Skip to main content

Developing an Interactive AI team.

Any ideas?
I have about 5 Gemini gems that I want to collaborate to help be more effective. I have come across one snag. The models I know of do not allow api access to a private gem.

GPT told me I could somehow host the Gem in Discord and then use n8n to connect it to Gemini. That is something I am going to try.

Currently, I am using the old scratch pad method. At the end of each conversation with a gem, it puts together the relevant information needed for the group. Then I go to the next gem and it reads the document and away we go.

It is much slower than it should be but it is better than just going back and forth myself.

Chris Mercer

I've handled this in the past by "networking" the gems through the same common knowledge (but that does require me to manually copy and paste to update the document they are all using).

However, I'm in the VERY EARLY stage of playing around with these...
https://flows.workspace.google.com/

If you have a workspace account it literally gives you the ability to bring your gems into an automation through the Googleverse. 😎

I'm no expert, but did take this screenshot so you could see some of the potential:
https://drive.google.com/file/d/1DLgoCyLvybDOrS36f8zgsuZlYo1YEMkF/view?usp=drivesdk

Albano Gonzalez

I kind of did something like this on N8N, but didn't use Gems (you can't do that), but I used AI nodes and grabbed all the content from the gems and set that up the system prompts, used the user message for the incoming query and built a main orchestrator, an ai agent who is in charge of deciding who (which AI agent) will reply to the user message.