Paperless-ngx is an extensive document management system. 3.0.0 has just been released.
It seems from a user/admin perspective that a lot of obsolete stuff under the hood got removed and they did some performance stuff, but it’s also full of AI features. There’s too many changes to read them all though, I wonder if there’s a blogpost or something somewhere summarizing the main changes.
Personally I’ll stay with 2.X for some days at least to see what people say about the new version. I’m not interested to have AI parse my documents, the alrogirthmic auto tagging etc works quite well.
You must log in or # to comment.
I accidentally upgraded by blindly using the
latestDocker tag, and the frontend of the new version is so much faster than before. Love to see it!At the risk of getting internet-assaulted I am one of the primary devs, feel free to ask me about it. To address some of the issues already raised:
- All the new “AI” (LLM) features are opt-in, meaning if you dont enable them you wont see them, literally. At all.
- All of them can also be run local-only, of course.
- There are a ton of other features as well.
- One of our devs has made use of Claude especially around database schema improvements, which is a particularly tricky thing to work on. I dont like AI, I wish I could go back to the world before it, but I cant. So at this point it is almost absurd to not use LLM tools to some degree in programming. AI slop or vibe-coding is to do so indiscriminately, and I hate that too. Believe me we get tons of PRs from people with basically no dev experience or abilities slinging up a Claude PR they dont understand… and we dont accept them. We expect the same from ourselves. We use it sparingly and very specifically, interrogate it aggressively and stand behind it after it’s merged. To somehow object to any / all use of it at this point either implies a) you dont do programming or b) you dont understand the nuances of how tools can be used in programming. Used properly, it’s no more offensive than using an IDE instead of a bare text editor, is that something you care about?
- Also, leaving Claude as a co-author I would argue seems more transparent, not doing so or not disclosing it in the PR seems like the shadier choice.
- The ethical concerns about LLMs, compute use etc are real, but I dont think one OSS project somehow needs to bear the entire weight of this societal problem.
Again, Im happy to discuss more, as long is its not the guy who opened the raving “issue” in our project.
Thank you for the amazing work. This is, along side immich, my most used self hosted project. I’ll definitely give this a look over this weekend!
Thanks for chiming in! I don’t fully agree with your reasoning re the usage of AI but I can accept your stance and am glad that you implemented it opt in with no intention of annoying me into using it. I also know many people have been asking for those features for a while though.
The person creating the issue has been downvoted here as well, even though this is a largely anti-ai place in my perception.
By now I also found the upgrade guide which lists the main features - I usually read the release notes, but this time it was a whole lot and hard to take it all in.
Respect for the nuance of the argument. Congrats on the major version.
Hey, dont believe me, Im just a random, but this guy seems like worth a listen: https://arstechnica.com/ai/2026/07/linus-torvalds-to-critics-of-ai-coding-in-linux-fork-it-or-just-walk-away/
No I’m saying I respect the fact you have more nuance than most towards the argument and didn’t leave it in such a reductive form that doesn’t help anyone.
Last time I looked at Paperless-NGX it insisted on taking the source files and moving/renaming them so it could manage them itself. In fact IIRC the only way to add files was to give them to the app (or upload via web UI) so it could take them over.
Does it still do that?
I don’t like tools that demand to take over the original files. I’d prefer it if it could watch a read-only folder for new files. Immich for example or Jellyfin can do that.
You can define the folder structure used by the app, so if needed, you can backup your documents as raw files (of course, you’ll lose metadata from Paperless).
Yeah it maintains it’s own folder structure and I’m not aware of a way past that (not that I looked).
There’s several more options to get the documents into it though. Ingest folder, via email, certainly more.
yes i have waited to start with my paperless journy because of the extra features und breaking changes.
FYI: they use Claude Code.
Just in case anyone has an issue with that.
Pro-tip: Block the user “claude” on GitHub and it’ll show you a warning for every repository where they forgot to remove the Co-Author tags:
Where
they forgot to removethere are co-author tags.There are plenty of people who are interested in using tools that are available and also recognize that it is underhanded to hide that fact for people who feel strongly about it.
