ECHOES — A Private, Open-Source AI Memory Keeper
Some stories deserve to outlive us.
This is my submission for Hacktoberfest Open-Source AI Challenge — Week 4: Ghost in the Machine.
What I Built
We create memories every day.
A family recipe passed down through generations. A story our grandparents tell us. An old photograph with a story behind it. A handwritten letter, a voice recording, or a small detail that might otherwise disappear with time.
But what happens when those memories get lost?
That's the idea behind ECHOES — a private, open-source AI memory keeper designed to help people preserve, organize, and rediscover the stories that matter.
ECHOES allows users to build a personal archive of memories, including stories, photographs, letters, recipes, and family histories. They can then ask questions and retrieve information from their archive using AI.
I interpreted the theme Ghost in the Machine as something more human than a ghost or a fictional AI personality.
What if the machine could preserve the echoes of the people and experiences that shaped us?
ECHOES isn't designed to impersonate people or pretend to be someone who is no longer here. Its purpose is to preserve their stories and help us revisit what was actually recorded.
Demo
GitHub Repository:
https://github.com/kmishra006/Echoes-Archive-
Live Demo: https://echoesarchive.netlify.app/
Code
The project is open source:
The goal is to make the project understandable, extendable, and accessible to developers interested in building privacy-first AI applications.
How I Built It
ECHOES is built around a local-first architecture, with the goal of keeping personal memories under the user's control.
The technology stack includes:
- Frontend: React, Vite, and TypeScript.
- Backend: Python and FastAPI.
- Database: SQLite.
- AI inference: Ollama with an open-weight language model.
- Memory storage: Local file storage.
- Retrieval: Searching stored memories and using relevant records as context for AI-generated answers.
The central idea is that the AI should answer from the user's actual archive rather than inventing personal history.
One important design principle is evidence-based answering:
- FOUND: Relevant supporting memories were found.
- PARTIAL: Some supporting information exists, but the answer is incomplete.
- NOT_FOUND: The archive does not contain enough information to answer.
- CONFLICT: Stored memories contain conflicting information.
Answers should be grounded in retrieved records, with references to their sources wherever possible.
If ECHOES cannot find something, it should say so.
I believe an AI memory system is only useful if people can trust what it tells them.
Why Does Open Innovation Matter?
For a project like ECHOES, open innovation isn't just a technical preference. It's part of the product's philosophy.
Personal memories are sensitive. Family histories, photographs, letters, and private stories deserve more control than simply sending everything to an external service.
Using open-weight models and local inference makes it possible to explore an architecture where AI processing can happen locally, without depending on a closed, cloud-only AI API.
It also gives developers the freedom to:
- Experiment with different open-weight models.
- Inspect and modify the underlying architecture.
- Build without being locked into a single AI provider.
- Adapt the project for different communities and use cases.
- Contribute improvements to an open-source project.
There is still work to do around deployment, storage, and privacy guarantees, but the direction is clear: personal memories should not have to become someone else's data to become useful.
The Spirit of It
Imagine finding an old family recipe years later.
You don't remember who first made it. You don't remember the small story behind it. But somewhere in your archive, there's a memory explaining exactly why that recipe mattered to your family.
You ask ECHOES.
Instead of inventing a sentimental story, it finds the original record and brings it back to you.
That's the experience I wanted to build.
Not an AI that pretends to bring people back, but one that helps keep their stories from disappearing.
Some stories deserve to outlive us.
TEAM MATES
Built with curiosity, open-source technology, and the belief that the stories we preserve today can mean something to someone tomorrow.
Thanks for checking out ECHOES!
Top comments (0)