[zozbot234]

I don't get it. The 'Modelfile' files are used to save and restore chat history as well, set custom system prompts and lots of other stuff that would require custom coding with most other local AI frameworks. Llama.cpp certainly doesn't offer anything like that out of the box. Those sorts of complaints seem pointless to me.

[homarp]

you tried a recent build of llama-server (from llama.cpp) ? the web interface does remember my chat, and obviously let me change all settings.