Afterlight: A Private Memory Garden, Powered by Local AI 🌼
* What I Built : * A memory rarely arrives as a polished story. It's a few details: cardamom tea on rainy afternoons, a garden someone tended, a handwritten recipe, a lesson that stayed. I built Afterlight, a local-first memory garden that helps turn those fragments into a keepsake. You enter a few
Hacktoberfest Open-Source AI Challenge: Ghost in the Machine Submission 👻🌼
A memory rarely arrives as a polished story. It's a few details: cardamom tea on rainy afternoons, a garden someone tended, a handwritten recipe, a lesson that stayed. I built Afterlight, a local-first memory garden that helps turn those fragments into a keepsake.
You enter a few details and Afterlight creates a respectful tribute, details to hold onto, an original keepsake line, gentle reflection questions, and an optional remembrance idea. It is a writing aid, not a claim to contact or speak for anyone who has died.
The idea sits between Halloween's ghost stories and DÃa de los Muertos' spirit of remembrance: the past can stay present through the details people choose to preserve.
The app runs locally on my Mac. The demo should show a fictional sample memory being entered, the generated keepsake, and the download/copy options. I am not claiming a public hosted demo until I actually deploy one. Code GitHub repository: https://github.com/Kishan-shah12/After-light.git
The model prompt emphasizes a few constraints: stay grounded in the details supplied, don't invent facts, don't impersonate the person being remembered, and don't assume a specific religion or belief about an afterlife.
The local model isn't an add-on; it makes the privacy model possible. Once the model and dependencies are installed, Afterlight can generate a keepsake offline. Personal stories don't need to be sent to a third-party inference API for the core workflow, and developers can inspect the code, change the prompt, or swap in another local Ollama model.
There are trade-offs: local inference needs disk space, memory, electricity, and suitable hardware, and generated text still needs human review. Open innovation gives people control and the ability to adapt the tool—not a guarantee that the model is perfect or that the software has no cost.
The Lazy Payoff Instead of starting with a blank page, I can begin with a structured first draft and spend my time correcting it, adding missing details, and deciding what I want to preserve. I haven't run a formal time-saving study, so I'm not claiming a specific number of minutes saved.
Overall challenge entry. I am not claiming partner categories unless I add and verify a meaningful integration with a qualifying partner technology. Built for the Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine.