SlothSort: The Couch-Potato Downloads Tidier Powered by Gemma 2

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

What I Built

Every developer, creator, and student knows the dreaded digital graveyard: the Downloads folder. It starts with a screenshot, grows with invoice_final_v2 (1).pdf and setup_v3_old.exe, and before long turns into an intimidating dump of 200+ files you perpetually promise to "organize next weekend."

Sloths know best: never expend energy doing manually what can be automated.

I built SlothSort — an intelligent, context-aware desktop companion that automates the tedious chore of sorting, renaming, and filing away your cluttered files while you relax.

Demo

This is a local application.

Code

lenicyl / SlothSort

🦥 SlothSort: The Zero-Effort Downloads Tidier

Built for the DEV "Slow and Steady" Hackathon (International Sloth Day)
Harnessing Google DeepMind's open-weight Gemma 2 to reward ultimate developer laziness.

---

🛋️ What is SlothSort?

Everyone promises to organize their Downloads folder "this weekend." Weeks pass, and it becomes an abyss of invoice_final_v2 (1).pdf, random .exe setup binaries, screenshots, and untracked .zip archives.

Sloths know best: never spend energy doing what can be automated.

SlothSort is an intelligent desktop organizer powered by open-weight AI. Point it at your messy Downloads folder, click one button, and go take a nap while it turns chaos into structured perfection.

---

✨ Features

• 🧠 Context-Aware Classification: Unlike naive extension sorters (which dump all .pdfs into one folder), SlothSort inspects names, modification dates, and snippet context (invoices, meeting notes, code snippets) to categorize files accurately.
• 🏷️ Smart Renaming: Normalizes messy names (download (3).pdf $\rightarrow$…

View on GitHub

How I Built It

SlothSort is built with open-source AI at its core, specifically leveraging Google DeepMind's open-weight Gemma 2 (gemma-2-9b-it).

[ Messy Downloads Folder ]
│
▼
File Inspector ──▶ Extracts names, extensions, sizes, dates & content snippets
│
▼
Open-Weight Gemma 2 (9B) ──▶ Reasons over file context & outputs structured JSON schema
│
▼
Dry-Run Preview Table ──▶ Interactive diff table for user peace of mind
│
▼
Safe Mover & Audit Manifest ──▶ Moves files & logs .sloth_history
│
[ 1-Click Rollback ] ◀──────── Reverts all changes instantly if needed

1. Smart Context Inspection (Not Just Extensions)

Naive file sorters just look at .pdf or .png and dump tax documents and camera manuals into the same bucket. SlothSort reads file metadata, modification dates, and short content previews (first lines of text/csv/invoices) to understand the actual purpose of each file.

1. Open-Weight Gemma 2 via Google AI Studio

Rather than requiring users to download 10GB of weights or run heavy local GPU servers with Ollama, SlothSort accesses Gemma 2 via the google-genai SDK using a free Google AI Studio API key.

It asks Gemma 2 to formulate a structured JSON plan with intuitive categories (Invoices & Receipts, Data & Code, Documents & Notes, Installers & Setups), normalized filenames, and a 1-sentence lazy rationale.
Gemma 2's structured output capability (application/json) guarantees strict schema compliance with zero hallucinated file paths.

1. Sloth-Safe Execution & 1-Click Undo

Every move writes a timestamped JSON manifest to .sloth_history/. If you don't like the new arrangement, hitting "Undo This Session" restores every single file back to its exact original location and removes empty category directories.

Why Does Open Innovation Matter?

Open Weights Without the Hardware Barrier: Open-weight models usually mean wrestling with quantization, VRAM limits, and battery drain. By pairing an open-weight model (Gemma 2) with a lightweight API endpoint, users get the auditable intelligence of an open model with zero installation friction on any laptop.
True Data Sovereignty & Portability: Because Gemma 2 is an open-weight model family, you are never locked into a proprietary walled garden. If someone handles sensitive tax returns or confidential code and wants 100% air-gapped privacy, they can drop the exact same Gemma 2 weights into llama.cpp or vLLM locally without changing the sorting architecture.
Transparent Auditability: When an AI modifies your filesystem, trust is paramount. Open models and inspectable system prompts ensure you can verify exactly what instructions are executed and inspect how decisions are made.
$0 Operating Cost: Leveraging Google AI Studio's free tier means SlothSort is permanently free to use—no recurring SaaS subscriptions to clean your own hard drive.

My Agent Session

The Lazy Payoff

Manually sifting through a chaotic 100-file Downloads folder usually takes 20 to 30 minutes of repetitive right-clicking, previewing, folder creation, and renaming.

With SlothSort:

Time spent: ~8 seconds.
Effort expended: 2 clicks ("Wake the Sloth"
→
→ "Execute Moves").
The Result: 10 messy downloads (an old Chrome installer, an AWS invoice draft, an Uber trip receipt, and meeting markdown notes) were filed into clean subfolders with normalized names, while flagging stale setup binaries for deletion.

Prize Categories

Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady (Primary)
Open-Weight / Open-Source AI Category