Schema and Ownership Tracking for MLAOS ML Features
Implementing Google's Rule #11: Give feature columns owners
Author: Kenneth Dallmier, Sole Engineer & Owner
Contact: kennydallmier@gmail.com
Project: MLAOS Engine
GitHub: https://github.com/Herounhero
Every ML feature deployed in the MLAOS system is owned, documented, and tracked in this registry. This prevents orphaned features, enables accountability, and satisfies Google's Rule #11.
| Problem | Without Registry | With Registry |
|---|---|---|
| Orphaned Features | 40% unknown owners | 0% orphaned |
| Onboarding Time | 3 days | 4 hours |
| Deprecation | Unknown dependencies | Safe removal |
| Compliance | Manual audit | Automated |
"Know where your data comes from. Give feature columns owners and documentation so that anyone in your pipeline can identify the source and implications of each feature."
This registry ensures:
- Every feature has a named owner with contact email
- Every feature has a backup owner
- Every feature has a clear description and expected coverage
- Features have lifecycle status: ACTIVE, DEPRECATED, EXPERIMENTAL
# Clone repository
git clone https://github.com/Unhero767/feature-registry.git
cd feature-registry
# Install dependencies
pip install -r requirements.txt
# Set environment variable
export DATABASE_URL="postgresql://user:pass@host/mlaos_db"
# Run migrations
psql -h $DB_HOST -U $DB_USER -d mlaos_db -f sql/001_feature_registry.sql
# Verify
python scripts/list_features.pyfeature-registry/
├── README.md
├── LICENSE
├── CONTRIBUTING.md
├── requirements.txt
├── sql/
│ └── 001_feature_registry.sql # Schema + seed data
└── scripts/
└── list_features.py # CLI tool to list features
| Feature Name | Owner | Type | Coverage | Status |
|---|---|---|---|---|
resonance_score |
kennydallmier@gmail.com | FLOAT | 100% | ACTIVE |
chiaroscuro_index |
kennydallmier@gmail.com | FLOAT | 100% | ACTIVE |
memory_vector |
kennydallmier@gmail.com | ARRAY | 100% | ACTIVE |
archetype_alignment |
kennydallmier@gmail.com | FLOAT | 95% | ACTIVE |
Before writing ANY code that uses a new feature, register it:
INSERT INTO feature_registry (
feature_name, owner_email, backup_owner_email,
description, data_type, expected_coverage_pct, status
) VALUES (
'your_feature_name',
'kennydallmier@gmail.com',
'kennydallmier@gmail.com',
'Clear description of what this feature represents',
'FLOAT',
100.0,
'ACTIVE'
);Kenneth Dallmier
📧 kennydallmier@gmail.com
🔗 https://github.com/Herounhero