Aathvan: An Offline Remembrance Card Maker for Family Memories

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

What I Built

Aathvan ("memory" in Marathi) is an offline remembrance card maker. You enter a loved one's name, relation, favourite food, favourite song and one memory, and a local open model writes a short, gentle remembrance card. You can print it as a PDF for an ofrenda, a family album or a memorial.

Family memories are deeply personal, so nothing here touches the internet. The model runs on my laptop through Ollama.

Demo

I tested it with a made-up example (Aaji, who loved puran poli, Lata Mangeshkar songs and telling stories on the terrace). The model wrote a short English card using only the facts I gave it, with no invented details. The Marathi output from the 3B model was poor (see second screenshot), so English is the supported language for now.

Code

sarvadnyasonje / Aathvan

# Aathvan

Offline family remembrance card maker. Enter a loved one's name, favourite food, song and a memory, and a local open model (Qwen2.5-VL 3B via Ollama) writes a short remembrance card you can print as PDF.

## Setup

1. Install Ollama and run: ollama pull qwen2.5vl:3b

2. pip install -r requirements.txt

3. py -m streamlit run app.py

## Why open-source?

- Runs fully offline, so family memories never reach a company server

- Nothing is stored in a service that could shut down

- Swapping the model is a one-line change

- Limitation: Marathi output from the 3B model was poor, so English is the supported language for now

View on GitHub

How I Built It

• Python + Streamlit for the UI
• Ollama running Qwen2.5-VL 3B locally (the same model I used for my Week 2 project, so no new download)
• A prompt that says "use only the facts given, do not invent new details", so the model doesn't make up stories about someone's loved one
• The card is generated as HTML and printed to PDF from the browser
• Limitation: the 3B model wrote poor Marathi (see screenshot). A dedicated text model like qwen2.5:3b or gemma2 would likely do better, and swapping it is a one-line change in app.py

Why Does Open Innovation Matter?

• Private by design. A remembrance card holds names, favourite foods and personal stories about someone who has passed away. With a local model, none of that is sent to a company server.
• Works with the lights (and internet) off. After the one-time model download, nothing needs a connection.
• Not locked inside a service. There is no account and no subscription, so the tool can't disappear on me. The code and the model are on my own machine.
• Swappable. The Marathi output from the 3B model was poor. With an open model I can try a different one by changing one line, which I couldn't do with a closed API.

My Agent Session

I built this step by step with an AI assistant guiding me through setup, debugging and the final code. No DevRelay session saved.

The Spirit of It

Día de los Muertos is about keeping loved ones close through food, music and stories. Aathvan follows the same idea: you give it a favourite dish, a favourite song and one memory, and it helps put them into a short card that could sit beside an ofrenda or in a family album. The "ghost" here is the memory itself, and the model only helps put it into words. Because it runs locally, these memories stay with the family.