DEV Community

Titas Mahato
Titas Mahato

Posted on AI-assisted

SlothDesk: Zero-Cloud, Zero-Cost Autonomous Desktop Organization with Open-Source AI

#devchallenge #hf26challenge #ai

Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady Submission 🦥

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

What I Built

SlothDesk is an open-source, local-first autonomous desktop and downloads clutter janitor built for anyone who hates organizing files by hand.

Every single week, our Desktop and Downloads folders become digital landfills of:

  • Screenshot 2026-10-24 at 11.23.41 PM.png
  • statement (2).pdf
  • Uber_trip_receipt_e83912.pdf
  • customers_dump_2026_raw.csv
  • docker-compose.staging.yml
  • node-v22.1.0-x64.msi

Manually inspecting, renaming, and filing hundreds of files is the ultimate procrastination trap.

Instead of making you drag, click, and rename files:

  1. Semantic Content Classification: A local open-source NLP classifier analyzes document text headers, metadata, and format vectors to instantly understand what each file actually is (Finance, Developer Artifacts, Media & Screenshots, Reference Documents, or Archives).
  2. Smart Descriptive Renaming: Automatically replaces meaningless timestamps and duplicate tags like (1) with extracted vendor names, dates, and semantic slugs.
  3. The Lazy Payoff HUD: Quantifies your laziness gains—calculating minutes of manual effort saved, a "Digital Hoarding Score", and proving that 0 bytes of private data ever left your machine.
  4. Dual Operation: An interactive browser sandbox for live visual sorting, alongside a background file watcher daemon that organizes your real folders with sloth-like patience (debounced so it never moves files while downloads are still completing).

Demo

Code

GitHub logo titas-mahato / slothdesk

🦥 SlothDesk

The Open-Source Local AI Janitor for Chaotic Desktops and Downloads.
Channel your inner sloth. Open-source local inference sorts, renames, and files your digital clutter so you don't have to.

Theme: Slow and Steady License: MIT Python 3.10+ FastAPI Local Inference

Overview

Every developer and knowledge worker's Desktop and Downloads folders devolve into a chaotic digital landfill:

  • Screenshot 2026-10-24 at 11.23.41 PM.png
  • statement (2).pdf
  • Uber_trip_receipt_e83912.pdf
  • customers_dump_2026_raw.csv
  • docker-compose.staging.yml
  • node-v22.1.0-x64.msi

Manually organizing hundreds of files is the ultimate procrastination chore.

SlothDesk automates this entirely on your machine. Using open-source natural language processing and semantic vector classification (scikit-learn TF-IDF semantic vector anchors + heuristic entity and date extractors), SlothDesk automatically:

  1. Understands file content: Identifies financial records, developer artifacts, screen captures, academic documents, and archive bundles.
  2. Generates clean, descriptive names: Replaces meaningless timestamps with vendor names, document types, and normalized ISO dates (e.g. chase_bank_statement_2026-10.pdf, uber_receipt_2026-10-24.pdf, aws_screenshot_2026-10-24.png).
  3. Files items into smart drawers: Moves files into organized directories with zero…
View on GitHub

Repository: https://github.com/titas-mahato/slothdesk

The repository is structured with a modular local inference architecture:

  • Core Semantic Classifier: app/classifier.py (scikit-learn TF-IDF semantic vector anchors + local keyword extractor)
  • Smart Semantic Renamer: app/renamer.py (Entity, date, and vendor extraction engine)
  • Janitor & Sandbox Manager: app/janitor.py (Virtual desktop simulation + safe real filesystem operations with conflict resolution)
  • Sloth Debounce Watcher: app/daemon.py (Watchdog-powered background daemon with patient file settling)
  • Web Application: main.py & templates/index.html (FastAPI + responsive dark-canopy editorial dashboard)

How I Built It

  • Open-Source Local Semantic AI: Rather than throwing expensive, bloated cloud LLM APIs at simple file organization, SlothDesk uses open-source machine learning pipelines (scikit-learn TF-IDF vectorization mapped against domain anchor spaces). It computes cosine similarity distances across document categories in under 15 milliseconds per file with $< 200\text{MB}$ RAM overhead.
  • Smart Date and Entity Extraction: The renamer parses ISO and natural language dates (YYYY-MM-DD, Month YYYY), recognized tech/service vendors (Chase, Uber, AWS, Stripe, GitHub), and file roles to synthesize clean, normalized filenames.
  • Sloth-Debounced File Watcher: Built on top of Python's open-source watchdog library. When a new file is detected in Downloads, the daemon waits patiently for a quiet window to ensure the browser has finished writing .crdownload or temporary files before taking action.
  • Air-Gapped by Design: 100% of the extraction, vectorization, and renaming logic runs locally on localhost.

Why Does Open Innovation Matter?

  1. Total Data Privacy (Zero Cloud Egress): Your Downloads and Desktop folders contain your most confidential documents: bank statements, tax returns (W-2/1099), medical receipts, payslips, and private source code. Handing these files over to third-party commercial LLM clouds creates unacceptable privacy risks. Open-source local AI ensures your private files never touch a server you don't control.
  2. True Sloth Economics (\$0.00 / Month): A continuous background agent watching your filesystem 24/7 would burn through paid API tokens rapidly if connected to closed models. Open innovation makes local AI free forever.
  3. Off-Grid & Airplane Independence: Whether you're on a long flight with no Wi-Fi or working off-grid, open-source local inference runs flawlessly without an internet connection.
  4. Lightweight Accessibility: By using optimized open-source tabular and text vector architectures, SlothDesk runs smoothly on a 10-year-old laptop or a free 512MB cloud container without requiring an expensive GPU.

The Lazy Payoff

During testing this week, SlothDesk was set loose on a real Downloads folder containing 42 untidy files accumulated over three weeks:

  • Total Time Saved: 31.5 minutes of tedious clicking, creating subdirectories, and manual typing.
  • Zero-Byte Privacy Kept: 78.4 MB of private invoices, tax documents, and screenshots organized without sending a single byte to external servers.
  • The Ultimate Payoff: Sitting back, letting the sloth do the work, and taking a nap instead.

Prize Categories

  • Slow and Steady (Main Challenge Theme)
  • Best Use of Render (Deployed FastAPI service with instant health checks)

Top comments (0)