Meet the Empress of Dirt,my niece.

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass

My niece is literally addicted to her phone, that's why I wanted to build something which keeps her on phone the shortest duration possible, so I built Empress of Dirt,keeping this rule in mind- 3 mins phone, 30 mins Garden, not Internet.

It takes an image of my garden (luckily,we have one), Gemma helps it to analyze which plants are there in our garden, with select/deselect option if it guesses something wrong.And,along with that, I have simulated Arduino Uno Q (as it was not physically available with me), but it would help get us the garden conditions, all this is saved on local memory with the help of SQLite.


I planned to keep this on local device, rather than on cloud, hence local memory.Then,it generates daily quests with different difficulties, for her to complete. Now, these quests are also being generated by Gemma. So, she gets her quests and leaves her phone inside to complete it in our garden, as my plan was. After getting completed, these quests are also saved in local memory with the help of SQLite for reference in the future quests. I also integrated voice instructions, so Jessica from ElevenLabs is practically her garden mentor.

For orchestration, it is planned to be handled by Mastra, and the physical world connectivity with Arduino Uno Q. My future plans includes getting levels and gamify it a little, but keeping my core plan intact, which is : 3 mins phone, 30 minutes Garden.

Frontend was built using Next.js, React and Typescript, using Framer for interactive background. Backend was handled using Python and Fast API, deployed on Render and GitHub for source control.
live - https://empress-of-dirt-frontend.onrender.com
GitHub repo- https://github.com/S-Bhardwaj21/empress-of-dirt
Smarter quest generation so she doesn't get bored easily and personalised conversations were both my one of the primary concerns while building this and that's exactly where open source frameworks like Gemma, ElevenLabs and Mastra primarily helped me, not to mention Render for smooth deployment.
I hope you liked what I built.
And I think, I just cooked this one.