valencity 0.1.0

A privacy & ML safety toolkit for production ML pipelines

🛡️ valencity

The ML Safety Fortress
Privacy Engineering • Data Validation • Leakage Prevention

Features • Installation • Quick Start • Documentation • Contributing

---

🚀 Why valencity?

valencity is a comprehensive toolkit designed to make machine learning pipelines safer, compliant, and reliable. It bridges the gap between data engineering and privacy compliance, offering a "fortress" of protection for your data assets.

Feature	valencity 🛡️	Great Expectations	Pandera	Custom Scripts
PII Detection	✅ 50+ Patterns	❌	❌	⚠️ Manual
Masking/Anonymization	✅ K-Anon, DP	❌	❌	❌
Leakage Prevention	✅ Full Suite	❌	❌	❌
Drift Detection	✅ 4+ Methods	✅	❌	⚠️ Manual
Safe Cross-Validation	✅ Built-in	❌	❌	❌
HTML Reports	✅ Beautiful	✅	❌	❌

---

✨ Features

🕵️ PII Detection & Anonymization

• Detect 50+ types of PII including IBAN, Passports, and API Keys.
• Smart Masking: Redact, hash, partial mask, or replace with fake data.
• NLP Support: Named Entity Recognition for names and addresses (via spaCy).

✅ Data Validation & Quality

• Drift Detection: Catch distribution shifts before they break your model.
• Schema Validation: Enforce strict data contracts on your DFs.
• Quality Checks: Nulls, duplicates, outliers, and cardinality.

🚫 Leakage Prevention

• Safe Cross-Validation: Automatic leakage checks during CV.
• Leakage Detectors: Find target leakage, train/test overlap, and temporal leakage.
• Safe Splitters: Time-series and group-aware train/test splitting.

---

📦 Installation

# Install core package
pip install valencity

# Install with NLP support (for names/addresses)
pip install valencity[nlp]
python -m spacy download en_core_web_sm

---

⚡ Quick Start

1. Protect Your Data (PII)

import pandas as pd
from valencity.pii import PIIDetector, PIIMasker

# Scan for PII
df = pd.read_csv("users.csv")
detector = PIIDetector()
report = detector.scan_dataframe(df)

if report.has_pii:
    print(f"⚠️ Found PII in columns: {report.pii_columns}")

# Mask it
masker = PIIMasker(strategy="partial")  # j***@example.com
safe_df = masker.mask_dataframe(df)

2. Validate Quality & Drift

from valencity.validation import DriftDetector, DataSchema

# 1. Define expectations
schema = DataSchema.from_dataframe(train_df)

# 2. Check for drift in production
detector = DriftDetector(method="ks_test")
detector.fit(reference_df=train_df)

drift_report = detector.detect(current_df=new_data)
if drift_report.has_drift:
    print(f"🚨 Data drift detected using {drift_report.method}!")

3. Prevent Model Leakage

from valencity.leakage import SafeCrossValidator
from sklearn.linear_model import LogisticRegression

# Safe CV ensures preprocessing only sees training fold
# Prevents subtle data leakage bugs
cv = SafeCrossValidator(n_splits=5, preprocessor=StandardScaler())

scores = cv.cross_val_score(LogisticRegression(), X, y)
print(f"Realistic Accuracy: {scores.mean():.4f}")

---

🗺️ Roadmap

• CLI Tool: valencity scan data.csv
• Privacy Module: Differential Privacy & K-Anonymity
• Synthetic Data: Generate safe, realistic test data
• Integrations: MLflow, Airflow, and dbt support

---

🤝 Contributing

We welcome contributions! Please check out the Issues page.

1. Fork the repo
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

---

Built with ❤️ for the ML Community