Echoes: A Private, Local AI Memory Lantern

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

What I Built

Echoes is a local-first writing companion inspired by the stories and memories we carry with us. Share a small memory and choose a form: a remembrance, a gentle ghost story, or a short literary scene. Echoes uses a local model to shape a draft while asking it to stay grounded in the details provided.

Demo

The project currently runs locally. Install the dependencies and follow the Ollama setup in the README. There is no public demo yet.

Code

View the source code on GitHub.

How I Built It

The interface uses Streamlit and sends the chosen style and memory to an open-weight qwen2.5:3b model running through Ollama on the same computer. Echoes does not save the input to an app database. The prompt asks the model not to invent biographical facts or present a fictional story as true, and the user can review, edit, or discard the generated draft.

Why Does Open Innovation Matter?

Family memories are personal. Local inference means the words stay on the writer's computer instead of being sent to a hosted AI API. After the initial model download, generation works offline and has no per-story API fee. The prompt and model are inspectable and replaceable, so the writing behavior can be adapted rather than locked inside a service.

My Agent Session

I do not have a shareable DevRelay session link for this project.

The Spirit of It

Echoes is a small invitation to write down the details that might otherwise fade: a phrase, a song, or a familiar light in the window. It does not claim to speak for anyone; it gives the writer a private place to shape a story in their own words.