Midzimu: Keeping Family Stories Close with Local AI
This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine . What I Built Midzimu is an offline-first family oral-history archive for preserving memories, proverbs, traditional practices, recipes, and the words elders choose to share. You can write and edit sto
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.
Midzimu is an offline-first family oral-history archive for preserving memories, proverbs, traditional practices, recipes, and the words elders choose to share. You can write and edit stories, record audio or attach a recording, search the archive, and save or restore a backup.
I built it as a place for families to keep these voices and memories close, with an optional local AI helper rather than making AI a requirement for using the archive.
There isn't a hosted demo yet. You can run the app locally with Python 3:
python -m http.server 4173 --bind 127.0.0.1 Enter fullscreen mode Exit fullscreen mode Then open http://127.0.0.1:4173. For microphone recording and local storage, use a supported browser; localhost counts as a secure context.
The optional AI helper sends a transcript to an Ollama server at the address the user configures. It can draft a summary, suggest themes, and translate into a chosen language. The original transcript remains separate and unchanged. With Ollama running locally, the story text stays on the user's device; a remote Ollama address would receive it, so users should only configure an endpoint they trust.
Family stories can be deeply personal. Keeping the archive in the browser, with backups the family controls, avoids requiring an account or cloud archive. The core archive works without AI, and the optional helper can use an open-weight model through a local Ollama installation. That gives people more control over where their stories go and lets them choose or change the model instead of depending on a hosted AI API.
The trade-off is that local AI is optional setup: users need to install Ollama and a model, and configure the local service to accept requests from the app. The archive itself remains useful without that setup.
Midzimu doesn't try to imitate an ancestor or invent their voice. It offers a place to preserve the words, recordings, and traditions people choose to share, so those memories can stay with the family.
These discussions explore related local-first and browser-storage trade-offs:
Read the first discussion · Read the second discussion