update TranscriptionResponse output to list of words and segments ins…

Apps.OpenAI/Actions/AudioActions.cs

@@ -12,7 +12,6 @@
 using Blackbird.Applications.Sdk.Common.Invocation;
 using Blackbird.Applications.Sdk.Utils.Extensions.Files;
 using Blackbird.Applications.SDK.Extensions.FileManagement.Interfaces;
-using Newtonsoft.Json;
 using RestSharp;
 using System;
 using System.Collections.Generic;
@@ -77,14 +76,14 @@ public async Task<TranscriptionResponse> CreateTranscription(
         }
 
         var response = await UniversalClient.ExecuteWithErrorHandling<TranscriptionDto>(request);
-        var words = response.Words?.Select(x => new WordResponse(x)).ToList() ?? new List<WordResponse>();
-        var segments = response.Segments?.Select(x => new SegmentResponse(x)).ToList() ?? new List<SegmentResponse>();
+        var words = response.Words?.Select(x => new WordResponse(x)).ToList() ?? [];
+        var segments = response.Segments?.Select(x => new SegmentResponse(x)).ToList() ?? [];
 
         return new()
         {
             Transcription = BuildTranscription(response, isDiarizationModel),
-            Words = JsonConvert.SerializeObject(words),
-            Segments = JsonConvert.SerializeObject(segments)
+            Words = words,
+            Segments = segments
         };
 
         static string GetResponseFormat(string model) => model switch

Apps.OpenAI/Apps.OpenAI.csproj

@@ -4,7 +4,7 @@
     <TargetFramework>net8.0</TargetFramework>
 	  <Product>OpenAI</Product>
 	  <Description>Creating safe artificial general intelligence that benefits all of humanity</Description>
-	  <Version>2.8.42</Version>
+	  <Version>2.8.43</Version>
 	  <AssemblyName>Apps.OpenAI</AssemblyName>
   </PropertyGroup>
 

Apps.OpenAI/Models/Responses/Audio/TranscriptionResponse.cs

@@ -1,4 +1,5 @@
-using Apps.OpenAI.Dtos;
+using System.Collections.Generic;
+using Apps.OpenAI.Dtos;
 using Blackbird.Applications.Sdk.Common;
 
 namespace Apps.OpenAI.Models.Responses.Audio;
@@ -7,14 +8,14 @@ public class TranscriptionResponse
 {
     public string Transcription { get; set; }
 
-    [Display("Words (serialized)")]
-    public string Words { get; set; }
+    [Display("Words")]
+    public List<WordResponse> Words { get; set; }
 
-    [Display("Segments (serialized)")]
-    public string Segments { get; set; }
+    [Display("Segments")]
+    public List<SegmentResponse> Segments { get; set; }
 }
 
-public record WordResponse(WordDto dto)
+public class WordResponse(WordDto dto)
 {
     public string Word { get; set; } = dto.Word;
 
@@ -23,7 +24,7 @@ public record WordResponse(WordDto dto)
     public double End { get; set; } = dto.End;
 }
 
-public record SegmentResponse(SegmentDto dto)
+public class SegmentResponse(SegmentDto dto)
 {
     [Display("Segment ID")]
     public string Id { get; set; } = dto.Id;

Tests.OpenAI/AudioServiceTests.cs

@@ -7,35 +7,30 @@
 using Blackbird.Applications.Sdk.Common.Files;
 using Blackbird.Applications.Sdk.Common.Exceptions;
 using Blackbird.Applications.Sdk.Common.Invocation;
-using Newtonsoft.Json.Linq;
 
 namespace Tests.OpenAI;
 
 [TestClass]
 public class AudioServiceTests : TestBaseWithContext
 {
-    [TestMethod, ContextDataSource(ConnectionTypes.OpenAiEmbedded, ConnectionTypes.OpenAi)]
-    public async Task CreateTranscription_OpenAi_ReturnsTranscription_DiarizedJsonFormat(InvocationContext context)
+    [TestMethod, ContextDataSource(ConnectionTypes.OpenAi)]
+    public async Task CreateTranscription_OpenAi_ReturnsTranscription(InvocationContext context)
     {
         // Arrange
         var handler = new AudioActions(context, FileManagementClient);
         var model = new AudioModelIdentifier { ModelId = "gpt-4o-transcribe-diarize" };
         var request = new TranscriptionRequest
         {
-            File = new FileReference { Name = "Transcription sample short.mp3" },
-            Language = "pt",
+            File = new FileReference { Name = "tts delorean.mp3" },
+            Language = "en",
         };
 
         // Act
         var result = await handler.CreateTranscription(model, request);
-        var segments = JArray.Parse(result.Segments);
 
         // Assert
-        TestContext.WriteLine(result.Transcription);
-        TestContext.WriteLine(result.Segments);
+        PrintResult(result);
         Assert.IsNotNull(result);
-        Assert.IsTrue(segments.Count > 0);
-        Assert.IsTrue(segments.Any(x => x["Speaker"] != null));
     }
 
     [TestMethod, ContextDataSource(ConnectionTypes.OpenAiEmbedded, ConnectionTypes.OpenAi)]
@@ -75,7 +70,6 @@ public async Task CreateTranscription_OpenAi_StandardModel_ReturnsSingleBlobText
 
         // Assert
         TestContext.WriteLine(result.Transcription);
-        TestContext.WriteLine(result.Segments);
         Assert.IsNotNull(result);
         Assert.IsFalse(result.Transcription.Contains("A:"));
     }
@@ -96,7 +90,6 @@ public async Task CreateTranscription_AzureOpenAi_ReturnsTranscription(Invocatio
 
         // Assert
         TestContext.WriteLine(result.Transcription);
-        TestContext.WriteLine(result.Segments);
         Assert.IsNotNull(result);
     }