What I Built
Sorting cluttered directories—especially messy Downloads folders filled with random PDFs, unnamed screenshots, and stray documents—is a chore everyone puts off indefinitely. SlothSort is a lightweight background automation concept designed around the "Slow and Steady" mindset. It passively listens for newly added files, extracts minimal text or metadata, and lets a local open-weight model figure out the proper category and readable name. The system automatically routes items into organized destination folders like invoices, study material, and receipts, letting users be completely lazy while their storage cleans itself up without any manual effort.
Demo
The project runs as a background service with an interactive dashboard that logs real-time file categorization. Whenever a new file is dropped into an unorganized folder, the interface highlights the detected item, shows the extracted summary, and visually updates the destination path as the file is sorted automatically.
Code
The core implementation consists of an automation script built around a file watcher loop and structured prompting. It intercepts file system creation events, invokes a lightweight local model via standard inference APIs to return clean JSON output containing the target directory and sanitized file name, and safely handles file relocation across local disks.
How I Built It
The architecture uses Ollama and llama.cpp to host compact, quantized open-weight models such as Qwen2.5-1.5B or Llama-3.2-1B directly on local hardware. File system monitoring is handled via Python's watchdog library to keep CPU usage negligible while waiting for new files. For reading incoming data, local open-source utilities like pypdf and pytesseract extract high-signal text snippets, which are then passed into the model using strict formatting to generate predictable, safe directory paths.
Why Does Open Innovation Matter?
Open innovation ensures complete privacy and zero running costs. Personal receipts, tax documents, and private notes never leave the machine or touch third-party cloud servers. Unlike closed proprietary APIs that charge per-token fees for constant background queries, local open weights run indefinitely for free. The setup works seamlessly without an internet connection and offers the freedom to swap, configure, or fine-tune models to fit any system hardware without external vendor locks.
Top comments (0)