home.social

adingbatponder :nixos: 👾

View on fosstodon.org
  1. This says rns.recipes/forum/general/re-d code is to be taken off or IP based repos. I was not able to join the discussion - was locked. Just wanted to say that github is the reference location for code (a shame, but still fact), and that I am using as a robust meshed git repo system & so no sheep. As mirror repo codeberg/github are ok locations just to get it out & get bootstrapped. To only get the code over reticulum itself seem not beginner-friendly.

  2. Can anyone point me to a serious /document-reading stack that actually works?

    I’m not looking for Tesseract/OCRmyPDF/Paperless/Stirling-style OCR. I’m looking for self-hostable vision-language OCR or document AI: printed documents only, no handwriting, but strong on bad scans, tables, columns, forms and layout. Hardware is negotiable; accuracy is not.

    Is there a known model/server/pipeline/GPU setup approaching ChatGPT/Claude OCR, or is this becoming another cloud API/token lock-in?

  3. Is there no site as good as what the big tech cos can offer with their LLM and VLM tools which are charged for or read your data? Is FOSS in any way making headroads into ?

  4. Does anyone have problems with this printer ET-16655 # ET16655 as such or under Linux as a LAN (cable) connected printer?

  5. @simondueckert @karstenpe Ich habe
    Qwen2-VL-2B-Instruct
    Q4 (~1.3 GB on disk )
    Multimodal projector (vision encoder bridge) at Q8
    probiert und es kann mit dem CPU (ThinkPad T480s i5-8350U 16Gb RAM nixos) sehr langsam *einzelnen* Tabelle Einträge oder < 60% sicher ergebnisse richtig OCRen wo tesseract scheitert aber es war eine irre Quälerei mit claude code.
    Haben Sie ein local LLM gefunden???

  6. Is there a FOSS model that does using machine vision learning tricks to get text of tricky docs where fails? Using to get & co to do it is stomach-turning....

  7. What is the best for of tricky input such as tables or just for high fidelity output in general?

  8. The prolonged attack on servers is affecting updates. I’ve been wondering whether a system like could maybe help as a secondary mirror and reduce the impact of outages or attacks. I mirror my Codeberg repo to a Radicle seed node to cope with outages. That does not mean I can push when Codeberg is down. The single-source-of-truth issue is still unclear to me, but downloading a pre-attack version should be possible. Pushing new work during an attack is tricky.

  9. Get and email saying first my payment method has changed
    (I think I requested this online about 6 months ago)
    then and email immediately afterwards saying account is being closed for one of the following reasons
    no vehicle registered
    not enough credit to pay for a crossing
    no valid form of payment
    without saying which.
    This is so pityful. Absolutely diabolically bad.
    Or is this a hacked account?
    Lord help us if this is how things work from now on.

  10. @freya I am trying out on my "homelab" server but was most surprised that it seems one has to do all the ZFS setup external to the which only lives within the manually made ZFS pool environment. That seems to mean I need a sort of orchestration script like tool to match flake /persist data settings (am using lite currently) to the ZFS setup hoo-ha. Works fine but is not declarative at all really. Any ideas there ?

  11. Solution in for machine ThinkPad seems to be:
    services.xserver.enable = true;
    services.displayManager.gdm.enable = true;
    services.desktopManager.gnome.enable = true;
    services.desktopManager.gnome.flashback.enableMetacity = true;
    # At the GDM login screen, click the gear # icon to choose between:
    # - "GNOME" (Wayland)
    # - "GNOME Flashback (Metacity)"
    and
    nixos-rebuild boot + reboot is required (not switch) when jumping across major nixpkgs versions

  12. I am trying to update my flake and discover that is not supported by the latest but I also know that some of my hardware ( , VM machines running xrdp and x11spice) fails to show correct windows under so this is very un-linuxy and I feel like a sort of macos hardware obsolescence vibe ... Or is the new way forward here for the good old tech? Thinkpad once failed with fifo underrun on pipe B when running wayland

  13. Have I understood it right that an update of a in , which has a .git , is performed without staging it first... i.e. the lock files are used independently of git and are an exception to the rule that changes have to be staged before testing a build and switch?

