A desktop app that lets an LLM reorganize your files—locally

What it does

AI File Sorter is a C++ / Qt6 desktop application for Windows, macOS, and Linux that scans folders and proposes new names and categories for your files. It handles images via built-in visual LLMs like Gemma 3 4B IT, reads documents with text LLMs, and cleans up audio and video filenames using embedded metadata tags. Every proposed move appears in a preview table for your approval before the app touches the filesystem.

The interesting bit

The app keeps categorization consistent across runs by blending raw LLM output with taxonomy normalization, session consistency hints, and a local cache of your past approved decisions—essentially building a lightweight, private feedback loop without retraining the model. It can run entirely offline using bundled GGUF backends, and it will only reach for a remote API if you explicitly plug in your own key.

Key highlights

• Runs fully offline with local visual and text models (Gemma 3 4B IT, LLaVA 1.6 Mistral 7B, or custom GGUFs) and sends no telemetry.
• Supports remote models via your own OpenAI, Gemini, or OpenAI-compatible API credentials.
• Offers two categorization modes: More Refined for detailed labels or More Consistent to keep folders uniform.
• Lets you constrain the model with category whitelists and review edits before any rename or move is applied.
• Includes persistent undo, dry-run previews, and a local learned-hints database that improves future suggestions based on your past approvals.

Verdict

Worth a look if your Downloads folder or NAS has become an unnavigable soup of generic filenames and you want a private, review-first cleanup tool. Skip it if you already have a rigid naming convention and zero interest in LLM suggestions.