Data and Storage Design

Vektagraf employs a model-aware object graph storage model. Every model and field in the schema influences storage layout.

Key design considerations:

Object Graph vs Relational vs Document

Traditional relational systems enforce global schemas unsuitable for dynamic, nested models. Document systems sacrifice rigor and relationships. Graph systems lack strong validation semantics. Hyperstores integrate all three:

• Relational principles → constraints, references, enumeration
• Document principles → structural flexibility, nested objects
• Graph principles → relationships, traversal semantics

This hybrid approach allows schemas to express domain models naturally.

Version Chains

Every object is versioned. Version chains produce an immutable log of state. These chains are cryptographically linked, enabling provable history.

Indexing

The schema drives index generation:

• Field indexes
• Composite indexes
• Vector indexes (HNSW, IVF, PQ)
• Full-text indexes
• Range/ordered indexes
• Tenant-specific indexes

Whether something should be indexed is a schema-level decision, not an implementation detail.