shaked_coffee

Framework just announced their Desktop computer: an AI powerhorse?

Recently I've seen a couple of people online trying to use Mac Studio (or clusters of Mac Studio) to run big AI models since their GPU can directly access the RAM. To me it seemed an interesting idea, but the price of a Mac studio make it just a fun experiment rather than a viable option I would ever try.

Now, Framework just announced their Desktop compurer with the Ryzen Max+ 395 and up to 128GB of shared RAM (of which up to 110GB can be used by the iGPU on Linux), and it can be bought for something slightly below €3k which is far less than the over €4k of the Mac Studio for apparently similar specs (and a better OS for AI tasks)

What do you think about it?

I've installed "Overcooked 2" and "Moving Out" through Heroic Games Launcher (flatpak) and then used the ad-hoc option to add it to steam (also installed as flatpak). If I try to launch the game from Heroic it works properly, and it works also if I try to launch it from the command line with the command it added to steam

flatpak run com.heroicgameslauncher.hgl --no-gui --no-sandbox "heroic://launch/legendary/8e29583ae4b44a21883038668f7e301e"

but if I click the PLAY button on steam, the game fails to launch and the following error is printed on terminal:

chdir "/var/home/lez"
Game Recording - would start recording game 12260096395435835392, but recording for this game is disabled
Adding process 532 for gameID 12260096395435835392
ERROR: ld.so: object '/var/home/lez/.var/app/com.valvesoftware.Steam/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/var/home/lez/.var/app/com.valvesoftware.Steam/.local/share/Steam/ubuntu12_32/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS32): ignored.
ERROR: ld.so: object '/var/home/lez/.var/app/com.valvesoftware.Steam/.local/share/Steam/ubuntu12_64/gameoverlayrenderer.so' from LD_PRELOAD cannot be preloaded (wrong ELF class: ELFCLASS64): ignored.
pid 534 != 532, skipping destruction (fork without exec?)
Game Recording - game stopped [gameid=12260096395435835392]
Removing process 532 for gameID 12260096395435835392

Anyone else faced this issue? If so how did you fixed it?

I've made the following backup script for my immich stack to be automatically run every day

# Load variables from the .env file
SCRIPT_DIR="$(dirname "$(readlink -f "$0")")"
set -a
source "$SCRIPT_DIR/../.env"
set +a

# Create a dump of the database and back it up
docker exec -it immich_db pg_dumpall -c -U immich > immich/latest_db_dump.sql
rustic -r $BUCKET_NAME/immich backup immich/latest_db_dump.sql --password=$REPO_PWD

# Backup the library, uploads and profile folders from the upload volume
rustic -r $BUCKET_NAME/immich backup immich/uploads_v/library --password=$REPO_PWD
rustic -r $BUCKET_NAME/immich backup immich/uploads_v/upload --password=$REPO_PWD
rustic -r $BUCKET_NAME/immich backup immich/uploads_v/profile --password=$REPO_PWD

# Apply forget policy
rustic -r $BUCKET_NAME/immich forget $RUSTIC_FORGET_POLICY --password=$REPO_PWD

and when I test it everything works properly, and the created sql dump file is complete and properly backed up.

However, when the execution is triggered automatically by a cronjob (as specified in this crontab line)

"30 3 * * *    root    /home/admin/WinguRepo/scripts/docker_backupper.sh"

(the line is taken from the nixos configuration file, that's why it also contains the user executing the operation)

it seems something breaks in the dumping process, because the script completes successfully but the sql dump file is an empty file (as can be noticed in the following output of rustic -r myrepo snapshots

snapshots for (host [wingu-box], label [], paths [immich/latest_db_dump.sql])
| ID       | Time                | Host      | Label | Tags | Paths                     | Files | Dirs |      Size |
|----------|---------------------|-----------|-------|------|---------------------------|-------|------|-----------|
| 10a32a83 | 2025-01-06 20:56:48 | wingu-box |       |      | immich/latest_db_dump.sql |     1 |    2 | 264.6 MiB |
| 1174bc2e | 2025-01-07 12:50:36 | wingu-box |       |      | immich/latest_db_dump.sql |     1 |    2 | 264.6 MiB |
| 00977334 | 2025-01-08 03:31:24 | wingu-box |       |      | immich/latest_db_dump.sql |     1 |    2 |       0 B |
| 513fffa1 | 2025-01-10 03:31:25 | wingu-box |       |      | immich/latest_db_dump.sql |     1 |    2 |       0 B |
4 snapshot(s)

(the first two snapshots were manually triggered by me executing the script, the latter two instead are triggered automatically by the cronjob)

Any idea about what is causing this behavior?

