March 15, 20251 yr Hi @ all, I am at a loss. Usually I try to solve me problems by myself googing or asking chatgtp/perplexity. I recent years I think I have become quite good at solving these problems even though IT stuff is just a hobby and my dayjob has nothing to do with any of it. To help you all understand my problem I will first tell you a little bit about my system: Hardware: CPU: AMD Ryzen Threadripper 1950X 16-Core @ 3400 MHz Mainboard: ASRock X399 Professional Gaming American Megatrends Inc., Version P4.01 BIOS dated: Tuesday, 02-05-2023 RAM: 128 GiB DDR4 Multi-bit ECC HDDs: Array 7x 10 TB WD red (2x Parity) Cache: 2x Kingsotn NVME 2TB Addtional Storage (a couple of usb Drives for Backup options) and Samsung SSD 980 PRO 2TB nvme and SD Ultra 3D 2TB ATA My Server is a mixture of FIle Storage and computing - what does that mean? - I use my server to make my life easier. I have mainly 2 VMs running (my General purpose Windows 11 VM and Home Assistant) Then I have quite a few Dockers running: Name of Docker Container (Autostart or Manual start) why I use it/what i use it for antennas (Autostart) to have dvb input on emby ApacheGuacamole (Autostart) to have access to my vm via web browser from everywhere binhex-emby (Autostart) to access my media binhex-krusader (manual start) to move files around calibre (Autostart) to sort my ebook library dashy (Autostart) to have a startpage for my wife so she can access all the docker containers etc Dozzle (manual start) for logging dockers evcc (Autostart) to charge my ev via my pv firefox (Autostart) so i can browse uncensored from anywhere in the world. frigate (Autostart) for home assistant all together 6 cams round the house jDownloader2 (manual start) i rarly use this mariadb (Autostart) database for nextcloud MusicBrainz-Picard (manual start) when i get around to rip a cd myself which rarly happens these days nextcloud (Autostart) as a google service replacement Nginx-Proxy-Manager-Official (Autostart) so i can access all my containers with my own domain ollama (Autostart) to use AI open-webui (Autostart) web UI for ollama Paperless-AI (manual start - since it keeps crashing) i want to automatically tag and sort my papers (see below) paperless-ngx (Autostart) - do digitalize my incoming mail - everything is saved inside a nextcloud folder which automatically syncs with nextcloud Redis (Autostart) this is needed for ollama stable-diffusion (manual start) I rarely use it but never had any troubles using it teddycloud (Autostart) to play custom tonies for my kids tvheadend (Autostart) to distribute my tv signals vaultwarden (Autostart) to store my passwords what happens is that i want to use paperless-ai the server completely freezes, it becomes unresponsive (i cant even ping it) and the log files do not help me either (or i am not looking at the right place). usually this happens whenever i scan documents using paperless-ai but lately it also happened when using the open-webui. The weird thing was, that after the next reboot the nvidia driver plugin seemed to be missing. Also as a sort of a fail safe restart method, I had a shelly plug between the PSV and the Server, so that when the server crashed i could use wireguard (running on my router) to turn the power to the server off and on again therefore restarting it from whereever. Now i had to remove the shelly plug since it kept shutting down my Server after a couple of seconds. So i replaced the shelly plug with another shelly plug - same exact behavior, the server starts (does not even come up to boot unraid) and just turns of. Without the shelly plug i now don't have this behavior anymore. I added a couple of logs hopefully someone can help me. unraid.log.1.gz unraid.log.2.gz
March 16, 20251 yr Author 3 hours ago, bmartino1 said: i would have you remove paperless and look into using local AI docker. Which one paperless nxg or ai? Also how is the local AI docker going to do the same as any of the paperless dockers? I didn't quite understand?
March 16, 20251 yr your have seperated your ai stack to run the paperless model. you have 3 docekrs when 1 can do that work and better. local ai you can download a model and interact with eh api / json to run the paper model... local ai already has a web interface to interact with the model and run the model you chose at start. as yoour doing to much at once casuing bottle necks review:
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.