Architecture
CADANCE™ Specification
A deep dive into the CADANCE™ Specification, the declarative heart of the MindLab platform.
The core philosophy of the MindLab platform is “Configuration-first.” We believe that for AI to be a trusted, enterprise-grade asset, its behavior must be deterministic, repeatable, and auditable. This cannot be achieved through prompting alone.
This is why we created the CADANCE™ Spec, a single, versioned specification that declaratively defines the entire lifecycle of an AI-driven workflow. It is a human-readable, machine-executable contract that ensures every action taken by the system is traceable, reproducible, and aligned with your explicit intent.
The 7 Components of the CADANCE™ Spec
CADANCE™ is an acronym for the seven core components of the specification:Config
Config
Defines the high-level objectives, success criteria, and operational budgets (cost, latency, SLOs) for the workflow.
Agents
Agents
Specifies the roles, capabilities, constraints, and escalation paths for the specialist agents involved in the workflow.
Data
Data
Declares the knowledge sources, memory scopes, data contracts, and citation requirements for the workflow.
Analysis
Analysis
Defines the output templates, evaluation rubrics, and schemas that the agents must adhere to.
Norms
Norms
Specifies the policies, governance rules, approval requirements, and human-in-the-loop (HITL) checkpoints for the workflow.
Controls
Controls
Defines the state graphs, retry/timeout logic, gating conditions, and deterministic seeds for the workflow’s execution.
Execution
Execution
Specifies the runbooks, handoffs, logging levels, artifact retention policies, and audit requirements for the workflow.