Skip to content

README.md

Latest commit

 

History

History
194 lines (145 loc) · 8.89 KB

README.md

File metadata and controls

194 lines (145 loc) · 8.89 KB

FeedbackExtractor.Client

Description

FeedbackExtractor.Client is a WPF desktop application targeting .NET 8 that provides a graphical user interface for extracting session feedback from document images or PDF files.

The application allows the user to:

  1. Select an extractor implementation from a drop-down list.
  2. Open a source document (image or PDF) from the local file system.
  3. Trigger the extraction and display the resulting SessionFeedback fields in the UI.

The application is built following the MVVM pattern using CommunityToolkit.Mvvm. Dependency Injection is configured at startup via Microsoft.Extensions.DependencyInjection, with all extractor implementations registered as keyed singletons against the IFeedbackExtractor interface.

Supported extractor implementations

ExtractorImplementations value Underlying implementation Accepted file types
OpenAI OpenAIFeedbackExtractor JPG, JPEG, PNG
DocumentIntelligence_Base DocumentFeedbackExtractor (prebuilt-layout) PDF, JPG, JPEG, PNG
DocumentIntelligence_Custom CustomFeedbackExtractor PDF, JPG, JPEG, PNG
ContentUnderstanding ContentUnderstandingFeedbackExtractor PDF, JPG, JPEG, PNG
Mixed_Models OrchestratorFeedbackExtractor JPG, JPEG, PNG
Mock MockFeedbackExtractor Any

Configuration

The application loads configuration from two JSON files at startup, in the following order:

  1. appsettings.local.json (required — contains environment-specific secrets)
  2. appsettings.json (required — contains shared defaults)

Values in appsettings.local.json override those in appsettings.json. The appsettings.local.json file is excluded from source control via .gitignore and must be created manually for each environment.

The full configuration template for appsettings.local.json is shown below. Each section maps directly to the corresponding extractor library's configuration. Only the sections for the extractors you intend to use need to be populated.

{
  "Logging": {
    "LogLevel": {
      "Default": "Information",
      "Microsoft.AspNetCore": "Warning"
    }
  },
  "DocumentIntelligenceDocument": {
    "Key": "<api-key>",
    "Endpoint": "<https://your-resource.cognitiveservices.azure.com/>",
    "MinConfidence": "0.5",
    "TenantId": "<azure-ad-tenant-id>",
    "ClientId": "<azure-ad-client-id>",
    "ClientSecret": "<azure-ad-client-secret>"
  },
  "DocumentIntelligenceCustom": {
    "Key": "<api-key>",
    "Endpoint": "<https://your-resource.cognitiveservices.azure.com/>",
    "ModelName": "<custom-model-name>",
    "MinConfidence": "0.5",
    "TenantId": "<azure-ad-tenant-id>",
    "ClientId": "<azure-ad-client-id>",
    "ClientSecret": "<azure-ad-client-secret>"
  },
  "OpenAI": {
    "Key": "<azure-openai-api-key>",
    "Endpoint": "<https://your-resource.openai.azure.com/>",
    "ModelName": "<deployment-name>",
    "ImageDetail": "high",
    "TenantId": "<azure-ad-tenant-id>",
    "ClientId": "<azure-ad-client-id>",
    "ClientSecret": "<azure-ad-client-secret>"
  },
  "ContentUnderstanding": {
    "Key": "<api-key>",
    "Endpoint": "<https://your-resource.services.ai.azure.com/>",
    "AnalyzerName": "<analyzer-name>",
    "ApiVersion": "2025-05-01-preview",
    "StorageEndpoint": "<https://youraccount.blob.core.windows.net/>",
    "StorageKey": "<storage-account-key>",
    "ContainerName": "<blob-container-name>",
    "SasExpiryMinutes": 15,
    "TenantId": "<azure-ad-tenant-id>",
    "ClientId": "<azure-ad-client-id>",
    "ClientSecret": "<azure-ad-client-secret>"
  }
}

For a detailed description of the individual configuration keys in each section, refer to the README of the corresponding library:

Security note: Never commit appsettings.local.json to source control. It is listed in .gitignore for this reason.

