# The Zero-Effort Local File Janitor > Source: > Published: 2026-10-20 21:42:01+00:00 *This is a submission for the [Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady](https://dev.to/challenges/hacktoberfest-week3-2026-10-19)* What I Built: Channeling the Inner Sloth Nothing drains human energy faster than staring at a Downloads folder filled with IMG_4829.png, scan_final_v2 (1).pdf, and invoice_88392.pdf. Cleaning it requires opening each file, figuring out what it is, thinking of a sensible name, and dragging it into the right folder. In honor of International Sloth Day, I built SlothSort—a background daemon that rewards peak laziness. You drop files into your messy folder and walk away. Running at a relaxed, low-priority pace in the background, SlothSort uses local open-weight models to inspect images and read documents, renames them into clean YYYY-MM-DD_Category_Description.ext formats, and moves them into organized folders (Finance, Work, Receipts, Screenshots, Personal). The Open-Source Core Inference Engine: Ollama (running locally on port 11434). Open-Weight Vision & Text Model: Llama 3.2 Vision (11B) or Qwen 2.5 VL (7B) for reading screenshots, receipts, and image clutter, paired with Llama 3.2 (3B) for lightning-fast text/PDF classification. Structured Output & Agent Harness: Python pydantic schemas enforcing strict JSON output directly from Ollama's local structured output API, alongside watchdog and pypdf. Why Open Innovation Matters Here (Open vs. Closed) If you tried to build an automated file organizer using a closed frontier API, you would immediately hit three walls that make the project impractical: Absolute Data Privacy (Zero Cloud Leaks): Your Downloads and Desktop folders contain tax returns, medical receipts, bank statements, private screenshots, and signed contracts. Piping every file that touches your hard drive to a third-party server is a massive privacy risk. Because SlothSort runs open-weight models entirely on local silicon, your bank statement never leaves localhost. The "Set It and Forget It" Economics ($0 Cost): A background daemon that processes dozens of PDFs and high-res screenshots every week racks up continuous token costs on a paid API. Worse, if a loop bugs out, you wake up to a surprise bill. Running on local open weights means marginal cost is zero—you set it up once and let it run forever. "Slow and Steady" Wins on Local Hardware: Closed APIs optimize for instant, high-throughput cloud responses. But a background file sorter doesn't need to finish in 200 milliseconds. Just like a sloth, if your laptop takes 6 seconds of idle GPU/NPU time to read a PDF and move it while you are making coffee, the result is identical to the user. Open-weight 3B–8B models fit comfortably in consumer RAM and thrive in asynchronous, low-urgency background tasks. No Internet Required: Whether you're sorting receipts on a flight or working from a cabin with no Wi-Fi, the agent never breaks.