detect-file-type-local 0.1.1

Security-focused local file type detection powered by Google Magika

detect-file-type-local

An OpenClaw skill for AI-powered local file type detection.

Wraps Google Magika to provide ML-based file type identification that runs entirely offline. No API keys, no network calls — just local inference on an embedded ONNX model.

Features

• 214 file types detected by content, not extension
• Fully offline — no network access required
• Fast — only reads the bytes needed for classification
• Batch support — process multiple files or entire directories
• Multiple output formats — JSON, human-readable, bare MIME type
• Stdin support — pipe content directly

Quick Start

pip install detect-file-type-local

# Detect a single file
detect-file-type-local document.pdf

# Batch detect
detect-file-type-local --human *.pdf *.png

# Recursive directory scan
detect-file-type-local -r ./uploads/

# Pipe from stdin
cat mystery_file | detect-file-type-local -

Compatibility alias: detect-file-type remains available.

Output Formats

JSON (default):

{
  "path": "photo.jpg",
  "label": "jpeg",
  "mime_type": "image/jpeg",
  "score": 0.99,
  "group": "image",
  "description": "JPEG image",
  "is_text": false
}

Human-readable:

photo.jpg: JPEG image (image/jpeg) [score: 0.99]

MIME-only:

image/jpeg

OpenClaw Skill

See SKILL.md for the OpenClaw skill definition, including structured output schemas and usage guidance for LLM integration.

OpenClaw skill metadata now auto-installs from PyPI package detect-file-type-local.

Development

pip install -e '.[dev]'
pytest tests/ -v
ruff check .

Release

PyPI publishing is automated via GitHub Actions (Publish Python Package workflow):

1. Create a GitHub release with a tag matching package version (for example, v0.1.0)
2. Workflow builds and validates artifacts
3. Workflow publishes to PyPI via trusted publishing

After PyPI release, update and republish the ClawHub skill metadata to enable auto-install from detect-file-type-local.

License

MIT — see LICENSE.

This project uses Google Magika (Apache-2.0). See NOTICE and THIRD_PARTY_LICENSES.md.