[N3cr0ph4g1st]

Easiest would honestly be just to use gpt api. You can literally fork a streamlit or chainlit app and run it locally and then use langchain for any further customization...ex:https://streamlit.io/gallery?category=llms

All of them are open source minus the gpt part so you can get a feel for how it works.

[pachico]

Thank you but for legal and compliance reasons data cannot leave my premises/cloud. The entire point is precisely that, to be honest.