June 2, 2026

Local AI Tidy with Ollama: Organize Files Without Sending Them to the Cloud

Set up optional on-device AI rename and organize suggestions in DupeZappa using Ollama: opt-in, local, and review-before-apply.

ai-tidy
ollama
privacy

DupeZappa does not require AI to scan duplicates, find junk, or rename batches. When you want help turning messy filenames into consistent patterns, AI Tidy can draft suggestions, and with Ollama, that draft stays on your machine.

Why local AI matters for file cleanup

Rename suggestions are built from filenames and small readable content samples from your folders. Sending that to a cloud API may be fine for some users and a hard no for others: legal archives, client work, medical imaging, or anything on an air-gapped PC.

Ollama runs models locally. DupeZappa talks to the local Ollama HTTP API directly. No Node sidecar is required for the built-in provider.

What you need installed

Ollama for Windows: installed and running as a background service.
Models pulled locally: DupeZappa expects:
- Text model: qwen3.5:9b
- Vision fallback (optional): qwen3-vl:8b
DupeZappa: open Settings, choose Ollama as the AI Tidy provider.

Use Check Ollama in Settings to verify the runtime and models before running a large folder.

Enable AI Tidy the safe way

Pick a small test folder first: ten to fifty files, not ten thousand.
Choose a tidy mode (clean names, archive layout, or similar).
Generate a plan and read the summary; counts, skips, and conflicts should be visible.
Open the per-file diff. Reject anything that renames a file you still recognize by its old name.
Apply the batch. Use undo if a pattern was too aggressive.

AI Tidy is opt-in at every layer: provider selection, plan generation, and apply.

OCR and vision fallback

For image-heavy folders with little useful text in filenames, enable OCR hints in Settings and pick a vision fallback policy. The vision model runs locally through Ollama as well, with still no cloud upload from DupeZappa's built-in path.

If you prefer a custom HTTP endpoint instead, use Custom API in Settings and point to your own service. That path is for advanced setups; Ollama is the default local story.

When not to use AI Tidy

Skip AI for:

Strictly regulated filenames that must not change (evidence chains, build artifacts).
Folders where the current names are already canonical.
One-off renames; the batch rename tools are faster and fully deterministic.

Use AI Tidy when patterns are obvious to a human but tedious to write as rules: event photos, download dumps, mixed document scans.

Privacy in one sentence

With Ollama selected, DupeZappa does not send your file contents to MerginIT servers for AI Tidy; processing stays in your local stack, and you still approve every change before it hits disk.

Local AI is a drafting tool. Your review step is what makes it production-safe.