  14. Could it be that if flake check crashes or freezes your machine then your flake and or your computer are officially silly? Is there a way of using an already extant remote build facility in the actual flake check also?

  15. Upgrading my on is making my computer sweat even when it has a remote builder to help... looks good... took 10 mins... Anyone got the AI tool with ? working on in nixOS , or at all ? Aha... need a new terminal for the ? to trigger the AI tool ... wowzers...

  16. @stadtkind @edwtjo Thanks for the input. FYI this is my playbook for this synch / reconciliation process when two flakes have deviated and you want to make them the same by using a repo, e.g. codeberg , as the single source of truth. The solution uses the branch main on the main computer which has the main flake that should be up to date always, & a branch on the other machine named after the machine. The other machine will always commit to its branch codeberg.org/adingbatponder/re

  17. @stadtkind @edwtjo Well I really got into a spin. When doing the above where you
    1. pull a new flake down into a directory with a new name /flake-new
    2. rename the old /flake directory /flake-old
    3. rename /flake-new as /flake and check this out as branch
    4. try nixos-rebuild switch --flake .#<machine_name>

    caused the rebuild command not to switch. It just kept giving me the old generation...because I was doing 4. from previous main but should be done from the new branch

  18. Any tips and tricks for using for : when the "mother flake" flake lives on a public repo & is set up to provide multiple for different machines built using:
    nixos-rebuild switch --flake .#<machine_name>
    I dug myself into a hole by copying the files from the mother flake repo to set up a new machine and the flake on my new machine has diverged & cannot be merged back into the mother flake repo because I did not *clone* it down so has no common git history

  19. 1/6
    I’ve been thinking through disaster recovery with , impermanence, and .
    A flake defines a system declaratively, but assumes storage (e.g. ZFS pools) and persisted data (e.g. /persist) already exist. Flakes don’t create storage—they only configure it. Right?

  20. Do you have a single for multiple devices using different in which for each device you have a specified /persist data location as part of an setup? I am looking for examples, tips and foot-gun alerts.... cheers.

  21. @yggverse Thanks. Might also be a groovy child access control system. Have been looking for this for a long time. I wanted to have a pc with necessary school links access but zero anything else. In the past the dependent domains problem has stopped me setting this up. I think schools when setting homework should be obliged to issue the whitelist for that task. Maybe with an AI tool using github.com/simonw/rodney one might be able to crawl the dependent domains and make a whitelist fast ??

  22. If you call DE on a Sunday you spend 10 minutes talking to a machine that does not understand "Ja". Then it puts you through to a person except the message then says we are closed. You cannot even BEGIN to imagine how utterly diabolically bad customer service is even when automated.

    #o2
  23. is a really useful and informative error that is thrown by my and Mastercard (begins with number 5) online payment attempt on the website but an payment card (also starting with 5) works fine. Bravo N26 which, as I have said as often as I can, utterly rocks.

  24. Thinkpad of 2013 (gift from friend),
    18.2 of 2017 (350 euros 64 Gb RAM 27" Retina display) ,
    CX43 (10 euros per month) online server,
    (SM in table on right) X11-WTR SYS-5019P-WTR Silver 4110 × 16 (basic parts 500 euros 2nd hand)
    CPU comparisons using

  25. @AAKL @theregister I have an excellent bank account which allows me for a small monthly fee to have 10 different sub accounts, and one of them has their own bank card. All my online subscriptions are linked to one of these sub account's bank card and the account cannot be negative and so has a float in it instead. This means that any charges that are unexpected exceed the float and are not paid. That way I can also force halt a subscription by emptying the float in the event this is needed.

  26. Why should the state take a percentage of electricity costs as tax? Why not a fixed amount per Kwh ? Why tax the essentials in a way geared to make those who can least afford the essentials even worse off?

  27. Seems command-line interface tools with good / help info are very powerful because automatic coding tools can easily interface with them & use them correctly, even without additional context. Interestingly, many machine-based coding assistants are extremely good at generating command-line oriented programs in the first place. Until now, I struggled to build solid GUIs, for example with . Keeping software using the CLI - as well - seems to provide big automation advantages.

