when using Google's gemini models I don't get feedback about the context use like I do with the antropic model

**Describe the bug**

When using ECA for Emacs with the anthropic models I get a breakdown of the token usage but I can't see this with Gemini. I don't know if this is a limitation of the Emacs frontend or the ECA server itself.

**To Reproduce**

M-x eca-chat, do things, note no token counts in UI

**Expected behavior**

You should see the token usage

