I made a mistake
I was so focused on using the right tech stack from the start that I forgot the most essential part of the hackathon: making it work
Gemini has a feature called “Function Calling” that’s supposed to make it easier to get structured data outputs from the model
But it didn’t make it easy for me. This feature is not yet well documented, and I only found this and this as the only examples. The other example is not even written in Swift, so I was not sure if it’s applicable. The Gemini in Firebase also wasn’t that much of a help
NOTE: “Gemini in Firebase” is another Google product that lets you chat with Gemini about Firebase products. This is different from other Gemini features that you can integrate into your Firebase app. Confusing, right? I know
It took some time to adapt the examples to my use case, but I managed to get the flow working by combining elements from both examples. However, it’s requiring me to tweak the prompt further to make it fully functional
This isn’t a path worth pursuing for now, so I have to drop “Function Calling” and just use basic Gemini features, or consider working with other models
I guess this is the price you have to pay for working with new tech – not much documentation (yet)
Tomorrow, I have to get something out regardless of what it looks like, inside and out 😆
Oh! and by the way, as a result of my procrastination, I managed to speed up my WordPress site’s loading speed. Did you notice it? 😅
Also added the v1.0 feature coverage to the Y-Verse pre-order page
And special mention to the nationwide internet outage that happened for a few hours in the morning from Converge 😜