  28. I was using with to play with and it became a bit of a mess, to say the least, for what I was trying to do. By accident my radicle server lost power & I had to restart & lost my claude session. So I took the opportunity to claude to and used github.com/ryoppippi/claude-co too in my . The results were astounding. The updated claude code explained the problem correctly (I had guessed what the issue was) & fixed it in minutes. Super.

  29. Found shell history tool in packages and signed up for synch too.

  30. Anyone using to manage a “mother flake” across multiple machines?
    I’m trying to confirm something: since Radicle is built on , does rad init still leave you with a normal .git working tree that nixos-rebuild switch --flake . is happy with?
    Or does Nix require a standard Git repo and can’t treat a rad://… repo as a flake input without an HTTP/SSH Git mirror? If you do this or similar, what’s your actual workflow?

  31. I am told is fab & it is ... but does it make sense to run on proxmox: one for each task e.g. ? Or is it best to put all one's eggs into 1 single basket in one bare metal server that does it all? Or what of a farm instead? Does that spread risk of downtime due to machine breakdown issues or is maintenance too hard? Any experiments or experiences to report ? Thanks. @homelab

  32. Video of flashing update via & bios using . Setup of . Detailed video youtu.be/z6JsDy1YC6w
    2017 Mainboard setup using setup utility . Setup & useful background info bmc network configuration to failover for hardware stress testing using 3.7GHz youtube.com/shorts/rmhnG48P8Rw

  33. Why on do we pay on 2nd hand products? If I buy a used computer that cost the supplier to buy in used 200 and it is sold on to me for 300 because it was serviced or upgraded and delivered, VAT should only be due on the 100 extra, right ? Would using such a VAT regime help us support reusing things and reducing waste? This could in a democracy be achieved surely?

  34. @stephen @markmcb

    Very interesting firewall settings. Thanks for sharing them. Question into the void: is there a “menu-style” FOSS setup for —prebuilt profiles with different strictness levels one can apply and test? I’m running , but I don’t have a systematic way to validate beyond “tweak until it works.” I’d love a repeatable approach to reach a good protection baseline with minimal site access .

  35. @joschi @homelab @homelab_de I chickened out and made and - see my repo flake under features/network-appliance repoducible.org
    Wanted first to see what was going on on my network....

  36. Is there an product that lives in a base as in image ?

  37. youtu.be/Zu87ZLwZXdA?si=8ZxzrL mediated of text data between two computers using
    and the loudspeaker of one to send to the microphone of the other at 300 to explain the building blocks of a system.

  38. @zer0unplanned This post looks very useful. I made a similar tool (still in beta with errors) to grab screenshots and raw data of manual pages for commands, in this case info, and to present the different layers that are the OS, app and nixos settings. The program ssh-key-manager is in my flake repoducible.org ( features/security/packages can be pulled out into another flake) & bit of a monster. Yours looks ultra compact

  39. @momo works! is an easy way to ensure you have all files of interest available even in aeroplane mode, because the repositories you chose to follow are stored locally & automatically kept at their most recent version from when you were last online. Easy setup. Practical for dev ops (systems that keep working when web is unreliable or services disappear). repoducible.org (app.radicle.xyz/nodes/radicle. /features/radicle) has a part example for setup.

  40. A that lets a be auto-added to any flake. Shows & processes.

    codeberg.org/adingbatponder/re

    (Handles existing installs: adds another . Import tested on a few machines. Feedback or issue reports welcome.)

    @arianvp @mdione @EduNET_LK @bustikiller

  41. @mdione @EduNET_LK @bustikiller Great tip. I just spun up a dashboard in a showing the data. I do not get any exciting data but that is good, right ? LOL Thanks. Details in pastebin.com/VU8RYPJq

  42. Background:
    for my I set up in and have a functioning system with a dashboard ( info here codeberg.org/adingbatponder/re ).
    Preliminary plan:
    I now want to go to and system that blocks threats detected. Current plan is with but there is no GUI for that it seems, and it is a bit clunky and black-boxy.
    Question: What are the more user-friendly options for an IPS front-end / GUI ?
    Thanks!