I Was Too Lazy to Organize My Files, So I Taught a Sloth (Gemma) to Do It

Description and use

The author built LazyDesk AI, a local LLM‑powered tool that cleans up file names, detects duplicates, and produces a ZIP after user approval.

Problem

The author’s Downloads folder contained messy filenames and duplicate files, and manual organization was tedious and unlikely to be performed.

Outcome

In a test with six messy files, the app detected a duplicate pair, processed all files in 32 seconds, and produced a ZIP containing only approved files in suggested folders. The author estimates saving ~3 minutes for six files, implying significant time savings for larger collections.

Workflow

• Upload files to the Streamlit UI
• Ollama runs local Gemma 3 (1B) to suggest clean names and folders
• SHA‑256 hashes identify duplicate pairs
• Duplicate files are flagged and can be removed
• An editable approval table (pandas st.data_editor) displays suggestions
• User reviews and edits suggestions
• Approved files are packaged into a ZIP for download

Evidence

I built LazyDesk AI 🦥, a file decluttering app where a local open-weight Gemma model reads your messy files and suggests clean names and folders.

Source

All 6 files processed in 32 seconds

Source

Duplicate detective (SHA-256) caught the identical pair: duplicate_copy.txt and duplicate_original.txt

Source