Main Classes

App

File: App.xaml.cs
Namespace: FeedbackExtractor.Client

Application entry point. Overrides OnStartup to:

  1. Build the IConfiguration by loading appsettings.local.json and then appsettings.json.
  2. Configure the IoC container (Ioc.Default) by registering all services:
    • All IFeedbackExtractor implementations as keyed singletons, keyed by ExtractorImplementations enum values.
    • IMessenger (using WeakReferenceMessenger.Default).
    • MainWindowViewModel as transient.
    • HttpClient factory and logging.

MainWindowViewModel

Namespace: FeedbackExtractor.Client.ViewModels
Visibility: public
Base class: ObservableRecipient (CommunityToolkit.Mvvm)

The sole ViewModel of the application. Exposes the following observable properties:

Property Type Description
ExtractionTypes ObservableCollection<ExtractorImplementations> The list of available extractor implementations shown in the UI.
SelectedExtractionType ExtractorImplementations? The extractor selected by the user. Enables/disables the SelectFileCommand.
DocumentFilePath string? The full path of the document selected by the user. Enables/disables the ExtractSessionFeedbackCommand.
ExtractedSessionFeedback SessionFeedback? The feedback result returned by the extractor, bound to the result panel.
IsExtractedSessionValid bool true when the extracted feedback passes SessionFeedback.IsValid().
IsBusy bool true while extraction is in progress. Disables both commands.

And the following commands:

Command Description
SelectFileCommand Opens a file picker dialog (via OpenFileDialogMessage) filtered to the file types accepted by the selected extractor. Disabled while busy or when no extractor is selected.
ExtractSessionFeedbackCommand Resolves the keyed IFeedbackExtractor from DI, opens the selected file, calls ExtractSessionFeedbackAsync, and updates ExtractedSessionFeedback and IsExtractedSessionValid. Disabled while busy or when no file is selected.

Messaging:

  • Sends OpenFileDialogMessage to request the View to open a file dialog.
  • Receives FileSelectedMessage to update DocumentFilePath when the user selects a file.

ExtractorImplementations

Namespace: FeedbackExtractor.Client.Entities
Visibility: public (enum)

Enumerates all available extractor implementations. Used as the DI key when registering and resolving IFeedbackExtractor instances, and as the item type in the extractor selection drop-down.

Value Description
Mock Returns hardcoded data; no Azure service required.
DocumentIntelligence_Base Uses the Azure AI Document Intelligence prebuilt-layout model.
DocumentIntelligence_Custom Uses a custom-trained Azure AI Document Intelligence model.
OpenAI Uses an Azure OpenAI vision model.
Mixed_Models Uses the orchestrator that combines multiple models.
ContentUnderstanding Uses Azure AI Content Understanding.

OpenFileDialogMessage

Namespace: FeedbackExtractor.Client.Messages
Visibility: public

A plain message object sent from the ViewModel to the View to request that a file-open dialog be displayed. Carries a Filter string (e.g., "Images (*.jpg;*.jpeg;*.png)|*.jpg;*.jpeg;*.png") that controls which file types are visible in the dialog.

FileSelectedMessage

Namespace: FeedbackExtractor.Client.Messages
Visibility: public
Base class: ValueChangedMessage<string> (CommunityToolkit.Mvvm)

A message sent from the View back to the ViewModel after the user selects a file in the dialog. Carries the full file path as its value.

Value Converters

All converters are located in the FeedbackExtractor.Client.Converters namespace and implement IValueConverter for use in XAML bindings.

Converter Conversion Typical use
BoolToOppositeBoolConverter bool → !bool Inverting IsBusy to enable/disable controls.
NotNullToBoolConverter object? → bool (null = false) Enabling a button only when a result is present.
NotNullToVisibilityConverter object? → Visibility (null = Hidden) Showing result panels only after extraction.
VisibilityToOppositeBoolConverter bool → Visibility (true = Hidden) Hiding elements when a condition is true.