Skip to content

README.md

Latest commit

 

History

History
93 lines (74 loc) · 4.47 KB

README.md

File metadata and controls

93 lines (74 loc) · 4.47 KB
Raw

Garden Shop AI Chat

A polished .NET MAUI sample that demonstrates AI Extensions in a real app surface. The assistant, Sage, can browse the catalog, manage the cart, open modal pages, recommend starter bundles, and review or reorder past purchases using source-generated tools.

What to try

  • Add 5 packs of tomato seeds and a trowel
  • Build me a basil starter bundle
  • Show compact cart
  • Open the catalog
  • Go to my past orders
  • Re-order my last order

App behaviors

  • Responsive main surface — chat stays centered and readable; the cart shows as a sidebar on wider windows and moves behind a header button on narrower layouts.
  • Live tool inventory — the welcome screen renders cards from GardenShopTools.Default.Tools, so any new exported tool automatically appears there.
  • Modal navigation — catalog, cart, and orders open through Shell routes as animated modal overlays.
  • Approval flow — checkout and destructive actions pause the chat and show an inline approve/reject banner.

Tool sources and lifetimes

GardenShopTools composes several very different source types with repeated [AIToolSource] attributes — no hand-written wrapper classes required. The sample uses an explicit context on purpose to curate the exact set of tools Sage should see, even though the library can also auto-generate an assembly-wide context for the whole app.

Source type Lifetime What it contributes
ProductCatalog static Catalog browsing tools like list_all_products, search_products, and get_product
CurrentCart singleton Cart inspection and mutation tools like show_list, add_to_list, change_qty, and remove_from_list
IOrderArchive singleton interface Past-order lookup, checkout_list, reorder, and clear_past_orders
MainViewModel singleton UI navigation tools: navigate_to_page and dismiss_page
CartViewModel singleton Accessor-level tools: get_cart_mode / set_cart_mode
CatalogViewModel transient recommend_bundle, a page-local bundle recommender that returns a starter kit without mutating the cart

This sample is especially useful if you want to see a transient view-model participate in a shared tool context while still writing through to singleton state.

Tool scenarios

Area Tools
Catalog discovery list_all_products, search_products, get_product
Cart management show_list, add_to_list, change_qty, remove_from_list, cancel_list
Cart presentation get_cart_mode, set_cart_mode
Orders list_past_orders, find_order, checkout_list, reorder, clear_past_orders
Page navigation navigate_to_page, dismiss_page
Recommendations recommend_bundle

Feature showcase

Feature Where
[ExportAIFunction] on a static property Services/Catalog/ProductCatalog.cs → All / list_all_products
[ExportAIFunction] on a static method with an optional param ProductCatalog.SearchProducts
Custom tool names (method ≠ tool name) ProductCatalog.FindByName → get_product, CurrentCart.SetQuantity → change_qty
[ExportAIFunction] on a singleton DI service Services/Cart/CurrentCart.cs
[ExportAIFunction] on an interface Services/Order/IOrderArchive.cs
[FromServices] parameter injection IOrderArchive.Checkout([FromServices] CurrentCart cart)
Accessor-level property tools ViewModels/Cart/CartViewModel.cs → get_cart_mode / set_cart_mode
Transient tool host ViewModels/Catalog/CatalogViewModel.cs → recommend_bundle
Shell modal navigation tools ViewModels/MainViewModel.cs + AppShell.xaml.cs
Responsive welcome cards and centered chat layout Views/ChatView.xaml + Pages/MainPage.xaml

Approval flow

checkout_list, cancel_list, and clear_past_orders carry [ExportAIFunction(ApprovalRequired = true)]. When the model requests one of those actions, the input bar is replaced by an approval banner until you accept or reject it.

Build & run

dotnet build samples/AIExtensions.Sample.Garden -f net10.0-maccatalyst

Configure user secrets (shared across AI Extensions samples):

dotnet user-secrets --id ai-attributes-secrets set "AI:Endpoint" "<your-endpoint>"
dotnet user-secrets --id ai-attributes-secrets set "AI:ApiKey" "<your-key>"
dotnet user-secrets --id ai-attributes-secrets set "AI:DeploymentName" "<your-deployment>"