Keepsake: Oral-History Archive That Keeps Stories Close

Keepsake: An Oral-History Archive That Keeps Stories Close

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

What I Built

I built Keepsake, an offline-first notebook for preserving oral histories: the stories, voices, and small details that can otherwise fade with time.

Code

https://github.com/poonamkumari24/Hacktoberfest.git

How I Built It

Keepsake is built with plain HTML, CSS, and JavaScript, browser storage, and a small Node.js server. It needs no hosted database or user account. Stories and recordings can be exported and imported as portable archives.

For AI assistance, Keepsake uses Ollama with a model installed on the user’s computer. The setup example uses Qwen 2.5 1.5B. To try it:

1. Install Ollama and download the example model.

2. Run node server.js.

3. Open http://127.0.0.1:4173.

Model inference could not be exercised in my environment because Ollama was not running. Audio transcription is not automated; it remains a manual step.

Why Does Open Innovation Matter?

These are personal memories, so keeping them private matters. Using Ollama and an open-weight model lets Keepsake provide AI assistance without sending stories to a hosted AI service. Story text is sent to Ollama on the same computer only when the user requests AI help, and audio stays in the browser.

After the initial setup, the app can be used without an internet connection. Users can also choose among the models they have installed in Ollama. That local control makes AI assistance more accessible while keeping the archive portable through import and export.