Skip to content
ORVIXLABSPrivate AI systems
ESStart a project
// ENGINE

CASANDRA

DDE

A divergent discovery engine for expanding sources, candidates, relationships and hypotheses with provenance and explicit limits before verification.

PRODUCTIONVersion 1.0.0Updated 2026-09-15
CASANDRAConceptual diagram: Seed, Documents, Entities, Relations, Hypotheses, Verify.SeedDocumentsEntitiesRelationsHypothesesVerify
// HOW IT WORKS
CASANDRA DDE · DIVERGENT DISCOVERY SEEDperson · entity · eventSOURCESrecords · documents · dataEXPANDheterogeneous agentsCANDIDATESrelations · hypothesesPROVENANCEorigin of each finding CASANDRA opens the search space. It does not decide what is true. hands off to verification
// WHAT IT DOES

Discover what deserves investigation before deciding what holds.

01 / INPUT

A seed: person, company, event, document, entity or initial problem.

02 / PROCESS

Expands sources and relationships, generates candidates and hypotheses, and preserves provenance.

03 / OUTPUT

A discovery package containing candidates, hypotheses, relationships, sources and provenance.

// SYSTEMS BUILT ON THIS ENGINE

The engine stays underneath. The product changes with the real-world problem.

This list is generated automatically from published systems that declare this engine as their cognitive base.

01
Ariadna

It is a corporate and relationship investigation system that reconstructs links among people, companies, addresses, documents and scattered events to reveal structures that are invisible when each fact is reviewed in isolation.

→02
Vector

It is a commercial-intelligence system for grain elevators that detects when there is a concrete reason to contact a producer without confusing a good market price with a good decision for the elevator.

→03
Nexus

It is an operational-intelligence system for companies that connects content, SEO/GEO, advertising, social channels, customer interactions and metrics to decide what should happen next from shared evidence.

→04
Geovox

It is a generative-presence intelligence system that measures how an organisation appears in AI engines and which public evidence influences those responses.

→05
Industrial Intelligence Systems

It is a family of industrial-intelligence systems for Oil & Gas, mining and plants that relates telemetry, energy, asset condition and maintenance without taking over the critical control loop.

→

The hard problem is often not deciding. It is discovering what deserves a decision.

CASANDRA DDE operates before verification. It starts from a seed, explores declared sources, discovers candidates, relationships and patterns, and turns relevant findings into new investigation paths. Expansion remains bounded by contracts, budget, depth and provenance rules.

Provenance from the first finding

A relation without origin is weak evidence. CASANDRA keeps where each item came from, when it was retrieved and what transformation turned it into a candidate. Another layer can return to the source instead of trusting model memory.

Diverge without falling in love with a hypothesis

The engine opens the search space. It can produce rival hypotheses and initially unlikely paths. That is precisely why it does not issue the final verdict: discovery and certification remain separate functions.

CASANDRA + CERBERUS

This is one of OrvixLabs' natural combinations. CASANDRA asks “where else should we look?” CERBERUS asks “does this actually hold?” One opens the field; the other closes it with evidence.

Real applications

Due diligence, corporate intelligence, commercial research, fraud, oversight, public-record analysis, M&A, legal investigation, investigative journalism, R&D and any problem where relevant material is distributed across too many sources for a human team to traverse without losing relationships.

// CONCRETE EXAMPLE

Input: one company and two known documents. Process: CASANDRA expands people, addresses, acts and related sources without asserting guilt. Output: new investigation paths with exact provenance for each finding.

// ORVIXLABS

The architecture is defined around the operation, its data, constraints and verification requirements.

Discuss an architecture