EDIT: Solution found thank's to @farcaller@fstab.sh comment:

You don’t need -it because you don’t run an interactive session in docker. It might be failing because you ask for a pseudoterminal in an environment where it doesn’t make sense.

cross-posted from: https://feddit.it/post/13363649

Hi! I'm a flutter developer and I've recently switched from Ubuntu to Bluefin (based on Fedora Silverblue) I was trying to install flutter and all the dependencies it asks for when running flutter doctor and while for clang++, cmake and ninja I have been able to install them with Homebrew (brew install llvm cmake ninja) and the warning went away, flutter seems not to detect GTK 3.0 development libraries when installed through homebrew (brew install gtk+3)

After having looked a little bit online, I've seen that the problem is probably that it's searching the libraries in the wrong place and in fact even when I run pkg-config --modversion gtk+-3.0 I see the library as missing. Therefore, I tried setting the PKG_CONFIG_PATH env variable to the folder where brew installed my libraries (/home/linuxbrew/.linuxbrew/opt/gtk+3/lib/pkgconfig) and after that the output of the pkg-config program above is the one expected. However, this doesn't seem to fix the issue with flutter doctor... why? where should I specify this path for flutter to notice it then?

EDIT: I found out that I was almost right! The PKG_CONFIG_PATH env variable is what flutter looks at to find GTK-3 libraries, but should be set to /home/linuxbrew/.linuxbrew/lib/pkgconfig (if pkg-config has been installed through brew as I did) and then everything will work properly.

Hi! I'm a flutter developer and I've recently switched from Ubuntu to Bluefin (based on Fedora Silverblue) I was trying to install flutter and all the dependencies it asks for when running flutter doctor and while for clang++, cmake and ninja I have been able to install them with Homebrew (brew install llvm cmake ninja) and the warning went away, flutter seems not to detect GTK 3.0 development libraries when installed through homebrew (brew install gtk+3)

After having looked a little bit online, I've seen that the problem is probably that it's searching the libraries in the wrong place and in fact even when I run pkg-config --modversion gtk+-3.0 I see the library as missing. Therefore, I tried setting the PKG_CONFIG_PATH env variable to the folder where brew installed my libraries (/home/linuxbrew/.linuxbrew/opt/gtk+3/lib/pkgconfig) and after that the output of the pkg-config program above is the one expected. However, this doesn't seem to fix the issue with flutter doctor... why? where should I specify this path for flutter to notice it then?

EDIT: I found out that I was almost right! The PKG_CONFIG_PATH env variable is what flutter looks at to find GTK-3 libraries, but should be set to /home/linuxbrew/.linuxbrew/lib/pkgconfig (if pkg-config has been installed through brew as I did) and then everything will work properly.

The ads of this company called Airback that was making this backpacks where you could fit more stuff thanks to a built-in vacuum system appeared on my feed multiple times, but since the smallest bag they were making wasn't small enough to be used as free bag for lowcost companies I never really considered them. Now though I see there is a new model on their website called Airback Go that is made exactly for that purpose, and so I was wondering: does anyone here have actually tried it (or its bigger brother)? How is the quality of the backpacks? Is the vacuum thing something worth the price or it's more of a marketing strategy?

I'm currently running a Nextcloud instance in docker, using the "multiple containers" method, but I recently discovered that the reccomended method to run it is using this "All In One" image I previously didn't know about, and so I wanted to try to migrate to that setup (also in order to have easily also the office and whiteboard features that atm I don't have on my instance + have easier Backup-restore process)

The problem is that on my server I'm using traefik as a reverse proxy to expose services to the internet, and it is working on a specific docker network (called traefik_net) where also every container that should be exposed is connected, and from the official documentation of Nextcloud AIO I really don't understand how am I supposed to configure it to work in a setup like mine (mainly because the mastercontainer creates all the containers it needs on a network called nextcloud_aio and I didn't find a way to change that and where to set the proper traefik lables)

Anyone that is running AIO behind a traefik reverse proxy maybe can help me to understand?

I was previously using PopOS! 22.04 on my tuxedo laptop and I'd installed on it Howdy to take advantage of the IR camera and have a windows hello alike face recognition feature.

Everything was working fine, but after some time GNOME 46 and its new goodies were too tempting to stick with Pop's old GNOME version (at least for me) and therefore I switched to Ubuntu 24.04

However, when I tried to install howdy using the PPAs as I did with Pop I noticed it wasn't working because of some changes that were made regarding on how Python is managed, and I couldn't find a solution for that. Looking at howdy's GitHub issues, there are a lot of them talking about this problem that seems to be started with 23.x versions already, but having so many issues created a bit too much confusion to me and I didn't manage to find a working solution from there.

Is there anyone here using Howdy on Ubuntu 24.04? How have you managed to install it?

Today thanks to a NetworkChuck video I discovered OpenWebUl and how easy it is to set up a local LLM chat assistant. In particular, the ability to upload documents and use them as a context for chats really caught my interest. So now my question is: let's say l've uploaded 10 different documents on OpenWebUl, is there a way to ask llama3 which between all the uploaded documents contains a certain information (without having to explicitly tag all the documents)? And if not is something like this possible with different local lIm combinations?

I recently finished the episode of The Verge's podcast #Decoder with the interview to Bluesky's CEO and it seems a quite interesting project. At the beginning I wasn't looking really into it because of their choice of using a new protocol instead of the existing ActivityPub, but after listening to her and the reasons behind this choice maybe I'll give them a chance.

What do you think? Do you use it alongside with the fediverse?