[!NOTE] Upgrading to Paperless-ngx v3 can only be performed from version 2.20.15. If you are running an older version, please upgrade to v2.20.15 before proceeding with the v3 upgrade.
Acronyms, initialisms, abbreviations, contractions, and other phrases which expand to something larger, that I’ve seen in this thread:
Fewer Letters More Letters Git Popular version control system, primarily for code NAS Network-Attached Storage Plex Brand of media server package RAID Redundant Array of Independent Disks for mass storage RPi Raspberry Pi brand of SBC SBC Single-Board Computer SSD Solid State Drive mass storage VPN Virtual Private Network
7 acronyms in this thread; the most compressed thread commented on today has 5 acronyms.
[Thread #60 for this comm, first seen 23rd Jul 2026, 10:40] [FAQ] [Full list] [Contact] [Source code]
@two9a@lemmy.world your bot is breaking again pulling in words not used in the discussions.
I may be mistaken, but I had looked into Paperless before and saw they had AI features before this. Is it just more, or more in your face? I was under the impression they were optional, at least when I looked into it.
Honestly, this is one of those things where you need to define what AI means. Paperless 2.x used a machine learning-backed OCR library called Tesseract that has been in development since 1985.
The current version of Tesseract does use a LSTM neural network, since 2018. Is that an AI? Fuck if I know because AI is a basically meaningless term.
It’s not an LLM or a diffusion model, which is what backs a lot of the things people are calling AI nowadays.
As long as it’s not gambling or a scam. (Points are most “technology” companies and their products)
They added an AI chatbot that you can ask for information about the contents of the document you’re viewing. There’s more in the linked PR.
Paperless always had some sort of primitive AI for document classification and auto tagging. I haven’t yet read the changes on 3.x but I’ve been using it for years and it’s like a dumb AI that needs a lot of guidance and manual tagging at first, but as time goes on, it “figures out” every document you throw in the ingest. I’ve been sending it emails, or uploading straight from the mobile app and its amazing, can’t go back to a folder filled with “Download(739).pdf” or “IMG_4926.jpg” files of random recipts and invoices
I agree. The AI features were very lightweight before. Just for auto tagging etc.
Hope this isn’t a shitshow.
I had a quick look and the AI features that are added:
- LLM driven tagging suggestions
- Remote OCR (Azure AI)
They do look to be optional at least. I’d probably be a bit more concerned if the code was largely written with LLMs, given the large number of commits in what looks to be only a few months.
I upgraded and havent enabled the AI stuff and I literally cant tell it’s there, not sure why people are freaking out. Seems like theres a lot more in it as well.
Just installed via Ansible and it’s running without any issues so far. As far as I could tell visually, it’s running much faster.
That’s good to hear. The performance of the UI was always an issue for me.
Goddamnit. I hope someone forks 2.20.
Why? The new LLM features are opt-in. Just don’t turn it on if you don’t want them. Or is there something else you don’t like?
If they weren’t using LLMs to code, I’d mostly agree with that. But projects that put effort into integrating LLMs likely use them as well (as is the case here).
I wonder if there’s a blogpost or something somewhere summarizing the main changes.
I bet you could use AI summarize all the changes. LOL
spoiler
Breaking changes
The biggest upgrade risks are removal of API v1 compatibility, dropping support for API version 9, dropping Python 3.10, removing support for document and thumbnail encryption, removing pybzar as a barcode reader, and changing the pre/post consume scripts so they no longer accept positional arguments.
There’s also a change that decouples OCR control from archive file control, plus a refactor of advanced database settings to allow more user configuration. Search and indexing
A major headline item is replacing the Whoosh search backend with Tantivy, along with unifying text search around Tantivy.
