Using standalone llama.cpp (not embedded)

[hopperath-dot]

With the latest update, I cannot seem to update the config to use my local llama.cpp server (built with proper rocm params). If I just change the port under llama c/c++ (offline) and do not start the internal server (which uses slow cpu mode) proxyai appears to default to ollama settings, trying to connect on localhost:11434. I had this working fine on the earlier version of the plugin. I may be missing something simple, but so far I have been able to figure it out. Any suggestions appreciated!