magika を Homebrew, apk, Nix, winget でインストール

プロジェクトリンク

• ホームページ
• リポジトリ
• 上流ドキュメント
• パッケージマネージャページ

コマンドとエイリアス

• magika

プロジェクトの歴史

Google created the public repository in August 2023 and announced the open-source release in February 2024. Official docs describe a model trained and evaluated on roughly 100 million samples across more than 200 content types, with about 99 percent average precision and recall on Google's test set.

The project moved beyond a Python package into a multi-language toolchain. The README documents a Rust command-line tool, Python API, JavaScript package, Rust bindings, web demo, and additional bindings, with release tags split by CLI, Python, JavaScript, and Go components.

採用の歴史

Magika had unusually fast visibility because Google said it was already used at scale inside Gmail, Drive, and Safe Browsing routing, processing hundreds of billions of samples weekly. Official docs also note integrations with VirusTotal and abuse.ch.

The open-source package spread into package managers quickly: the README documents Homebrew, pipx, pip, cargo, npm, and installer scripts, while the supplied Homebrew input also lists apk, Nix, and winget packages. The GitHub repository has tens of thousands of stars and more than a thousand forks, reflecting broad interest from security and developer-tooling users.

使われ方

The CLI scans files or directories and can emit descriptions, labels, MIME types, scores, JSON, or JSONL. Typical package-manager usage is `magika PATH`, often recursively, as a faster and ML-backed complement to extension- or magic-number-based detection.

Library users embed Magika to classify uploads, repositories, malware samples, or large file corpora. The docs emphasize near-constant inference time because the model uses a limited subset of file content.

パッケージ好きにとっての重要性

Magika is package-nerd interesting because it packages an ML model as a normal fast CLI. It belongs in the same mental drawer as `file` and libmagic, but with model assets, language bindings, and security-scanning use cases layered on top.

Its multi-artifact release style matters to maintainers: Homebrew ships the CLI, Python users install `magika`, Rust users can install `magika-cli`, JavaScript users have npm bindings, and downstreams need to track model/data changes as part of versioning.

タイムライン

• 2023: Public GitHub repository created.
• 2024: Google open-sourced Magika and announced it on the Google Open Source Blog.
• 2024: CLI, Python, Rust, and JavaScript packaging paths documented.
• 2025: Research paper published at ICSE according to official docs.
• 2026: CLI and binding release tags continue in the public repository.

Related projects

• libmagic and the Unix `file` command are the traditional content-type detection tools Magika is commonly compared with.
• VirusTotal and abuse.ch are named by Google as integrations.
• Gmail, Drive, and Safe Browsing are named by Google as internal-scale consumers.

ソース

• https://github.com/google/magika
• https://github.com/google/magika/releases
• https://opensource.googleblog.com/2024/02/magika-ai-powered-fast-and-efficient-file-type-identification.html
• https://securityresearch.google/magika/introduction/overview/
• source_facts.package-manager
• source_facts.repo