12
[AIT] Local LLM for Server Setup Guidance?
(lemmy.world)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 26 Jul 2026
12 points (66.7% liked)
Selfhosted
60951 readers
770 users here now
A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don't control.
Rules:
-
Be civil.
-
No spam.
-
Posts are to be related to self-hosting.
-
Don't duplicate the full text of your blog or readme if you're providing a link.
-
Submission headline should match the article title.
-
No trolling.
-
Promotion posts require active participation, with an account that is at least 30 days old. F/LOSS without a paywall has exceptions, with requirements. See the rules link for details. Tags [CBH] or [AIP] are required, see the links in Rule 8 for details.
-
AI-related discussions and AI-involved promotional posts have additional requirements for tagging, as noted in Rule 7 and the AI & Promotional Post Expanded Rules post, and find example disclosures here.
Resources:
- selfh.st Newsletter and index of selfhosted software and apps
- awesome-selfhosted software
- awesome-sysadmin resources
- Self-Hosted Podcast from Jupiter Broadcasting
Any issues on the community? Report it using the report flag.
Questions? DM the mods!
founded 3 years ago
MODERATORS
Easy setup would be to use koboldcpp + SillyTavern + Gemma 4 26B A4B GGUF of the largest quant your graphics card can fit together with your context.
Remember to setup SillyTavern to allow network connections and create a user and password, default installation is localhost only.
Set the temperature to 0.3 if using IQ3_M, higher quants allow higher temperature, and make sure all the formatting templates inside Advanced Formatting are set to Gemma 4.
Then use either Gemma 4 itself or the free google.com AI to create some "W++ Denze with Horizontal Lines summaries for Gemma 4" of the relevant documentation of the latest version of the home server apps your friends are gonna install.
You should ensure the resulting lorebook entries are no more than 1k tokens each to leave some context for your friends chats, add it to a character card in SillyTavern and set some keywords to allow them to load dynamically and not stay in context memory all the time.
Regarding the character card you can ask Gemma 4 to write that for you too, I find "Write a character card in W++ Denze with Horizontal Lines style for Gemma 4 with this name, personality, attitude and skillset" works well for that.
Then ask it to write a "First message prompt that starts with X, continues with Y and ends with Z for that character card" and you get a first draft to rewrite and paste into the "First Message" of the character card. The first message works as a template that Gemma 4 will imitate when you chat with it. Then simply try the chatbot out before letting others use it.
I find my own Gemma 4 26B A4B IQ3_M works well for practicing hiragana and katakana, discussing programming or troubleshoot existing code or writing a small function but it can't be expected to write a correct DatabaseService.cs from scratch and stuff like that.
Oh and forget about finding good cards and lorebooks for SillyTavern use online, most users use it for NSFW Roleplaying chats. It is a very easy UI to use to create harnesses for your local LLM though.
Depending on the use and the size of your graphics card you might be better of trying Gemma 4 12B too.
On my Nvidia 5060 Ti 16GB I can use either Gemma 4 A4B IQ3_M (larger model but highly compressed) or Gemma 4 12B Q8_0 (smaller model but much less compression) and after a night of testing with the Gemma 4 12B Q8_0 I'd say it's a better helper when it comes to coding but a worse gm for my dnd inspired fantasy world roleplaying adventure.
I imagine your "setup a selfhosted environment" is closer to the first.