Hybrid retrieval

Search by structure, meaning, relationships, and time.

MaluDB combines relational filters, vector similarity, graph traversal, and bitemporal validity into permission-aware retrieval plans for AI applications.

select * from match_memory( query => 'why did renewal risk increase?', search_modes => array['relational','vector','graph','temporal'], compartment => 'enterprise_sales' );

Relational precision

Constrain memory retrieval by tenant, account, source, role, time window, confidence, and validation state.

Vector recall with compartments

Semantic candidates are generated inside authorized compartments so irrelevant or forbidden memories stay out of context.

Graph and temporal reasoning

Follow relationship edges and valid-time windows to explain how a decision, incident, or workflow evolved.

SQL

Structured filters

Use SQL for exact constraints and reporting.

VEC

Semantic recall

Find memories by meaning and context.

BIO

Bitemporal truth

Ask what was known, when it was true, and when it changed.