Controller Layer (Miso)

Prev Next

# Controller Layer (Miso)

Governance, identity, policy, and lifecycle

Miso is the Controller layer of AI Fabrix. It is the platform control plane responsible for:

  • Identity & access (who can do what)
  • Policy & governance (what is allowed and under which constraints)
  • Environment lifecycle (how configurations are promoted and controlled)
  • Audit & observability (how actions are recorded and evidenced)

Control Plane vs Data Plane

AI Fabrix is intentionally split into:

  • Controller (Miso): governs identity, policy, lifecycle, and evidence.
  • Dataplane: executes integrations and processes business payloads (CIP and dataplane runtime).

This separation ensures governance is structural, not re-implemented in each app, agent, or integration.

Why the Controller never touches business data

Miso is designed so it does not handle business payloads. Instead, it:

  • Issues and validates identity context (Entra ID claims, groups, roles)
  • Applies governance decisions (policies, quotas, allowed endpoints)
  • Controls environments (dev/test/prod) and deployment promotion
  • Records evidence (audit trails, access decisions, configuration change history)

Business data access and transformation occur only in the Dataplane execution boundary.

Result: Governance can be audited and explained without requiring the Controller to become a data-processing system.

What architects validate

  • Control-plane components do not store or process business payloads.
  • Identity context is preserved end-to-end (no "service account as user" pattern).
  • Policies are evaluated consistently and centrally.
  • Environment promotion is controlled and reproducible (Dev → Test → Prod).
  • Auditing is deterministic (not forensic reconstruction).

Controller Layer sub-articles

  • Miso Overview — Role of the Controller, control plane vs data plane separation, platform capabilities
  • Identity & Access — Entra ID, RBAC and ABAC, SCIM, workspace and environment scoping
  • Policy & Governance — Policy packs, egress, model access, quotas, compliance
  • Environment Lifecycle — Dev → Test → Prod promotion, drift prevention, versioning, approval gates
  • Audit & Observability — Audit trails, policy decisions, model usage tracking, cost attribution

Miso in the architecture

Miso in the architecture


Key takeaway: Miso governs who can do what, what is allowed, how environments are promoted, and how actions are evidenced — without touching business data. All data access happens in the Dataplane.