Digital Archaeologist：以開放式 AI 重新發掘遺忘檔案

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady.

What I Built

Digital Archaeologist is a project designed to help people rediscover forgotten information hidden in their digital files.

Over time, we collect PDFs, notes, research documents, code snippets, and project ideas. Finding one specific detail months later can mean searching through countless folders and opening files one by one.

Digital Archaeologist aims to simplify this process. The goal is to let users search their documents, discover related information, and ask questions about their digital archive.

Think of it as an archaeologist for your digital life: instead of digging through the ground to uncover the past, it searches through forgotten files to uncover useful ideas.

Current status:

Demo

Code

How I Built It

I want Digital Archaeologist to be lightweight, accessible, and affordable to run, even for developers without powerful computers or a budget for paid AI services.

My planned technology stack includes:

• React + Vite: To build the user interface.
• PDF.js: To extract text from PDF documents.
• IndexedDB: To store imported files and metadata in the browser.
• MiniSearch: To provide fast, lightweight text search.
• Gemma: An open-weight AI model to summarize documents, understand their contents, and identify connections between notes.
• Google Colab or another free-tier runtime: To experiment with running AI inference without requiring a powerful local machine.

The basic search functionality is designed to work without AI. The AI layer can then add document summaries and more intelligent answers.

I will document the final implementation and any limitations of the free-tier services used.

Why Does Open Innovation Matter?

Digital archives can contain personal notes, unfinished projects, research, and sensitive information. Developers should have control over how this information is processed.

Open-weight models such as Gemma give developers more flexibility to experiment with AI, adapt their applications, and potentially run models on infrastructure they control.

For Digital Archaeologist, open AI makes it possible to explore document understanding without making a paid, closed AI API the only option.

Combining open-weight AI with browser-based storage can also help keep documents on the user's device, provided they are not sent to an external inference service.

Open models do not automatically guarantee privacy, so the application should clearly explain when information leaves the user's device.

The Lazy Payoff

The idea is simple: spend less time searching and more time using the information you already have.

Instead of opening dozens of documents to find an old project idea, a user should be able to search their archive and jump directly to the relevant source.

I plan to demonstrate this with a small collection of sample documents and measure how easily the application can retrieve a specific piece of forgotten information.

Prize Categories

Intended category: Best Use of Gemma

I intend to target this category by making Gemma an important part of the document-understanding workflow, rather than simply mentioning the model in the project.

Thanks for checking out Digital Archaeologist!