github.com/ggml-org/llama.cpp/…

Then find the right model depending on your RAM, these are guf files but this Phi 3 is hard to get as you need to log in in many ways trough git or on site or wget and curl return a 404. Don't ask me how I worked it around but then you have something like this after you build your script (Also all is in git how to's) make sure it listens to local host and taraa! then you have this . This model works with <[ ..]> brackets as system message

OK call it what it is no API and local network and easier then searching endlessly the web, there were AI is omnipresent and sold to Big Corps (your data)

So I integrated Vim (no extension but with bash) to see outputs code and added a Hallucination checker to it

then shift = v

then F5 to verify

Now tell me is this any slop?

edit: Keep it clean, keep it local, keep it yours. 🔒
Set up is 100% legal, both the LLama.cpp as the model are MIT license, no data leaves your machine and no terms of services violated.
I'll add does not scrape the web, Phi-3 Mini is a static, offline model trained by MicroSoft mid 2023 if I'm correct and It cannot browse, fetch, or see anything online