Consensus

Consensus is the autonomous semantic layer that powers every Colrows runtime. It's a living, governed graph of business meaning - entities, metrics-as-state, events, concepts, constraints, and policies - that AI agents and analysts reason over before any data is touched.

Why analytics semantics aren't enough

Traditional semantic layers were built to support dashboards and BI tools. They focus on simplifying query construction for humans and standardizing metric definitions. AI agents are not analysts: they operate autonomously, continuously, and across domains. They need semantics that go far beyond metrics and dimensions - state, causality, constraints, event lifecycles, and organizational rules.

Pointing an analytics semantic layer at AI workloads produces brittle systems that work only for narrow query patterns and collapse under real-world complexity. Consensus is built differently: it is AI-native semantic infrastructure.

Semantic primitives

Consensus models meaning through first-class primitives. Together they define the semantic state space in which agents and humans operate.

Metrics as state, not queries

In Consensus, metrics are not stored SQL. They are modelled as derived semantic state - a continuously interpretable representation of business reality. A metric like Net Revenue does not merely encode how to compute a number; it encodes:

• Business meaning - what Net Revenue is, and how it differs from Gross Revenue or Bookings.
• Valid grain - per order, per invoice, per customer, per day.
• Dependencies - orders, refunds, chargebacks, discounts.
• Constraints - how it can be filtered, grouped, compared.
• Downstream impact - which dashboards, agents, or signals rely on it.

When an agent observes a decline in Net Revenue, Consensus lets it reason semantically, not procedurally - distinguishing volume-driven decline from refund-driven erosion, because those relationships are explicit in the metric's state.

The semantic execution graph

All primitives connect through the semantic execution graph. It is not a passive knowledge store or a documentation artifact; it is an active runtime structure that agents consult while reasoning, deciding, and acting. Every step in the reasoning process is validated against the graph before any query is generated or any action is taken.

When an agent is asked to analyze a drop in Net Revenue, it does not simply join tables and aggregate. It follows semantic paths that link Net Revenue to its contributing events - orders, discounts, refunds, chargebacks - while honoring constraints such as valid grain, time windows, and organizational policies. Invalid joins or semantically incorrect aggregations are blocked before execution.

Multi-vector embeddings

Each concept in Consensus carries up to three vectors:

• Definition vector - derived from the canonical, governed definition.
• Usage vector - derived from observed query, alert, and dashboard patterns.
• Combined vector - a weighted blend used for recall when natural language drifts ("lapse" vs. "churn", "GMV" vs. "gross merch value").

Vectors identify candidates; structural reasoning makes the final determination. Embeddings are never the source of truth.

Scopes & personas

Consensus supports four levels of scope:

• Global - definitions shared across the enterprise.
• Datastore - source-specific definitions and join paths.
• Persona - role-based overrides and policy bindings.
• User - personalized context from prior interactions.

Personas are first-class graph nodes, not just access lists. Binding a request to a persona resolves an allowed subgraph at compile time. Compilation occurs entirely within that subgraph, which is what makes governance structural rather than procedural.

Autonomous semantic agents

Consensus is operated by a coordinated set of autonomous agents. They are not workflow automations - they are reasoning components that operate over the semantic execution graph to keep enterprise meaning accurate, consistent, and actionable.

• Discovery agentsIngest schemas, metadata, documentation, and Confluence pages to identify candidate entities, events, metrics, and relationships.
• Architecture agentsValidate grain, dependencies, and constraints - refusing to publish definitions that violate business logic.
• Learning agentsObserve how humans and agents use semantics in practice and refine definitions, examples, and synonyms accordingly.
• Monitoring agentsDetect semantic drift, anomalies, and broken assumptions as schemas, business processes, and behaviors evolve.

Drift, conflict, and equivalence

Consensus runs continuous structural and statistical checks:

• Statistical fingerprinting of column distributions to detect data drift.
• Structural diffing of dataset nodes to detect schema evolution.
• Hybrid vector + structural equivalence to detect duplicate or conflicting metrics - vectors locate candidates, structure makes the final call.

Governance & explainability

Every change to a node in Consensus is versioned, attributed (human or agent), timestamped, and linked to the agents, dashboards, and signals it impacts. This is the substrate the audit trail rests on. Every executed query carries a trace describing which definitions were used, which paths were traversed, and which constraints were enforced - and any historical query can be re-executed against the exact semantic state that was active at the time.

What Consensus enables

• An AI Analyst whose answers are repeatable, auditable, and explainable across teams and time (see Colrows AI).
• Agent-grade infrastructure - the same graph powers your custom agents over the API.
• Cross-system consistency - definitions remain stable across dashboards, signals, agents, and notebooks.
• Self-maintaining semantics - autonomous agents keep the graph current without a team of curators.