gets completely gets messed up because the orientation is wrong and it can't read anything and everything gets messed up.
4. I rotate it with rotate-pdf-addon, but this does not reprocess any metadata.
5. Reprocessing the file once again uses the original image and the problem repeats.

So I guess I could upload, rotate the image, save the converted pdf, delete the old entry, upload the new pdf,.. but at scale this is quite an issue.

So I guess my question is... Anyway to not make the converted pdf rotate in the first place ^^; - is there any setting in joex config where I can tell the proper component to not rotate the converted PDF or to try and "find" The correct orientation?

Or could we go about an addon to reprocess the metadata of the rotated file...?

What do you think?

I guess this brings me back to Issue #1 - maybe merge issues?

[eikek]

hi, in theory, the rotate addon would reprocess the rotated file. not sure why this doesn't work.

[tenpai-git]

Hi Eikek, Okay I checked it out - it turns out it was just a cache issue. The rest server rotates the image and pdf preview after the addon job, but doesn't update the metadata until you refresh the page. No big deal. So that solves a big part of it for the addon! Also ocrmypdf has a `--rotate-pages` option, I will see if I can get it to work. That leaves me with one question. I am wondering if I wanted to make an add-on that gets the text output from the yellow "View Extracted Data" area and reformat it through a bash script and put it back in that tab, how would I go about that? I can do this directly in ocrmypdf with `ocrmypdf --sidecar output.txt input.pdf` and piping the output.txt to a bash script that takes output.txt and formats it. Maybe the add-on could just add the output to the notes section? So something like; # Addon takes the current pdf input.pdf # Addon runs ocrmypdf on it with --sidecar output.txt input.pdf # That output is stored somewhere temporairly # Whatever script I want to cleanup that output.txt runs on the script (this would be subject to some parameters) # The final output.txt is added to notes. Is something like this feasible? I know how to do it in theory entirely on the system, just not how I would hook or find the "current item" for the addon, the input pdf, or add it to the notes.

…

On 1/26/25 00:02, eikek wrote: hi, in theory, the rotate addon would reprocess the rotated file. not sure why this doesn't work. — Reply to this email directly, view it on GitHub <#3 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BAIX63B5TC64N7O4V2EF7BD2MORP7AVCNFSM6AAAAABVZLMRFCVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMJTHE4TINBUG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>