[Eisenstein]

They are using Ollama which is based on llama.cpp; florence is not supported on that backend.