If you have options and the space go for 64bit CPU and midi sized office tower for expandability. If you want hardware acceleration for playback you need the check the docs (like immich vp9 > intel 9th gen CPU). What do you mean by cloud service?
This community won’t be much helpful but there is one specific for local LLMs. I think it was ai_ and some other…
Casual user here and by far no experts. Just wanted to leave my 2 cents about OMV and starting a homelab. At the beginning with OMV, it was nice to have everything tied up into one space with an gui but as time progressed I had more and more problems to do the stuff I wanted, like Home assistant VM with specific network settings. Maybe it was the lack of experience at the time but I switched to plain Debian and had no headache since.
About the Hardware, my first plans were just backup pictures from phones so I went with an used Lenovo SFF 7th gen Intel with 16GB. With time it’s escalated to Nextcloud, jellyfin, paperless ngx. Idle is still around 15W and only paperless hammers the CPU when importing scans. The only downside off the SFF is the lack of space.
My advice for the hardware would be to start small and cheap, preferably used. With some time you will see what you really need and have the experience to make it happen.
Just wondered who would still be on Reddit while self hosting and behold it’s corpo… Good luck to anyone who wants to participate though
This kind of posts from random user, especially with those names are a clear sign of cracks in the bubble…
So yeah, space will be an issue if you would go small… Some SSF may be enough but you will mostly need some special cables though. I got an SSF thinkcentre and managed to put 3x 2,5" HDDs inside (needed some machining).
Basically I would recommend something like an i5-9400 with 16gb ram in an tower case. Also have the PSU in mind, because some office PCs come with really low wattage, that could limit your expandability.
If you really want to stay with your mentioned software, 8GB could be enough though.