Skip to content

Commit e0714e3

Browse files
GarrettBeattyGarrettBeatty
committed
Add live integration tests for the Polly and Transcribe speech clients
Integration tests belong with the code they cover, so the speech-clients PR now carries them. AWS.Speech.MEAI.IntegrationTests is a net8.0 project that runs against live AWS with the ambient credential and region chain. The tests always run; they are not gated on credentials, so a missing or unauthorized environment fails them rather than skipping. - Polly GetAudioAsync returns non-empty audio/lpcm. - Polly GetStreamingAudioAsync emits SessionOpen, audio chunks, SessionClose. - Transcribe transcribes Polly-synthesized speech: the phrase is synthesized to 16 kHz PCM, a second of trailing silence is appended so Transcribe endpoints and stabilizes a final result, and the recognized text is asserted to contain a distinctive word. This exercises both clients end to end with no checked-in audio fixture. Registered in AWS.DotNetAI.slnx. Also suppresses the xUnit1051 analyzer advisory in the unit-test project (those tests drive the loop with their own tokens). Verified: all three integration tests pass against live AWS in us-west-2.
1 parent 9dfd09e commit e0714e3

4 files changed

Lines changed: 139 additions & 2 deletions

File tree

‎AWS.DotNetAI.slnx‎

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -26,5 +26,6 @@
2626
<Project Path="test/AWS.AgentCore.Hosting.IntegrationTests/AWS.AgentCore.Hosting.IntegrationTests.csproj" />
2727
<Project Path="test/AWS.Bedrock.MEAI.UnitTests/AWS.Bedrock.MEAI.UnitTests.csproj" />
2828
<Project Path="test/AWS.Speech.MEAI.UnitTests/AWS.Speech.MEAI.UnitTests.csproj" />
29+
<Project Path="test/AWS.Speech.MEAI.IntegrationTests/AWS.Speech.MEAI.IntegrationTests.csproj" />
2930
</Folder>
3031
</Solution>

‎test/AWS.Speech.MEAI.IntegrationTests/AWS.Speech.MEAI.IntegrationTests.csproj‎

Lines changed: 26 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,26 @@
1+
<Project Sdk="Microsoft.NET.Sdk">
2+
3+
<PropertyGroup>
4+
<OutputType>Exe</OutputType>
5+
<!-- Amazon Transcribe streaming and the VoiceAgent loop are net8.0-only, so the integration
6+
tests target net8.0 and run against live AWS. -->
7+
<TargetFramework>net8.0</TargetFramework>
8+
<Nullable>enable</Nullable>
9+
<LangVersion>Latest</LangVersion>
10+
<IsPackable>false</IsPackable>
11+
<!-- MEAI001: uses evaluation-stage Microsoft.Extensions.AI APIs. -->
12+
<NoWarn>$(NoWarn);MEAI001</NoWarn>
13+
</PropertyGroup>
14+
15+
<ItemGroup>
16+
<PackageReference Include="Microsoft.Extensions.AI.Abstractions" Version="10.8.3" />
17+
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="18.5.1" />
18+
<PackageReference Include="xunit.v3" Version="3.2.2" />
19+
<PackageReference Include="xunit.runner.visualstudio" Version="3.1.5" />
20+
</ItemGroup>
21+
22+
<ItemGroup>
23+
<ProjectReference Include="..\..\src\AWS.Speech.MEAI\AWS.Speech.MEAI.csproj" />
24+
</ItemGroup>
25+
26+
</Project>

