Joly0

Hey guys, i noticed after rebooting my unraid server, HA-in-a-box is not able to start the vm and throws this error: Version check passed. Continuing... Host path for '/domains' is /mnt/user/domains Highest Q35 machine type available is pc-q35-10.2 The default VM network type is eth0 Putting default ha vm icon in place Icon copied successfully. Home Assistant directories are already present...continuing. There is already a vdisk image here...skipping. As this is an OVMF VM, I need to create an NVRAM file. Creating now .... Formatting '/etc/libvirt/qemu/nvram/d8e05b6b-07e0-4a41-b1b4-2a961f908de2_VARS-pure-efi.fd', fmt=raw size=65536 All required files are present. Attempting to define the VM... Failed to define the VM. There was an error during the VM definition process. Not sure though how to solve this problem, i have no idea why its failing there and how to fix it. Starting the container manually works though and starts the vm properly

Hey guys, someone maybe knows why these "yunkan" containers keep re-appearing under "recently added" every few days? The templates are found here GitHubyunkan-unraid-templates/templates at main · mrtian2016/yu...云瞰 YunKan — Unraid Community Applications 模板仓库(CPU / OpenVINO / CUDA 3 个 amd64 变体) - mrtian2016/yunkan-unraid-templatesbut i cannot see any reason why they are re-added all the time, i just see they update the "changes" part in the templates when its updated. Its super annoying to see them every few days again and again.

Just looked through it, couldn't find it, then realized I linked the wrong repository (but still correct for container templates). The actually file you should edit is this one: https://github.com/Squidly271/AppFeed/blob/3127c70c026d256115bce4d64f3777afdd6c0862/statistics.json#L1856

The repository is here https://github.com/Squidly271/Community-Applications-Moderators If you search for your username in the repo, you will find the blacklist file I was mentioning. Just delete the part for your main branch and PR it

For your information, squid has the community applications moderations repository on his GitHub, which includes everything for CA ,including blacklisted stuff. You basically have to create a PR to un-blacklist (remove your repository from the blacklist file) and let squid merge it. It's super simple and straightforward. I have done that myself 3 times already (not for plugins, but for docker container templates, which got falsely marked as duplicates)

Hey, i noticed after a reboot the agent is simply gone. When i try to access it for example through the normal unraid webui terminal and run something like "claude" i get this error: bash: /usr/local/emhttp/plugins/unraid-aicliagents/agents/claude-code/node_modules/.bin/claude: No such file or directory The "agent store" says its installed. I either have to uninstall and reinstall or (if available) update to get claude (and others) back working.

Hey ,after the latest update something broke regarding removing and creating zram swap. So i had one running before the update. Then updated and wanted to re-create the swap file to use the higher swappiness (as far as i could read in the changelog, this is only applied to new zram´s?). I removed the existing zram, which looked to be ok, then the page had to be reloaded, but after reload it still showed the "remove" button. After doing that for a another 3-4 times, it finally reloaded and showed the "create" button instead of "remove", but when i clicked create, the log showed that it couldnt create it, because the file already exists. I tried thisn a few times over an over, always getting the same problems with removing and re-creating and finally gave up. Then randomaly i looked on the dashboard card and saw, that the zram appears to be created correctly with the correct 150 swappiness.

Yes, trying to from the default unraid shell, would be great if there was a way to do this for testing and debugging (currently trying to troubleshoot working with an claude code to opencode bridge). Another thing: This has been running for over an hour, it was stuck for most of the time now at the "Backing up active RAM sessions..." step. Not sure how long it will take for the next step, but this rough for a plugin update

Also another thing i have noticed: When i have to debug mcp servers using npm/npx they do not use the correct paths and usually defualt back to /root/ directory instead of the /tmp/unraid-aicliagents/work/root/. Even when starting them like this /usr/local/emhttp/plugins/unraid-aicliagents/bin/npx it doesnt work either and they still usually place their place under /root/. It would be great here if there was some way to easily get into the same environment as cli´s would run under. I assume you are doing some path shenanigans or something similar

Btw, just did the latest update with the migration and everything and afterwards the directory /tmp/unraid-aicliagents/work/root/ was empty, so all sessions, configurations, etc were gone. I had to manually copy back those files from the backup location under /boot/config/plugins/unraid-aicliagents/persistence/root

Just came here to say the incognito feature is pretty crazy good, in my opinion volumes should be masked aswell, because they unusually contain data to make it obv which container is which, which for example wouldnt or usually isnt the case for things like networks, ips or ports

