06 / 06 · 2025

OpenClaw

Your files, organized offline

A privacy-first local AI agent that sorts teaching files into a structured archive — running fully offline with scoped, hardened permissions.

Building in progressSolo build · Local-AI agent

// screens

01 · Dry Run
Proposed file moves before anything is touched
02 · Local Index
Semantic catalog built and queried fully offline
03 · Permissions
Scoped filesystem access, one folder at a time

// The problem

Teachers sit on a decade of unsorted lesson PDFs, slides and worksheets. Cloud AI tools can sort them but require uploading student-adjacent material — a non-starter in Swiss schools.

// The approach

  1. 01Runs entirely on the teacher's laptop via Ollama — no files leave the device.
  2. 02Scoped filesystem permissions and a dry-run mode so the agent proposes a tree before touching anything.
  3. 03Indexes content in a local SQLite catalog with semantic tags for fast re-finding.

// The outcome

  • →Used to migrate 10+ years of my own teaching archive into a clean tree.
  • →Zero cloud calls; safe to use with classroom-adjacent material under nFADP.

// Tech stack

PythonOllamaLocal LLMSQLite
← Previous
GAVL
Next →
CodeZug