‎test/AWS.Speech.MEAI.IntegrationTests/SpeechClientsIntegrationTests.cs‎

Lines changed: 108 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,108 @@
1+
// Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.
2+
// SPDX-License-Identifier: Apache-2.0
3+
4+
using Amazon.Polly;
5+
using Amazon.TranscribeStreaming;
6+
using Microsoft.Extensions.AI;
7+
using System;
8+
using System.Collections.Generic;
9+
using System.IO;
10+
using System.Linq;
11+
using System.Threading;
12+
using System.Threading.Tasks;
13+
using Xunit;
14+
15+
namespace AWS.Speech.MEAI.IntegrationTests;
16+
17+
/// <summary>
18+
/// Live integration tests for the standalone speech clients. These call Amazon Polly and Amazon
19+
/// Transcribe with the ambient AWS credential and region chain and always run (they are not gated on
20+
/// credentials): a missing or unauthorized environment fails the test rather than skipping it.
21+
/// </summary>
22+
[Trait("Category", "Integration")]
23+
public class SpeechClientsIntegrationTests
24+
{
25+
private const string SpokenPhrase = "the quick brown fox jumps over the lazy dog";
26+
private static readonly string[] DistinctiveWords = { "quick", "brown", "fox", "lazy", "dog" };
27+
28+
[Fact]
29+
[Trait("IntegrationTest", "Speech")]
30+
public async Task Polly_GetAudioAsync_ReturnsPcmAudio()
31+
{
32+
using var polly = new AmazonPollyClient();
33+
var tts = polly.AsITextToSpeechClient();
34+
35+
var response = await tts.GetAudioAsync("Hello from Amazon Polly.", cancellationToken: TestContext.Current.CancellationToken);
36+
37+
var audio = Assert.IsType<DataContent>(Assert.Single(response.Contents));
38+
Assert.Equal("audio/lpcm", audio.MediaType);
39+
Assert.False(audio.Data.IsEmpty, "Amazon Polly returned no audio.");
40+
}
41+
42+
[Fact]
43+
[Trait("IntegrationTest", "Speech")]
44+
public async Task Polly_GetStreamingAudioAsync_EmitsSessionFramedChunks()
45+
{
46+
using var polly = new AmazonPollyClient();
47+
var tts = polly.AsITextToSpeechClient();
48+
49+
var kinds = new List<TextToSpeechResponseUpdateKind>();
50+
long audioBytes = 0;
51+
await foreach (var update in tts.GetStreamingAudioAsync("Streaming from Amazon Polly.")
52+
.WithCancellation(TestContext.Current.CancellationToken))
53+
{
54+
kinds.Add(update.Kind);
55+
foreach (var content in update.Contents.OfType<DataContent>())
56+
{
57+
audioBytes += content.Data.Length;
58+
}
59+
}
60+
61+
Assert.Equal(TextToSpeechResponseUpdateKind.SessionOpen, kinds.First());
62+
Assert.Equal(TextToSpeechResponseUpdateKind.SessionClose, kinds.Last());
63+
Assert.Contains(TextToSpeechResponseUpdateKind.AudioUpdating, kinds);
64+
Assert.True(audioBytes > 0, "Amazon Polly streamed no audio.");
65+
}
66+
67+
[Fact]
68+
[Trait("IntegrationTest", "Speech")]
69+
public async Task Transcribe_TranscribesPollyGeneratedSpeech()
70+
{
71+
// Synthesize the phrase with Amazon Polly (16 kHz mono PCM), then transcribe it with Amazon
72+
// Transcribe. This exercises both clients end to end without a checked-in audio fixture.
73+
using var polly = new AmazonPollyClient();
74+
var tts = polly.AsITextToSpeechClient(defaultSampleRateHertz: 16000);
75+
var synthesized = await tts.GetAudioAsync(SpokenPhrase, cancellationToken: TestContext.Current.CancellationToken);
76+
var speech = synthesized.Contents.OfType<DataContent>().Single().Data.ToArray();
77+
Assert.True(speech.Length > 0, "Amazon Polly returned no audio to transcribe.");
78+
79+
// Append ~1s of trailing silence (16 kHz, 16-bit mono => 32000 bytes) so Amazon Transcribe
80+
// detects end-of-utterance and stabilizes a final result rather than only partials.
81+
var pcm = new byte[speech.Length + 32000];
82+
Array.Copy(speech, pcm, speech.Length);
83+
84+
using var transcribe = new AmazonTranscribeStreamingClient();
85+
var stt = transcribe.AsISpeechToTextClient("en-US", 16000);
86+
87+
using var timeout = CancellationTokenSource.CreateLinkedTokenSource(TestContext.Current.CancellationToken);
88+
timeout.CancelAfter(TimeSpan.FromSeconds(60));
89+
90+
string? latestRecognized = null;
91+
bool sawFinal = false;
92+
using var audioStream = new MemoryStream(pcm);
93+
await foreach (var update in stt.GetStreamingTextAsync(audioStream, cancellationToken: timeout.Token))
94+
{
95+
if ((update.Kind == SpeechToTextResponseUpdateKind.TextUpdating ||
96+
update.Kind == SpeechToTextResponseUpdateKind.TextUpdated) && !string.IsNullOrEmpty(update.Text))
97+
{
98+
latestRecognized = update.Text; // successive results refine the same utterance
99+
sawFinal |= update.Kind == SpeechToTextResponseUpdateKind.TextUpdated;
100+
}
101+
}
102+
103+
Assert.False(string.IsNullOrWhiteSpace(latestRecognized), "Amazon Transcribe recognized no speech.");
104+
Assert.True(sawFinal, "Amazon Transcribe produced no final (TextUpdated) result.");
105+
var transcript = latestRecognized!.ToLowerInvariant();
106+
Assert.Contains(DistinctiveWords, word => transcript.Contains(word, StringComparison.Ordinal));
107+
}
108+
}

‎test/AWS.Speech.MEAI.UnitTests/AWS.Speech.MEAI.UnitTests.csproj‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -12,8 +12,10 @@
1212
<Nullable>enable</Nullable>
1313
<LangVersion>Latest</LangVersion>
1414
<IsPackable>false</IsPackable>
15-
<!-- MEAI001: uses evaluation-stage Microsoft.Extensions.AI APIs. -->
16-
<NoWarn>$(NoWarn);MEAI001</NoWarn>
15+
<!-- MEAI001: uses evaluation-stage Microsoft.Extensions.AI APIs.
16+
xUnit1051: these unit tests deliberately drive the loop with their own tokens (or none);
17+
they do not need TestContext.Current.CancellationToken threaded through every call. -->
18+
<NoWarn>$(NoWarn);MEAI001;xUnit1051</NoWarn>
1719
<SignAssembly>true</SignAssembly>
1820
<AssemblyOriginatorKeyFile>..\..\public.snk</AssemblyOriginatorKeyFile>
1921
</PropertyGroup>

0 commit comments

Comments
 (0)