I have uploaded a more basic version of my chatbot to https://huggingface.co/spaces/gorkemgoknar/moviechatbot .
If you checkout app.py code you can see how you can use a transformers model types which is not included in standart interface (do not forget requirements.txt to import modules)
I also shared my model and information to use https://huggingface.co/gorkemgoknar/gpt2chatbotenglish
While building gradio app I noticed that save_state(history) get_state() does not work in embedded application. While gradio app is full screen it works but in HF spaces information is lost (likely due to Flask session not being statefull inside embedded fram).
To overcome this I use a global history but likely this would make models history being shared between different users (not a problem for a demo and it may be fun), yet for information there should be a way to save state if HF Spaces would be a different product.