I won’t type it in completely, but I have written a rather complete guide to getting JAN, cline, vscodium, and the qwen a3b models working as fast as I could (50-70 tokens per second). Here is the URL if anyone is interested in the guide. It does assume a bit of knowledge, and provides mostly the items I needed to get local ai working.
This also includes a script that gets proper RAG searching working on JAN as well.
Here is the guide link