What I Built
Some memories are too precious to lose. A favourite story, a meaningful conversation, a familiar saying, or a moment spent with someone special can stay with us long after the moment has passed.
I built Memory Lantern, a local-first AI remembrance writing application that helps people transform personal memories into meaningful letters, poems, and captions.
The idea is simple: instead of treating AI as just another content generator, I wanted to create something more personal—a tool that helps people preserve the words and memories that matter to them.
Memory Lantern uses an open-weight language model through Ollama, allowing AI-generated writing to run locally rather than relying on a hosted AI API.
The Inspiration: Ghost in the Machine 👻
The theme inspired me to think about the stories that remain even when a moment has passed.
Memories are a little like ghosts: they aren't physically present, but they can continue to influence how we feel, what we remember, and how we understand the people who matter to us.
I interpreted the theme by building a tool that gives people a way to turn their memories into something they can keep, revisit, and share.
Rather than creating a conventional chatbot, I wanted Memory Lantern to focus on a more human purpose: helping people put meaningful experiences into words.
Features
Memory Lantern brings together a simple interface and locally running AI.
AI remembrance writing: Transform personal memories into letters, poems, and captions.
Local AI inference: Generate writing through an open-weight model running with Ollama.
Model selection: Choose from the models supported by the application.
Sample memory: Explore the experience with an example.
Copy and download: Keep the generated writing or share it elsewhere.
Privacy-focused design: The application is designed around local AI inference instead of sending prompts to a hosted AI API.
How I Built It
I used GitHub Copilot as my coding assistant to help develop the application and bring the different components together.
The project uses:
Python for the application backend.
HTML and CSS for the interface and visual design.
JavaScript for frontend interactions.
Ollama to run an AI model locally.
Qwen 2.5 3B as the configured default language model.
GitHub to host the project's source code.
The application connects its interface to a Python backend, which communicates with the local Ollama service to generate remembrance writing.
GitHub Copilot helped me develop the project, while the open-weight model powers the AI functionality experienced by the user.
Why Does Open Innovation Matter?
For a project involving personal memories, privacy matters.
Using a locally running model means the application can generate writing without requiring every prompt to be sent to a third-party hosted AI API.
It also gives users and developers more control over the model that powers the application.
Open-weight models make experimentation possible. Developers can explore different models, adapt their applications, and build experiences without making a proprietary cloud API the foundation of every interaction.
For Memory Lantern, this approach supports three important goals:
Privacy: Keep AI generation local rather than depending on a hosted inference service.
Accessibility: Make AI-assisted remembrance writing possible on a personal computer with compatible hardware.
Control: Give developers the flexibility to experiment with different supported models.
Local inference does not automatically guarantee that every part of an application is private, so the complete data flow still matters. Memory Lantern's goal is to keep the AI experience local-first and avoid unnecessary external dependencies.
Code
Explore the project on GitHub:
https://github.com/shivalika11/hack-project****
Demo
I will share a demonstration of Memory Lantern showing how a personal memory can be transformed into a remembrance letter, poem, or caption using a locally running AI model.
The Spirit of It 🏮
Technology is often discussed in terms of productivity, automation, and speed.
But technology can also help us express things that are difficult to put into words.
Memory Lantern is my attempt to explore that side of AI: not replacing human memories or emotions, but helping people give their own memories a lasting form.
The most important part of a remembrance is not the model that generates the words. It is the person, experience, or story that inspired them.
Some lights don't need to burn forever to leave a lasting glow.
Top comments (0)