cd /news/artificial-intelligence/ofrenda-exe-a-candle-for-the-ghost-i… · home › topics › artificial-intelligence › article
[ARTICLE · art-161198] src=dev.to ↗ pub= topic=artificial-intelligence verified=true sentiment=↑ positive

Ofrenda.exe: A Candle for the Ghost in the Machine

A developer built Ofrenda.exe, a single-file Python command-line tool that turns a name and a single memory into a short remembrance card, using the open-weight Gemma model run locally through Ollama. The tool keeps memories on the user's machine, requires no account or internet connection, and falls back to a deterministic offline mode when the local Ollama server is unavailable.

by read2 min views1 publishedOct 30, 2026

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

Ofrenda.exe is a small local AI memory keeper inspired by Día de los Muertos.

You give it a name and one memory. It creates a short remembrance card that feels like a digital candle placed on an ofrenda.

I interpreted the theme as a meeting point between old memories and new technology. The ghost is not a person trapped inside a machine. It is the feeling that a story can disappear unless we choose to keep it alive.

The project runs from the terminal and keeps personal memories on the user's machine.

Install Ollama and download the small Gemma model:

ollama pull gemma3:1b
ollama serve

Then run:

python3 ofrenda.py "Abuela Rosa" "She taught me to make sweet bread on rainy Sundays."

Example output:

🕯️ OFRENDA FOR ABUELA ROSA

A short remembrance card appears here, written by the local model.

The project also has a deterministic offline mode:

python3 ofrenda.py "My first open source project" "It taught me that unfinished things can still help someone." --offline

The complete source code is available here:

Ofrenda.exe is a single file Python CLI with no third party Python dependencies.

It calls the local Ollama HTTP API with Python's standard library. The prompt asks the model to write a short remembrance without inventing facts or mentioning the prompt.

The project uses the open weight Gemma model locally through Ollama. If the local server is unavailable, the program falls back to a deterministic candlelight mode so the basic experience remains usable.

The main flow is:

Memories are private by default.

A closed hosted model could require up a person's name, family story, or personal history to a service we do not control. Ofrenda.exe uses local inference instead, so the story can stay on the user's laptop and work without an account or an internet connection.

Open models also make the project replaceable. Users can choose a smaller model for an older laptop, a multilingual model for their family, or a fine tuned model with a different cultural voice.

The application does not depend on one provider's pricing, availability, or personality. A remembrance tool should be able to outlive the service that created it.

I wanted the result to feel gentle instead of frightening.

A terminal window is usually associated with commands, logs, and errors. Ofrenda.exe gives it a quieter job. It lets a small memory sit in the light for a moment.

Technology can help us remember without taking the memory away from us.

── more in #artificial-intelligence 4 stories · sorted by recency
── more on @ollama 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/ofrenda-exe-a-candle…] indexed:0 read:2min 2026-10-30 · —