This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine
What I Built
Ofrenda Digital is an offline, private virtual sanctuary and oral history keeper built to honor the spirits of those who have crossed over, inspired by the Mexican tradition of Día de los Muertos (Day of the Dead).
In folklore, death has two stages: the day your physical body leaves this earth, and the day your name and memories are spoken for the very last time. Ofrenda Digital exists to ensure that second death never comes.
Users place offerings upon an interactive digital altar:
- A cherished photo placed within a golden memorial frame
- Glowing, flickering candlelight and traditional offerings (cempasúchil marigolds, pan de muerto, water, and salt)
- Intimate personal recollections: their loved one’s laughter, life wisdom, favorite tunes, and comfort foods
Powered by Google’s open-weight Gemma running completely on-device, the altar meditates on these memories and weaves an emotionally resonant, culturally respectful Día de los Muertos eulogy and remembrance blessing.
Code
🕯️ Ofrenda Digital — Local Memory Sanctuary
Ghost in the Machine Challenge (Hacktoberfest 2026 / Día de los Muertos)
Built with Google Gemma (Open-Weight), FastAPI, and Docker.
🌺 What I Built
Ofrenda Digital is an offline, private virtual altar (ofrenda) inspired by the traditions of Día de los Muertos (Day of the Dead). It allows families and individuals to honor loved ones, ancestors, or cherished companions who have passed on.
Users place offerings on the altar: their loved one's name, favorite meals, songs, and cherished memories. Powered by Google's open-weight Gemma model running completely on your local machine, the application weaves these intimate recollections into a poetic, heartwarming eulogy and oral history blessing.
💀 How I Built It
- Open-Source AI Core: Gemma 2 (2B) open-weight model by Google, executed locally through Ollama.
- Backend: Python with FastAPI providing the local REST endpoints.
- Frontend: Pure HTML5, CSS3, and JavaScript…
How I Built It
The entire application was designed from the ground up to operate containerized, reproducible, and 100% locally:
- Open-Source AI Core: Powered by Google's open-weight Gemma 2 (2B) model. Inference runs completely on-device via Ollama, using structured system prompts grounded in the lore, symbolism, and reverence of Día de los Muertos.
- Backend: A lightweight Python FastAPI service that manages local model orchestration, health checks, and fallback synthesis without touching external servers.
- Frontend: A responsive single-page experience written in semantic HTML5, CSS3, and JavaScript. It features authentic papel picado banners, dynamic candlelight flicker animations, and an interactive offering shelf.
- Voice Narration: An integrated audio reader (with Web Speech API for offline speech and optional ElevenLabs synthesis) that breathes vocal life into the generated eulogies.
-
Containerization: Packaged with Docker and
docker-composeso anyone can launch their own altar with a singledocker compose upcommand.
Why Does Open Innovation Matter?
When working with artificial intelligence, privacy is often framed in terms of passwords, telemetry, or trade secrets. But the most sacred data we possess is our grief, our memories of deceased grandparents, childhood pets, and departed friends.
Open innovation was the only ethical way to build this project:
- Sacred Data Privacy (Zero Cloud Leakage): You should never have to submit the memory of a lost loved one or a private family photo to a closed cloud API where it could be logged into telemetry databases, stored on third-party servers, or used to train commercial advertising algorithms. With Gemma running locally, zero bytes ever leave your machine.
- True Permanence: Proprietary cloud services shut down, pivot, deprecate endpoints, or lock features behind monthly subscriptions. An open-weight model packaged in a Docker container belongs to you. It will still run 15 years from now, offline, on a laptop in a cabin with no internet and the lights out.
- Cultural Sovereignty: Closed models often apply rigid corporate filters or sterilize cultural traditions into generic corporate prose. Open-weight models empower developers to steer, prompt, and fine-tune AI toward regional dialects, indigenous oral narratives, and community lore freely.
The Spirit of It
During initial testing, I input the memory of an elderly mentor who was known for brewing strong black coffee, listening to gentle acoustic guitar, and reminding everyone to "build things that outlive you."
Watching the digital candles flicker on the screen as Gemma wove those exact details—the smell of dark coffee beans, the strum of guitar strings, and the guiding presence of a mentor walking beside us once more—gave me genuine chills.
Technology is so often focused on chasing the future that it forgets the people who helped us get here. Ofrenda Digital proves that open-source AI can be a tool of remembrance, tenderness, and human connection.
Prize Categories
- Top Prize — Ghost in the Machine ($250)
- Best Use of Gemma ($200) — Built entirely around Google's open-weight Gemma model, running local inference on-device inside Docker to synthesize personal memories into Día de los Muertos tributes.
- Best Use of ElevenLabs ($100) — Integrated voice narration functionality giving voice to the altar's blessings.
Top comments (0)