AtelierOntology-native application generation

Declare the interesting 20% — entities, actions, surfaces, identity. The platform generates the 80%: a staff console, a citizen portal, and the APIs behind both. Multi-tenant, from one declaration.

Introduction

The mental model

Subsystems

📐

Declare, don't code

An application is data — ontology rows, expressed canonically as one Vertical Sheet (YAML) or edited live in operator editors. Entities, actions, surfaces, notifications, identity, reports, and portal pages are all declared, never hand-written.

Vertical Sheets

🧬

Generate, don't build

From the declaration Atelier generates the running, multi-tenant experience with zero per-vertical code — one shared renderer drives both the admin console and the citizen portal.

The Surface Plane

🧱

Compound software

A growing catalog of building blocks — entity types, widgets, action shapes, view kinds — that compose without breaking. Any entity renders in any widget; any action fires from any surface.

The thesis

🪞

Fork a worked example

A new tenant comes up fully configured by forking one template, then diverges by editing its own copy with the exact same editors the platform author used. One cross-plane bridge, zero special-case code.

App Provisioning

🛡️

Authorization as data

Every read resolves under a per-app, ontology-native namespace with the caller's own identity. Policy is declared as acl_rule rows and compiled to SQL — never branched in Python. Fail-closed by construction.

Authorization & Identity

🩺

Prove it from zero

A from-zero discipline stands up an isolated stack and asserts it self-provisioned with no hand-seeding — backed by build-time guards, an in-product diagnostics gate, and read-only canaries.

Diagnosis

What this is. The canonical specification of the Atelier platform — the basis of every platform evolution going forward. Eleven code-anchored subsystem sections plus the connective tissue: vision, the plane model, the compound-software thesis, the end-to-end lifecycle, cross-cutting invariants, a consolidated glossary, an honest gaps register, and the north-star roadmap.
Inspired by — and a sibling to — Martha: same "declare the 20%, the platform provides the 80%" clarity, turned toward civic and operational apps instead of chat agents. Martha is also Atelier's durable-execution backend.