I Built slothd: A Local-First, Offline-Capable AI Daemon That Patiently Organizes Your Messy Workspace

Description and use

slothd automatically organizes local files using a two‑stage funnel—fast TabPFN metadata triage followed by local Gemma 3 vision. All actions run offline and are reversible via SQLite WAL.

Problem

Users spend time manually organizing cryptic, cluttered files on their local machines, risking privacy by uploading to cloud AI services.

Outcome

Within a second, 10 files were scanned, 2 clutter anomalies flagged, 1 duplicate set detected, and no cloud API calls were made, demonstrating 100% air‑gapped operation.

Workflow

• Detect idle state via OS APIs
• Scan filesystem
• Fast metadata triage with TabPFN
• Local Gemma 3 vision/text comprehension via Ollama
• Generate organization plan
• Show dry‑run preview
• User approves moves
• Log moves to SQLite WAL
• Index files in Tiger Data pgvector (or local SQLite)
• Provide semantic search UI
• Enable 1‑click rollback

Evidence

The entire core workflow—idle detection, filesystem scanning, tabular metadata triage, Gemma 3 document understanding via local Ollama, SQLite WAL manifests, and 1‑click rollback—executes 100% locally on your machine with zero internet calls.

Source

In under a second, you will observe the verified metrics: 10 files scanned, 2 clutter anomalies flagged in 1 ms, 1 exact duplicate set detected via SHA‑256, 0 cloud API calls required (100% air‑gapped isolation).

Source