7.9 KiB
Implemented architecture
Application shape
AutoForge is an in-process modular monolith with exactly one application executable: the Windows-first MAUI Blazor Hybrid Client. The application uses Razor component libraries for the complete routed UI and generic shared controls. There is no web/server host, WASM host or backend process.
flowchart TD
Client[AutoForge.Client: MAUI startup, BlazorWebView and DI] --> UI[AutoForge.UI: routes and feature presentation]
UI --> Components[AutoForge.Components: reusable generic UI]
UI --> Application[AutoForge.Application: cohesive feature slices]
Client --> Application
Application --> Core[AutoForge.Core: platform-neutral contracts]
UnitTests[AutoForge.UnitTests] --> Components
UnitTests --> Core
IntegrationTests[AutoForge.IntegrationTests] --> UI
IntegrationTests --> Application
IntegrationTests --> Core
Projects and dependency boundaries
All seven projects are in AutoForge.slnx:
| Project | Target and responsibility | Project references |
|---|---|---|
| Core | net10.0; platform-neutral target and immutable vehicle-profile/module/capability/provenance invariants |
None |
| Application | net10.0; target-selection and vehicle-profile import/catalog/select/export feature slices |
Core |
| Components | net10.0 RCL; generic StatusNotice with accessible tone and safe text rendering |
None |
| UI | net10.0 RCL; all application routes, navigation, layout and presentation |
Application, Components |
| Client | net10.0-windows10.0.26100.0; only MAUI Blazor Hybrid application executable |
Application, UI |
| UnitTests | net10.0; descriptor/resolver, rendered component, profile-domain and strict JSON tests |
Application, Components, Core |
| IntegrationTests | net10.0; real DI, synthetic profile fixture, rendered RCL and actual project-graph tests |
Application, Components, Core, UI |
Components has no AutoForge project dependency. UI never references Client. Core and Application never reference UI/Components/Client or tests. There are no cycles or production-to-test edges. Tests render UI directly and never reference or launch Client.
The Client contains only MAUI lifecycle/startup, its Windows platform entry
point, the BlazorWebView, host HTML/CSS and in-process composition. It registers
an empty target metadata set deliberately, so the real selection use case reports
not-registered until an explicit future composition supplies real metadata.
Vehicle profiles use a distinct per-process in-memory catalog; importing one does
not register or alter an execution target. No fake connected/available provider
is added.
UI ownership and real routes
All Razor routes and layouts are owned by the UI RCL; Routes scans the UI
assembly, not the native Client assembly.
| Route | UI owner | Implemented behavior |
|---|---|---|
/ |
Features/Foundation/Home.razor |
Describes the current foundation and accurately distinguishes implemented from absent features |
/execution-targets |
Features/ExecutionTargets/SelectExecutionTarget/ExecutionTargetSelectionPage.razor |
Requires explicit provider identifier and target kind, calls the real injected application resolver and displays success/errors |
/vehicle-profiles |
Features/Vehicles/Profiles/VehicleProfilesPage.razor |
Explicitly imports schema-v1 JSON, lists profile revisions, requires exact selection, renders metadata/provenance/unknowns and exposes read-only JSON export |
| Any unmatched route | Features/Foundation/NotFoundPage.razor via UI Routes.razor |
Presents an explicit not-found page and supported navigation |
Layout/MainLayout.razor supplies main navigation, shared layout and a permanent
metadata-only limitation statement. The generic
StatusNotice is shared by the
home, target-selection, profile and not-found pages; isolated CSS bundles from both RCLs
are imported through the Client scoped-style bundle.
The target-selection page's feature-local ExecutionTargetSelectionViewModel
adapts explicit input to the existing Application resolver. It contains only
presentation state/input validation and expected error display; it does not
reimplement provider resolution. It receives the actual IExecutionTargetResolver
through DI. The same AddAutoForgeTargetSelection registration used in the
headless integration tests is composed by Client.
Application feature organization
The accepted target-selection use case is grouped under
Application/Features/ExecutionTargets/SelectExecutionTarget/; its original
public namespace and API remain unchanged. Core still owns
Simulation, RecordedReplay, PhysicalHardware and immutable target descriptor
invariants. Application snapshots registrations, rejects duplicate IDs using
ordinal case-insensitive comparison, resolves only exact requested metadata and
returns the registered descriptor. Existing typed failures and no-fallback
behavior remain unchanged.
The vehicle profile slice is organized under
Application/Features/Vehicles/Profiles/. Core owns immutable profile identity
and revision, vehicle facts, arbitrary module roles, operation declarations,
capability states and provenance context. Application owns strict schema-v1
JSON interchange plus an in-memory catalog with explicit exact
(profileIdentifier, revision) selection. An import validates fully before it
registers; duplicate identities do not replace data. The Client catalog is
process-local and is not persistent.
The profile UI invokes the real import/list/select/export service through DI and
never auto-selects an imported/latest profile. Unknown optional facts remain
explicit null. Capability states and provenance are declarations only. Detailed
field, identifier, applicability and unknown-property rules are in the
schema-v1 guide and
ADR-0003.
New use cases should be cohesive and independently testable slices. Each slice owns its request/coordination and relevant application contract without forcing unrelated projects to depend on Application. Domain engines/protocols remain platform-neutral and can be extracted only when a real boundary warrants it. Do not add mediator boilerplate, generic manager/repository layers or empty future slices to demonstrate architecture.
Honest limitations and safety boundaries
Current work implements application-shell/navigation behavior, accessible status
presentation, explicit execution-target metadata selection and synthetic,
in-memory vehicle/module profile metadata interchange and selection. The
profiles are not vehicle discovery or compatibility data. The product still does
not implement connection/discovery, simulation, diagnostics, telemetry, replay,
backups, calibration values/editing, persistent storage, AI, cloud services,
operation execution or hardware integration. A SUPPORTED declaration, a
VERIFIED provenance label, physical-kind metadata or a calibration reference
does not implement or authorize an operation.
Future simulation remains deterministic and AI-independent with an isolated verification oracle. Future initial hardware integration is read-only. Physical programming requires separate operation-specific authorization and evidence. These are architectural requirements, not current app behavior.
The original ADR-0001 preserves valid foundation decisions but marks its dual-shell/deferred-host decision superseded. ADR-0002 is the current host and feature-slice decision.