Also could you explain how and where the different cli´s are installed, cause i would like to use things like claude code usage monitor or meridian/opencode-with-claude and they require things like the opencode.json file or the config location for claude, and i cannot find them. I searched under /tmp/unraid-aicliagents but couldnt find for example the opencode.json file

Hey, two other things i noticed is that the sync is running indefinitely (atleast for me was running for over 10 minutes and didnt finish) and the schedule for that is not saved nor applied in any way, so saving doesnt work

Yes, can confirm, bug is fixed and looks normal again (like it used to be). Thanks for quick fix

Tripple checked, but no, there are no other plugins that would modify the docker page, just folderview3. Also tried to replicate the same problem in chrome and there its exactly the same as in edge (so same problem as in firefox, but with thicker scrollbar)

So this is in firefox browser (latest version) on windows, but i can replicate the same problem in edge, though there the scrollbar at the bottom is super thick: Compose plugin version is: Compose plugin plus 2026.03.28 Folderview3 version is: 2026.03.28.1 Also yes, i tried a refresh, logging out and logging back in, clearing cache, etc

Hope you can see that here. That is a full screen screenshot from left to right and there is the scrollbar at the bottom. And containers on the right edge ar cut off here: This was not the case before. There was no scrollbar, containers were just hidden when there were too many. Its not even all containers showing, the meilisearch container in the bottom row is not even the last in that folder, there are still 2 more containers

For some reason after the latest update the folders in the docker tab (not sure about the vm tab, i dont use that) are basically full width, adding a scrollbar at the bottom. This was not the case prior to this udpate

Something else that is kinda annoying is, that i cannot scroll up or down in the cli itself. I can only scroll through my inputs, but i cannot see earlier repsponses

Yes, works. For some reason after installing the claude command was not there (installed it 2 days ago; always showed claude as not available command), but now its working as it should. Another thing i noticed, not sure if its a bug or intended, i do not have gemini cli installed, only claude. When i create a new workspace it shows in the popup on top "claude", so i assumed it would use calude then. But when i created the workspace i got the error, that gemini-cli was not installed. I had to remove the workspace, create a new one and click on "claude" separetely to make the workspace use claude. The "claude" button also gets a yellow-ish border were before its grey. Though when i create a new workspace in the same tab as before, i do not ahve to select claude again.

Hey, cant seem to find a way to add mcp server to cli´s (in my case claude, which would usually be done with `claude mcp add`). Is there a way?

Hey, not that i have anythign against you or your templates (they are very well done, gj on that :D ) but you might have noticed that i already maintain templates for open-webui https://github.com/Joly0/docker-templates/blob/main/templates/open-webui.xml and ollama https://github.com/Joly0/docker-templates/blob/main/templates/ollama.xml and why you didnt just create a PR for those templates? The thing is, CA regularly blacklists duplicate templates and two templates creates confusion https://www.reddit.com/r/unRAID/comments/1riwnhc/which_ollama_you_choosing_ollama_or_ollama/ Also wanted to mention rule 16 here https://forums.unraid.net/topic/87144-ca-application-policies-privacy-policy/

Looks good by the way

I did not claim anything, i just noticed, that you added a license, but all good if you have already added a notice. Its not liek i am scanning your repo 24/7. Also, it appears to me, that you might have misunderstood my initial questions or i might have asked them in a way that you could interpret as rude. Just to clarify, this was not and is not my intention and i am sorry if my questions were not well written and i apologize for it. I am not against you or the project that you are trying to take over. I am also not against you vibe-coding this project to continue developing it. I just wanted to clarify that maintaining an open-source project is more than just prompting an ai to do stuff and it makes a difference if you know how to code from the beginning and use ai to help you code or if you have zero to low knowledge in any programming language and use ai as your only source of truth. So my initial questions wasnt against you or against vibe-coding in itself, but just wanting to know what experience you have in coding in general and maintaing an open-source project.

I would also like to add another thing to the input i am giving here: You cannot just add your license to the codebase and call it a day. That license is not legally binding and infact its not even allowed (would count for me aswell as i did the PR) per copyright law. Infact folder.view and folder.view2 are "all rights reserved" projects (wasnt even allowed for VladoPortos to fork it in the first place). There are only 3 options you have that would make the license hold: 1. You have fully rewritten the code and no original code is left in there, than you can re-license the code 2. You add a notice that only your own code (and contributions) are covered by the MIT license. 3. You ask all original creators and contributors for permission to change the license (so scolcipitato, VladoPortos and all the contributors). Just wanted to tell you that, as i just saw you added a license. Was in a same situation a while ago and just wanted to share what i know.