Technical architecture for AI Fabrix — how control, execution, knowledge, Role Assistants, and evidence layers fit together without bypassing trust boundaries.
Purpose
Concept and pillar pages explain what and why. Architecture pages explain how components enforce Operational Trust, Enterprise Knowledge, Role Assistants, and Evidence Fabrix.
Start here
- Platform architecture — four pillars and enabling layers
- Trust boundaries — who may access what, and where enforcement happens
- Capability gateway — mandatory gate before governed execution
- Platform sizing and environments — presets, environments, scale-out
Pillar architecture (deep dives)
| Pillar | Architecture focus |
|---|---|
| Operational Trust | Roles, policies, certification, capability mediation |
| Enterprise Knowledge | Metadata, relationships, Enterprise MCP |
| Role Assistants | Role-scoped assistants, tasks, outcomes |
| Evidence Fabrix | Evidence, operational memory, audit |
Enabling components
| Component | Role |
|---|---|
| Miso (controller layer) | Identity, policy, lifecycle, deployment governance |
| Dataplane | Execution boundary — CIP, connectors, MCP/REST publication |
| Builder CLI | Integration authoring, validate, upload, certify |
| Orchestration / interface layers | Optional AI and user surfaces consuming governed contracts |
Legacy layer docs (docs/dataplane/, docs/controller-layer-miso/, docs/architecture-overview/) remain available during migration — prefer this section and pillar pages for the current product story.