Architecture designed
The architecture corpus remains in doc/architecture/**. This page is the
reading map for the AI Cloud system model, runtime shape, ownership boundaries,
shared services, and deep engineering packets.
Read This In Order
If someone asks for the grand vision and then the details, the order should be:
- Architecture and Design Principles
- Overall Platform Architecture
- System Overview
- Code Structure And Layer Model
- Platform Shared Services
- Detailed Design Index
What The Architecture Corpus Already Proves
The architecture work is stronger than a typical “control plane plus web app” design. It already defines:
- a shared-platform model that AI Cloud product surfaces and future products consume;
- explicit repo/package ownership across platform, product, and shared layers;
- a real runtime-access model for terminal, browser-app, API-app, metrics, and operator surfaces;
- a deployment and promotion model that treats production as a profile, not a separate hand-built environment.
Persona Routes
| Persona | First-read path | Decision points | Next-action pages |
|---|---|---|---|
| Architecture reviewer | Design Principles, Overall Platform Architecture, System Overview | Am I reviewing runtime topology, domain ownership, code/package boundaries, API/event contracts, state transitions, data/security lifecycle, or workload access surfaces? | Code Structure and Layer Model, Workload Access and Runtime Surfaces, API Domain Authoring, State Machines, Billing And Ledger, Storage Lifecycle |
| Internal developer | Developer Implementation Map, Code Structure and Layer Model, System Overview | Which contract starts the change, which route/domain owns it, which package owns it, and which state or data model constrains it? | Detailed Design Index, Domain Ownership, API Domain Authoring |
| App developer | Build on GPUaaS, App SDK Overview, Developer APIs | Which platform responsibilities does my app rely on, and which manifest/runtime contracts must I satisfy? | Manifest Model, Artifact Runtime Lifecycle, Contract Artifacts |
| Security reviewer | Security & Production Readiness, Terminal Session Security, Storage Lifecycle | Which flows carry auth material, tenant data, ledger evidence, terminal access, or release authority? | Current Controls, Release Evidence, External Architecture Path |
Start With These
- Architecture Review Pack when the reader is preparing for ARB, JAD, or a structured design review.
- Source of truth map for the current authority across platform foundation, Fairway, v3 routes, release gates, observability, proxy, terminal, provider, and App SDK docs.
- System overview for how GPUaaS works at runtime across API, workers, events, data stores, terminal access, billing, and GPU nodes.
- Platform foundation README for the current planning bundle.
- PSSM v2 for shared services and platform/product boundaries.
- Code and deployment architecture for maps, guards, and facades.
- Domain operating model and API modularization docs for v3 ownership.
Most Undersold Architecture Strengths
| Strength | Best page to open |
|---|---|
| reviewer-first status of what is implemented vs partial | Platform Capability Summary |
| GPUaaS as first product on a reusable AI platform | Platform Shared Services |
| runtime surface taxonomy beyond “SSH and web” | Workload Access and Runtime Surfaces |
| package/layer ownership rather than repo sprawl | Code Structure And Layer Model |
| production-as-profile promotion path | Production Deployment Model |
Pages
- Architecture Review Pack
- Overall Platform Architecture
- Platform Capability Summary
- Design Principles
- What GPUaaS Already Proves as a Platform
- Platform Proof Points
- Architecture Guard and CI Enforcement
- Developer Implementation Map
- Code Structure And Layer Model
- System Overview
- Platform Shared Services
- Platform Foundation
- GPU Slicing And Scheduler Layers
- Node Agent Runtime Depth
- Source Of Truth Model
- Domain Ownership
- API Domain Authoring
- State Machines
- Billing And Ledger
- Terminal Session Security
- Workload Access and Runtime Surfaces
- Storage Lifecycle
Implementation Discipline
For the platform-foundation track, the order stays explicit: maps first, guard visibility second, facade implementation third. The portal should make that sequence easier to find, not replace the architecture docs that define it.
Canonical sources