GhostMemory AI — Memories That Refuse to Disappear

Hacktoberfest Open-Source AI Challenge: Ghost in the Machine Submission 👻🌼

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

I built GhostMemory AI, a browser-based AI storytelling application that gives forgotten memories a second life.

Users can write a personal memory, story, person, place, or moment and choose how they want the AI to transform it:

I interpreted the "Ghost in the Machine" theme as the idea of memories becoming echoes inside a machine.

The "ghost" doesn't have to be supernatural. It can be a person, a place, a story, or a moment from the past that we don't want to disappear.

GhostMemory AI takes those memories and turns them into something that can be read and remembered again.

🌐 Live Website: https://ayush-kasaudhan-01.github.io/-GhostMemory-AI/

💻 GitHub Repository: https://github.com/ayush-kasaudhan-01/-GhostMemory-AI

Note: The file:///C:/... URL only works on my own computer, so I used GitHub Pages for the public demo.

The complete source code is available on GitHub:

https://github.com/ayush-kasaudhan-01/-GhostMemory-AI

The core of GhostMemory AI is an open-weight language model running in the browser.

I used SmolLM2-135M-ONNX with Transformers.js for text generation.

User Memory ↓ GhostMemory AI Interface ↓ Prompt Builder ↓ Transformers.js ↓ SmolLM2 Open-Weight Model ↓ Generated Memory ↓ Local Browser Storage Enter fullscreen mode Exit fullscreen mode There is no commercial AI API involved.

The application sends the user's memory to the locally running model in the browser, which generates the response.

Generated memories are stored using browser localStorage, so the project doesn't require a database or backend server.

For a project dealing with personal memories, privacy matters.

A closed AI API would mean sending someone's memories to an external service.

With an open-weight model running locally in the browser, the project can keep the AI interaction much closer to the user's own device.

Open innovation also gives the project more freedom.

This was one of the biggest reasons I wanted the AI itself to be an open component rather than simply connecting a website to a closed API.

I did not use an external agent session for this project.

The project was built around a local/browser-based open-weight AI workflow instead.

The idea behind GhostMemory AI is simple:

What if a machine could help us keep the things we are afraid to forget?

A small memory can disappear from a conversation, a notebook can be lost, and a story can stop being told.

But when that memory is transformed into a story, it becomes another way of preserving it.

That is where the "ghost" in GhostMemory AI comes from.

The machine isn't bringing back a supernatural ghost.

It is bringing back an echo of something that already happened.

I'm entering this project for the categories relevant to:

GhostMemory AI started as a simple idea: combine the atmosphere of Halloween with the human need to remember.

Building it with open AI made the idea even more interesting because the "machine" doesn't have to be a distant server.

GhostMemory AI gives them a place to stay. 👻