DocumentationNeurondB Documentation
Feature Store
Create Feature Store
Register features, apply transformations, and serve consistent features for training and inference.
Initialize Feature Store
Create feature store
-- Create feature store for user features
SELECT neurondb.create_feature_store(
'user_features', -- Store name
'users', -- Entity table
'user_id' -- Entity key column
) as store_id;Register Features
Register feature definitions including raw features and engineered features.
Register features
-- Register raw features
SELECT neurondb.register_feature(
1, -- store_id
'total_spent', -- feature name
'numeric', -- data type
'total_spent' -- SQL expression
) as feature_id;
-- Register engineered features
SELECT neurondb.register_feature(
1,
'avg_transaction_value',
'numeric',
'total_spent / NULLIF(num_transactions, 0)'
) as feature_id;Serve Features
Retrieve features for training or inference.
Serve features
SELECT neurondb.serve_features(
'user_features', -- Store name
ARRAY[1, 2, 3], -- Entity IDs
ARRAY['total_spent', 'avg_transaction_value'] -- Feature names
) as features;Next Steps
- Hyperparameter Tuning - Optimize model parameters
- Unified ML API - Consistent training interface
- Model Management - Deploy models