Call a Gemini GEM from API [via n8n]

All on linux.
All apply here: Docker, python venv, and systemd (web server).

The idea was to clean up some LLM API calls via n8n, but it works from curl, etc. There is a bit to it, as we need to run a n interim web server to catch n8n calls.

I really don't like to have the advantages of an API shit on because you need to send a large context/system or user message with every call. Don't know if it adds efficiency, but it looks sloppy.

Not all LLMs will likely bend to my will, but I will try...