The release also adds a number of search-related improvements, including better fuzzy matching, highlighting in title/content searches, better handling of special characters and CJK text, and fixes for date/query compatibility during the migration. AI and document intelligence
Paperless-ngx v3 adds a larger AI feature set: Paperless AI, remote OCR with Azure AI, support for Ollama embeddings, direct LLM language selection, LLM timeout and context settings, and a revamped AI vector store approach.
It also includes fixes and safety improvements around AI indexing, chat, embeddings, and permissions, suggesting this area was a big focus of the release. Documents and workflow
The release introduces document file versions, sharelink bundles, a document parser plugin framework, and several workflow enhancements like more actions and trigger filters.
There are also new or improved document-management behaviors such as moving to trash, password removal actions, better merge dialogs, live document updates, saved view sharing, and more detailed document attributes. UI and tasks
The task system was redesigned, with a new Tasks UI, task summaries in system status, and better monitoring access.
On the frontend side, the release updates Angular to v22, moves toward zoneless reactive behavior, improves mobile search behavior, refines thumbnails and image loading, and generally modernizes the UI. Performance and storage
Performance work is a major theme: there are database indexes for common query patterns, SQLite tuning, faster imports/exports, memory reductions in document importing, and indexing optimizations.
The release also standardizes checksums on SHA256, improves bulk operations, and adds a variety of backend and file-response fixes that should help reliability and speed. Practical impact
For a fresh install, v3.0.0 mainly means newer architecture, better search, and more AI/document features. For an existing install, the main things to watch are the breaking changes around API compatibility, Python version, encryption support, and the search/index migration from Whoosh to Tantivy.
deleted by creator
I do understand and share people’s hatred of corporate-owned, centralised cloud AI.
I understand (though share to a lesser degree) people’s ethical concerns about how these models were trained (copyright has been broken for a while now, this just exacerbates it).
But this level of outrage about using local models on an opt-in basis strikes me as hysterical.
Feel free to change my mind.
It reads as especially hysterical in this context, because Paperless is an automatic document categorization system, and I’m really sure what they think the automatic part of that is if it’s not “AI” of some broad description. Paperless, Papermerge et al are basically wrappers for machine learning tools and have been for as long as they’ve existed. LLMs are a natural and obvious fit for the kind of work these applications exist to do.
This just feels like someone reading “Improved AI pathfinding” in the patch notes for a video game and screaming “OH MY GOD IS NOWHERE SAFE?!”
True, although we must concede there is a difference between the simple machine learning already present in paperless and LLMs where, depending on quantization, CPU-based inference doesn’t get you very far.
You should be embarrassed by this.
Guess you won’t be making issues there for a while eh
Lol, you suck
The saddest part is that you clearly generated that message using an LLM. Or are you going to tell me that em-dashes are part of your everyday writing?
2026
complaining about em-dashes
You are completely ignorant about where do AIs even get their data from, yes?
A free knowledge cookie for you:
The em-dash is the language standard sign in Español for dialogue tags. Literally.
I have a dedicated button for em dashes on my keyboard /s
… I do, on my custom macro pad.
FUCK IT THEY GOT PAPERLESS-NGX TOOO???
Fedis, what’s the best alternative?
I’m going to go ahead and try to use this comment as a teachable moment.
For one, this comment barely skirts the line to me as acceptable. If you aren’t interested in AI topics, thats what the [AIT] and [AIP] tags are for - for you, as a user, to choose what you are interested in and what you are not.
The first part of your comment, by itself, I would have simply removed. You asked for alternatives, and are being rightfully replied to pointing out that this is all under the subject of automation, which (to me) barely puts this above low effort.
If you don’t want to see AI discussions in the future, use the tags to filter. If I see more comments like this that are barely, and entirely subjectively just above low-effort, the bar is going to raise just a hair and they will be removed, as they will come across like trying to skirt the rules.
what’s the best alternative?
You’re asking for suggestions for an automatic document categorization tool that doesn’t use any automatic document categorization?
You don’t need a knife to make a knife, and things like OCR are different from AI (or from “AI”).
How about one that doesn’t use LLMs to code?
