Alpha Version: You are viewing the ALPHA documentation. This is an experimental version and may contain breaking changes.
Skip to main content

AI-Generated Walkthrough

This walkthrough shows what happens when you tell an AI assistant:

"Create an online shop with a Catalog plugin (DCB) and an Ordering plugin (DCB). The Catalog has Products and Categories. Ordering has Customers and Orders. Catalog exposes product availability to Ordering. When an order is placed, ship it automatically and send a confirmation email."

What the AI Does​

Phase 1: Requirements Analysis​

The AI identifies:

  • 2 plugins: Catalog, Ordering
  • 4 entities: Product, Category (Catalog); Customer, Order (Ordering)
  • Architecture: DCB for both (stated preference)
  • Extension points: Catalog.Products → Ordering subscribes
  • Automation: AutoShipOrder (trigger: OrderPlaced, resolution: OrderShipped)
  • Outbound translation: SendOrderConfirmation (email on OrderPlaced)

Phase 2: Architecture Decision​

For each entity, the AI evaluates against the decision guide:

EntityCross-Entity?Sync?Automation?Decision
ProductNoNoNoDCB (user preference)
CategoryNoNoNoDCB (user preference)
CustomerNoNoNoDCB
OrderYes (checks products)NoYes (auto-ship)DCB
CatalogProduct (Ordering)—Yes (from Catalog EP)—DCB (sync entity)

Phase 3: Code Generation​

The AI generates 30+ files across 5 packages:

catalog-spec/ — 1 file: Products_ExtensionPoint.res

ordering-spec/ — 1 file: Orders_ExtensionPoint.res

catalog/ — each component is split into a spec file (@@reventless.spec) and a body file (_Behavior.res / _Projection.res):

  • StateChangeSlices: AddProduct, ChangeProductName, ChangeProductDescription, ChangeProductPrice, AddCategory, RenameCategory, ArchiveCategory (each + _Behavior.res)
  • StateViewSlices: Products, Categories (each + _Projection.res)
  • ExtensionPoint: ExtensionPoint/Products_ExtensionPointMapping.res
  • Extension: Extension/Orders_Extension.res (subscribes to Ordering's EP)
  • Plugin.res (generated — "AUTO-GENERATED — do not edit")

ordering/ — same spec + body split:

  • StateChangeSlices: RegisterCustomer, ChangeEmail, ChangeAddress, DeactivateCustomer, PlaceOrder, ShipOrder, CancelOrder, SyncCatalogProduct (each + _Behavior.res)
  • StateViewSlices: Customers, Orders, AvailableProducts (each + _Projection.res)
  • AutomationSlice: AutoShipOrder (+ _Automation.res)
  • OutboundTranslationSlice: SendOrderConfirmation (+ _Translation.res)
  • Extension: Extension/Products_Extension.res (subscribes to Catalog's EP)
  • ExtensionPoint: ExtensionPoint/Orders_ExtensionPointMapping.res
  • Plugin.res (generated)

platform-local/ — 1 file: src/Main.res

Phase 4: Build and Verify​

pnpm install
pnpm run build   # zero warnings
pnpm test        # all tests pass

The build runs generate-plugin src/ (via each plugin's prebuild script) to produce the Plugin.res composition root from the discovered components, then compiles with ReScript.

The Result​

Running node src/Main.res.mjs from the platform-local package starts:

  • The domain GraphQL API on port 4000 — every command and query
  • The domain MCP server on port 3001 — the same surface for AI assistants
  • The platform admin API on port 4001, with its MCP server on 3002

The generated code is identical in structure to the hand-written DCB-based example in this documentation — the assistant is not producing a different kind of application, only producing the same kind faster.

Doing it yourself​

This page describes the output. For the workflow — the slash commands, what to say to get a good result, and how to grow an existing project — see AI-assisted development.