ikaris 0.1.2

Ikaris - Open-source Hugging Face model safety checker

Ikaris

Ikaris is an open-source CLI tool for verifying the security of machine learning models from Hugging Face. Ikaris performs multi-layer checks to help users assess risks before deploying models in production environments.

✨ Main Feature

Ikaris performs model verification based on three methods:

🔍 Source Verification

Identifies the model creator and publisher, as well as basic metadata.

📁 File and Folder Safety Review

Review the structure and contents of the model repository for suspicious or non-standard content.

📄 Model Card and Metadata Review

Validate model documentation and metadata.

📦 Installation

Make sure you are using Python >=3.8, then install Ikaris via pip:

Using pip:

pip install ikaris

🚀 Usage

Basic usage example for checking models:

ikaris check hf-model tensorblock/Llama-3-ELYZA-JP-8B-GGUF

Another example :

ikaris check hf-model tencent/HunyuanVideo-Avatar

Output Sample

----------------------------------------
Source Verification
----------------------------------------
INFO: Source: https://huggingface.co/tensorblock/Llama-3-ELYZA-JP-8B-GGUF
INFO: Creator: tensorblock
WARNING: Model Publisher: Unknown
INFO: Tags: [...]
INFO: Downloads: 196
----------------------------------------
File & Folder Review
----------------------------------------
INFO: README.md is Safe
... (file lainnya)
----------------------------------------
Model Card Review
----------------------------------------
INFO: This model contains clear documentation.
INFO: This model contains clear metadata.

Summary: 20 Info, 1 Warning, 0 Critical

📁 Project Structure

Ikaris 
├── checker 
│   ├── __init__.py 
│   ├── file_verification.py 
│   ├── model_card_verification.py 
│   └── source_verification.py
├── helpers 
│   ├── __init__.py 
│   └── logging.py
├── __init__.py
└── main.py

🔧 Prerequisite

• Python >= 3.8
• huggingface_hub >= 0.32.4
• colorama >= 0.4.6

📝 License

Distribution under license: Apache 2.0

👤 Contributor

• Haydar Miezanie Abdul Jamil ([e-mail])

⚠️ Disclaimer

Ikaris is not a substitute for a full security audit. Use Ikaris results as an initial analysis, and combine them with manual checks for model usage on sensitive production systems.