Get started

ArchiveTuna surveys old folders and helps you find work worth revisiting. Start with an inventory, then add a local model when you want descriptions and a ranked review queue.

Requirements

• Node.js 22 or newer.
• Ollama with a downloaded model for description, ranking, and optional STL vision.
• f3d and ImageMagick only when rendering STL previews.

Inventory, duplicate analysis, saved-report rebuilding, and facts-only STL metadata work without a model. ArchiveTuna is developed on Windows; core inventory uses portable Node APIs. Linux/macOS behavior needs platform testing.

Published package

Install the published CLI from npm (version 0.2.0):

npm install -g archivetuna
archivetuna --version
archivetuna --help

First inventory

Choose a small representative archive first. Keep outputs outside the source tree, or use the default output directory (which the scanner excludes).

archivetuna scan "/path/to/archive" --out ./review
archivetuna dupes --out ./review

Read review/dupes.md. These are metadata comparisons; no file contents were read and no originals were moved or deleted.

First ranked report

archivetuna init profile.yaml

Edit the profile to describe what matters to you. Then download a model that fits your hardware and start the local Ollama service:

ollama pull gemma4:12b
archivetuna doctor
archivetuna run "/path/to/archive" --profile profile.yaml --out ./review --limit 20 --dry-run
archivetuna run "/path/to/archive" --profile profile.yaml --out ./review --limit 20

The default model is gemma4:12b. Select another downloaded model with --model. Read review/report.md and review/novel.md. Try a focused profile before spending time on a much larger tree.

Continue with the archive workflow or write a profile.