Ofrenda: a remembrance that never leaves your machine

K.S. Karan Reddy — Fri, 30 Oct 2026 13:09:21 +0000

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

What I Built

Ofrenda is a small web page that helps you write a remembrance for someone who has died, then sets it on a little altar. You give it a name, who they were to you, the foods and songs they loved, and one memory in your own words. A language model running on your computer turns that into a short piece of writing in your voice, and the altar lights a candle while it writes.

I read "Ghost in the Machine" as the people we carry around, not the monsters. The version I wanted was the one where the machine helps you say something about a person who is gone, and then gets out of the way.

The rule that shaped everything: the model may only use what you typed. It is told never to invent a name, a place, a date or a quote. If you give it three details, you get three details back. A remembrance with a made-up fact in it is worse than none.

Demo

Code

kskaranreddy / Ofrenda

Write a remembrance for someone you miss using a local Gemma model via Ollama. Runs offline in one HTML file; nothing leaves your computer. Hacktoberfest 2026: Ghost in the Machine.

Ofrenda

Write a remembrance for someone you miss, with a model that runs on your own computer.

You give it a name, who they were to you, the foods and songs they loved, and one memory. A local Gemma model, served by Ollama, writes a short piece in your voice using only those details. A small altar lights a candle while it writes. Nothing is uploaded: the page talks to localhost and nowhere else, so it works with the internet off.

Built for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine.

Run it

You need Ollama and Python (only to serve the page locally).

Pull a model:
```
ollama pull gemma3
```
On a machine with 8 GB of RAM or less, use gemma3:1b.
Start Ollama and allow the page to call it.

Windows (PowerShell):
```
$env:OLLAMA_ORIGINS="*"; ollama serve
```
Mac / Linux:
```
OLLAMA_ORIGINS="*" ollama serve
```
If the Ollama…

View on GitHub

The whole app is one index.html. No build step, no dependencies, no backend.

How I Built It

Model: Gemma (gemma3), served by Ollama on localhost:11434. Any model you have pulled shows up in the dropdown, and the page prefers a Gemma if it finds one.
Page: plain HTML, CSS and JavaScript. It calls Ollama's /api/tags to list your models and /api/chat with streaming on, and reads the newline-delimited JSON as it arrives so the text appears word by word.
Altar: three CSS tiers and a paper-cut banner. The candle's flame fades in when generation starts, and the animation turns off if your system asks for reduced motion.
Hardware: my laptop has no usable GPU for Ollama, so everything ran on CPU. A short remembrance takes about a minute, which suits the subject.

To run it:

ollama pull gemma3
$env:OLLAMA_ORIGINS="*"; ollama serve    # PowerShell
python -m http.server 8000 --bind 127.0.0.1

Why Does Open Innovation Matter?

The input to this tool is grief: a person's name, a story about their family, the song that was playing. I did not want that text to pass through anyone else's server, be logged, or sit in a database that outlives a startup's runway.

With an open-weight model the privacy is structural rather than promised. The page cannot send your memory anywhere because there is no remote to send it to. A closed API can offer a privacy policy. It cannot offer that.

It also means the tool keeps working. An ofrenda is something you come back to every year. A hosted model gets deprecated, repriced or shut down. A model file on your own disk keeps working, and when a better one comes out you swap it in with a dropdown.

The Spirit of It

Ofrenda is built around a simple belief: people never truly leave us as long as their stories live on.

It's about preserving the little things — their words, traditions, favorite memories, and the moments that made them who they were.

Not recreating those we've lost, but creating a space to remember them, celebrate their lives, and keep their stories alive.

Because sometimes, remembering is the closest thing we have to holding on.

Prize Categories

Best Use of Gemma

DEV Community: K.S. Karan Reddy