Hardware Requirements?

[SilverWulf212]

I'd like to begin by saying thanks; I've installed this over the weekend on a local core i5+ 10th gen/24 gb ram/ 1 TB m2. I'm loving the google docs retrieval and rag, as well as Q and A with contextual search.

So, some quick questions if I could:

1. What hardware would you recommend? I tossed it on a spare box, honestly, not expecting it to work (I've been disappointed many times before). I was amazed when I discovered it.... did what it said it would do: )! But, I've noticed quality is, of course, dependent on model and information. I've noticed it will retrieve correct answers more often with GPT5 than, say, with gpt 40 mini. Also fewer hallucinations. I was surprised as we "only" have 1200 google docs that are about 2-3 pages each if that. I've also found deep research queries will time out on several models, but I could be maxing out the ram on it - didn't check my bottlenecks yet. (back to - what hardware would you recommend : )?

2. Like the others , I'm trying to reduce hallucinations. I have custom directions set to return "I don't know" answers if info isn't found, but, especially with the less expensive models, it'll create or farm disparate info. Now, most of our Google docs aren't well structured, but neither are the docs of most people I had in mind for this product as a life improvement item and into to personal, and if they want it, private AI.

3. Any tips or best practices? I'm about to attempt to try to figure out how to get Onyx to read from Repairshoppr's tickets API and see what I can do there to really leverage this system for context and bridging info.

Thanks again for an incredible OSS product; I'm really enjoying it so far, and I believe it will bring a lot of value to our team. I had attempted to build something like this but was so far off from this level of awesome it's laughable. Thanks again.