equalyze 0.2.0

AI Fairness & Compliance SDK — CI/CD pipeline gate for EU AI Act compliance, bias auditing, and Equalized Odds evaluation.

Equalyze SDK

AI Fairness & Compliance SDK for CI/CD Pipelines

Equalyze is a Python SDK that integrates AI bias auditing directly into your CI/CD pipeline. It evaluates datasets against multiple fairness metrics — including Disparate Impact Ratio, Equalized Odds, and FPR Parity — and blocks deployments that fail compliance thresholds with a standard sys.exit(1).

Built for EU AI Act compliance.

Installation

pip install equalyze

Quick Start

from equalyze import EqualyzeSDK

sdk = EqualyzeSDK(api_key="your_api_key", base_url="https://your-equalyze-instance.com")

# Upload a dataset
dataset_id = sdk.upload_dataset("hiring_data.csv")

# Run a fairness audit with a compliance threshold
# This will sys.exit(1) if the model fails — blocking your pipeline automatically.
sdk.run_audit(
    dataset_id=dataset_id,
    protected_attributes=["gender", "race", "age"],
    threshold=0.85
)

CI/CD Integration

GitHub Actions

- name: Equalyze Fairness Gate
  env:
    EQUALYZE_API_KEY: ${{ secrets.EQUALYZE_API_KEY }}
    DATASET_PATH: ./data/model_output.csv
  run: equalyze-audit-gate

GitLab CI

fairness-gate:
  script:
    - pip install equalyze
    - equalyze-audit-gate
  variables:
    EQUALYZE_API_KEY: $EQUALYZE_API_KEY
    DATASET_PATH: ./data/model_output.csv

Metrics Evaluated

Metric	Description
Disparate Impact Ratio	Outcome ratio between protected groups (4/5ths rule)
Equalized Odds	True Positive Rate parity across groups
FPR Parity	False Positive Rate parity across groups
Statistical Parity	Outcome distribution equality

License

MIT