Veladora

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine

What I Built

A private ofrenda for Día de los Muertos. You place a name, a photo, a food, a song, and one memory on a table. Gemma 3 270M, an open-weight model, writes the inscription in your browser. Any sentence that is not grounded in what you placed is struck through and labeled Not on the table.

Photos, words, and the finished inscription stay in this browser. The weights download once from Hugging Face. After that, lighting another candle does not need a network connection, and nothing you type is sent to a model host.

Demo

https://djrcx.github.io/Velaroda/

Code

https://github.com/DJRCX/Velaroda

How I Built It

At first, I used Figma Make to fine out the design, then I used Cursor to build the project. It used Transformers.js, Gemma 3 270M, local grounding that strikes any sentence the notes do not support.

Why Does Open Innovation Matter?

the remembrance never goes to a server, and the candle still works when the network is off after the one-time download.

Prize Categories

• Best Use of Gemma.