reference implementation

Not designed
on a whiteboard.
Extracted from production.

Aurora Composable comes out of a live consumer product and the backend ecosystem behind it — two native mobile clients, a cloud backbone of single-purpose functions, an administrative companion, and the key collections that drive all of them. What follows describes that system generically.

100
single-purpose backbone functions, and not one framework between them

Everything separating one tier from another is a key. That is the claim this platform exists to evidence.

the pieces

What is actually running.

01

Two native clients

An iOS client and an Android client, each written entirely in its platform's first-party UI toolkit with no third-party dependencies at all. They map one to one at the file level and behave identically, because their behaviour comes from the same key collections rather than from either codebase.

02

A backbone of functions

Roughly a hundred small cloud functions, each owning one operation against one data shape, reached over signed endpoints. No application server, no framework runtime, and no shared library that every function must be redeployed to update.

03

One storage primitive

A managed key-value store with partition and sort keys chosen from the access patterns. Configuration, identity, content and telemetry all live in the same primitive, which keeps the number of storage technologies the team must understand at one.

04

Platform-native identity

Short-lived credentials issued per identity, with every client call signed. Authorisation is enforced by the cloud platform itself, so there is no bespoke auth layer to audit and no session framework to keep patched.

05

An admin companion

A separate internal application for authoring key collections, managing identities and reading the transaction log. Its own surface is scoped by privilegekeys, and its vocabulary is built from a registry rather than hard-coded — so extending the system is a data edit, not a rebuild.

06

A refresh cycle

Clients fetch the merged key collections on a short interval and cache them locally. This single mechanism is what turns configuration into operation: one write, and the entire installed base is running the new behaviour within a cycle.

by the numbers

Indicative figures, not a benchmark.

Offered as a sense of scale from the reference implementation.

Third-party dependencies in the iOS clientnone
Third-party dependencies in the Android clientfirst-party UI toolkit only
Application serversnone
Backbone functions~100, single-purpose
Standard key collections5, plus bespoke
Time for a configuration change to reach every clientone refresh cycle
Releases required for a tier or limit changenone

Subscription tiers, metered features, messaging, contextual help, notifications, purchasing and a full back office — with no third-party framework anywhere in it.