In case anyone wants to see the Gemini API function calling in action, as well as some interesting things like dynamic context windows using my ibgib protocol, you can check it out at https://ibgib.com (EDIT: It’s BYOK, so you’ll need a Gemini API key. You can find this at aistudio.google.com or in your GCP credentials).
I have made a couple videos on YT on how to use it, which you find at https://www.youtube.com/@ibgib . The most recent video was on “minigames”
, which is currently like agent-created dynamic flashcard sets. Right now, they are text only, but in the very near future I’ll have agent-generated images to be used as the stimulus for the flashcard. This is for learning/reinforcing foreign languages in a more fun, dynamic way, which maybe most importantly, raises the throughput capabilities compared to existing basic flashcard approaches (even with spaced repetition and other tactics).
Because my ibgib protocol will ultimately subsume version control, I am avoiding git + GitHub right now, but you can find the source in the @ibgib/blank-gib package on npm.