Verticals

The kernel thinks. You define the domain.

Vitruvyan is a cognitive operating system with a complete epistemic pipeline — perception, memory, reasoning, verification. It doesn't know your domain. That's the point. You bring the vocabulary, the rules, the data sources. The kernel handles everything else.

What is a vertical

A vertical is a thin domain layer that sits on top of the Vitruvyan kernel. It contains three things: a vocabulary (what entities exist in your domain), connectors (where data comes from), and rules (domain-specific validation and business logic).

The kernel provides the cognitive infrastructure — ingestion pipelines, semantic memory, LLM orchestration, reasoning chains, audit trails. Your vertical doesn't rebuild any of that. It configures it.

Think of it like a database schema: you define the shape of your domain, and the system knows how to store, query, reason about, and validate it. Except here, the system also knows how to perceive, remember, and explain its reasoning.

You don't need to understand the kernel internals. You need to understand your domain. That's where your expertise goes.

From zero to deployed

Four steps. No boilerplate. The CLI scaffolds the structure, you fill in the domain knowledge, and deploy.

/01

Scaffold

vit vertical create my-domain

One command generates the full vertical structure: entity definitions, domain connectors, validation rules, and a ready-to-use manifest. You start with a working skeleton, not a blank page.

/02

Define

my-domain/vertical_manifest.yaml

Declare your domain vocabulary — entities, relationships, signals. The manifest tells the kernel what your domain looks like. No code required at this stage — just configuration.

/03

Connect

my-domain/connectors/

Write thin adapters for your data sources — APIs, databases, file systems, IoT streams. The kernel handles ingestion, normalization, and semantic indexing. You just point it at the data.

/04

Deploy

vit vertical deploy my-domain

Your vertical plugs into the running kernel. Perception, memory, reasoning, verification — all active. Every output carries provenance, confidence scores, and a full audit trail. From day one.

What the kernel gives you

Every vertical inherits the full cognitive stack. You don't build these — they're already running.

Perception pipeline

Ingestion, parsing, entity extraction, language detection, normalization. Data enters clean.

Semantic memory

Vector storage, coherence scoring, temporal decay, retrieval with provenance. The system remembers.

Reasoning engine

LLM orchestration, chain-of-thought tracing, confidence scoring. Every conclusion is traceable.

Verification layer

Invariant checks, contradiction detection, source cross-referencing. Nothing ships unchecked.

Audit trail

Every input, every decision, every output — logged immutably. Full reproducibility.

Event bus

Redis Streams-based cognitive transport. Services communicate asynchronously. Payload-blind by design.

Example verticals

Reference implementations that show how different domains plug into the same kernel.

Finance

Risk scoring, market perception, regulatory compliance

risk-enginemarket-perceptionaudit-ledger

Healthcare

Clinical reasoning, differential diagnosis, patient memory

patient-intakeclinical-memorydiagnosis-engine

Legal

Contract analysis, precedent search, compliance verification

contract-parserprecedent-memorycompliance-engine

Industrial

Predictive maintenance, anomaly detection, sensor fusion

sensor-intakeanomaly-detectormaintenance-planner

Ready to build?

Install Vitruvyan, scaffold your vertical, and start building. The kernel is ready — your domain is the only missing piece.