TidyDesk AI：讓開源 AI 整理雜亂資料夾

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

What I Built

Meet TidyDesk AI — a privacy-first desktop file organizer designed to help students, developers, freelancers, and anyone with a cluttered computer spend less time managing files.

We've all been there: hundreds of files in Downloads, confusing filenames, duplicate documents, and screenshots we promise to organize later.

TidyDesk AI aims to make that cleanup effortless.

Instead of manually opening files, renaming documents, and dragging everything into folders, users can select a directory and let AI suggest a clean organization plan.

What it does

• Smart Categorization: Suggests categories for documents, images, invoices, academic notes, and other files.
• Intelligent Renaming: Recommends meaningful filenames based on available file information.
• Folder Organization: Suggests appropriate destination folders.
• Duplicate Detection: Identifies potential duplicates for review.
• Safe File Operations: Lets users review and approve proposed changes before applying them.
• Undo Support: Keeps a record of supported file operations so users can reverse changes safely.

The goal is simple: spend less time organizing files and more time doing something useful.

Demo

Try the live application: https://tidydesk-ai.onrender.com

TidyDesk AI is designed as a desktop application. The demo will show a cluttered test folder being analyzed, organized, and reviewed through a safe approval workflow.

Code

GitHub Repository: https://github.com/Charlie-ai786/tidydesk-ai.git

The project is being developed as a standalone application, with a focus on local file processing, safe filesystem operations, and open-weight AI.

How I Built It

TidyDesk AI combines native desktop capabilities with local AI inference.

Technology Stack

• React + Vite + TypeScript: User interface.
• Tauri 2: Lightweight desktop application framework.
• Rust: Native filesystem operations and application commands.
• Ollama: Local inference runtime.
• Qwen or Gemma: Candidate open-weight language models, depending on the final tested configuration.
• SQLite: Optional persistence for operation history and settings.

How it works

1. The user selects a folder using the native directory picker.
2. The application scans file metadata and supported document content.
3. Local AI suggests useful categories, filenames, and destinations.
4. The user reviews and edits the suggestions.
5. Approved operations are applied to the selected files.
6. The application records successful changes and supports undo where safe.

The model proposes changes, but it does not control filesystem operations directly. The application validates paths, checks for filename conflicts, and requires user approval before modifying files.

Why Does Open Innovation Matter?

Open innovation makes it possible to build a file organizer that prioritizes privacy, control, and independence from paid AI APIs.

1. Your files stay yours

File collections can contain personal documents, financial records, academic work, and private screenshots. Local processing reduces the need to send this information to external AI services.

2. No per-request cloud AI charges

When a compatible open-weight model runs locally, inference does not consume hosted LLM API credits. The user still needs suitable hardware, and the application may have other costs, but its core AI workflow need not depend on a paid API subscription.

3. Freedom to choose your model

A modular architecture makes it possible to experiment with compatible models and choose one that balances quality, speed, memory usage, and licensing requirements.

4. Offline capability

After installation and model downloads, supported file-organization workflows can run without an internet connection.

5. AI with guardrails

Open models provide flexibility, but they can still make mistakes. That's why TidyDesk AI separates model suggestions from file operations. Users review the plan before changes are applied.

For me, open innovation is not just about making AI cheaper. It's about building useful software that users can control and trust.

My Agent Session

DevRelay Session:

I'll include the recorded development session once it has been successfully submitted through DevRelay.

The Lazy Payoff

The goal is to turn a tedious manual cleanup session into a quick review-and-approve workflow.

Instead of deciding where every file belongs, users can review AI-generated suggestions in batches and apply the changes they approve.

What I still want to measure: How long it takes to organize a sample folder manually compared with using TidyDesk AI.

I'll report actual timings after testing the application on a representative folder rather than inventing a time-saving figure.

Prize Categories

Potential category: Best Use of Gemma

If Gemma is integrated and tested in the final implementation, I'll document its role in file categorization and filename suggestions and confirm that the project satisfies the official category requirements.

I'll list only the partner categories for which the finished project is genuinely eligible.

Thanks for checking out TidyDesk AI!

The idea behind this project is simple: let open-source AI handle the repetitive work, keep users in control, and make organizing files something you no longer have to put off.