|
| 1 | +// Licensed to the .NET Foundation under one or more agreements. |
| 2 | +// The .NET Foundation licenses this file to you under the MIT license. |
| 3 | + |
| 4 | +using System; |
| 5 | +using System.Collections.Generic; |
| 6 | +using System.Diagnostics; |
| 7 | +using System.Diagnostics.CodeAnalysis; |
| 8 | +using System.Diagnostics.Metrics; |
| 9 | +using System.IO; |
| 10 | +using System.Runtime.CompilerServices; |
| 11 | +using System.Threading; |
| 12 | +using System.Threading.Tasks; |
| 13 | +using Microsoft.Extensions.Logging; |
| 14 | +using Microsoft.Shared.Diagnostics; |
| 15 | + |
| 16 | +#pragma warning disable S3358 // Ternary operators should not be nested |
| 17 | +#pragma warning disable SA1111 // Closing parenthesis should be on line of last parameter |
| 18 | +#pragma warning disable SA1113 // Comma should be on the same line as previous parameter |
| 19 | + |
| 20 | +namespace Microsoft.Extensions.AI; |
| 21 | + |
| 22 | +/// <summary>Represents a delegating speech-to-text client that implements the OpenTelemetry Semantic Conventions for Generative AI systems.</summary> |
| 23 | +/// <remarks> |
| 24 | +/// This class provides an implementation of the Semantic Conventions for Generative AI systems v1.37, defined at <see href="https://opentelemetry.io/docs/specs/semconv/gen-ai/" />. |
| 25 | +/// The specification is still experimental and subject to change; as such, the telemetry output by this client is also subject to change. |
| 26 | +/// </remarks> |
| 27 | +[Experimental("MEAI001")] |
| 28 | +public sealed class OpenTelemetrySpeechToTextClient : DelegatingSpeechToTextClient |
| 29 | +{ |
| 30 | + private readonly ActivitySource _activitySource; |
| 31 | + private readonly Meter _meter; |
| 32 | + |
| 33 | + private readonly Histogram<int> _tokenUsageHistogram; |
| 34 | + private readonly Histogram<double> _operationDurationHistogram; |
| 35 | + |
| 36 | + private readonly string? _defaultModelId; |
| 37 | + private readonly string? _providerName; |
| 38 | + private readonly string? _serverAddress; |
| 39 | + private readonly int _serverPort; |
| 40 | + |
| 41 | + /// <summary>Initializes a new instance of the <see cref="OpenTelemetrySpeechToTextClient"/> class.</summary> |
| 42 | + /// <param name="innerClient">The underlying <see cref="ISpeechToTextClient"/>.</param> |
| 43 | + /// <param name="logger">The <see cref="ILogger"/> to use for emitting any logging data from the client.</param> |
| 44 | + /// <param name="sourceName">An optional source name that will be used on the telemetry data.</param> |
| 45 | +#pragma warning disable IDE0060 // Remove unused parameter; it exists for consistency with IChatClient and future use |
| 46 | + public OpenTelemetrySpeechToTextClient(ISpeechToTextClient innerClient, ILogger? logger = null, string? sourceName = null) |
| 47 | +#pragma warning restore IDE0060 |
| 48 | + : base(innerClient) |
| 49 | + { |
| 50 | + Debug.Assert(innerClient is not null, "Should have been validated by the base ctor"); |
| 51 | + |
| 52 | + if (innerClient!.GetService<SpeechToTextClientMetadata>() is SpeechToTextClientMetadata metadata) |
| 53 | + { |
| 54 | + _defaultModelId = metadata.DefaultModelId; |
| 55 | + _providerName = metadata.ProviderName; |
| 56 | + _serverAddress = metadata.ProviderUri?.Host; |
| 57 | + _serverPort = metadata.ProviderUri?.Port ?? 0; |
| 58 | + } |
| 59 | + |
| 60 | + string name = string.IsNullOrEmpty(sourceName) ? OpenTelemetryConsts.DefaultSourceName : sourceName!; |
| 61 | + _activitySource = new(name); |
| 62 | + _meter = new(name); |
| 63 | + |
| 64 | + _tokenUsageHistogram = _meter.CreateHistogram<int>( |
| 65 | + OpenTelemetryConsts.GenAI.Client.TokenUsage.Name, |
| 66 | + OpenTelemetryConsts.TokensUnit, |
| 67 | + OpenTelemetryConsts.GenAI.Client.TokenUsage.Description |
| 68 | +#if NET9_0_OR_GREATER |
| 69 | + , advice: new() { HistogramBucketBoundaries = OpenTelemetryConsts.GenAI.Client.TokenUsage.ExplicitBucketBoundaries } |
| 70 | +#endif |
| 71 | + ); |
| 72 | + |
| 73 | + _operationDurationHistogram = _meter.CreateHistogram<double>( |
| 74 | + OpenTelemetryConsts.GenAI.Client.OperationDuration.Name, |
| 75 | + OpenTelemetryConsts.SecondsUnit, |
| 76 | + OpenTelemetryConsts.GenAI.Client.OperationDuration.Description |
| 77 | +#if NET9_0_OR_GREATER |
| 78 | + , advice: new() { HistogramBucketBoundaries = OpenTelemetryConsts.GenAI.Client.OperationDuration.ExplicitBucketBoundaries } |
| 79 | +#endif |
| 80 | + ); |
| 81 | + } |
| 82 | + |
| 83 | + /// <inheritdoc/> |
| 84 | + protected override void Dispose(bool disposing) |
| 85 | + { |
| 86 | + if (disposing) |
| 87 | + { |
| 88 | + _activitySource.Dispose(); |
| 89 | + _meter.Dispose(); |
| 90 | + } |
| 91 | + |
| 92 | + base.Dispose(disposing); |
| 93 | + } |
| 94 | + |
| 95 | + /// <summary> |
| 96 | + /// Gets or sets a value indicating whether potentially sensitive information should be included in telemetry. |
| 97 | + /// </summary> |
| 98 | + /// <value> |
| 99 | + /// <see langword="true"/> if potentially sensitive information should be included in telemetry; |
| 100 | + /// <see langword="false"/> if telemetry shouldn't include raw inputs and outputs. |
| 101 | + /// The default value is <see langword="false"/>, unless the <c>OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT</c> |
| 102 | + /// environment variable is set to "true" (case-insensitive). |
| 103 | + /// </value> |
| 104 | + /// <remarks> |
| 105 | + /// By default, telemetry includes metadata, such as token counts, but not raw inputs |
| 106 | + /// and outputs, such as message content, function call arguments, and function call results. |
| 107 | + /// The default value can be overridden by setting the <c>OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT</c> |
| 108 | + /// environment variable to "true". Explicitly setting this property will override the environment variable. |
| 109 | + /// </remarks> |
| 110 | + public bool EnableSensitiveData { get; set; } = TelemetryHelpers.EnableSensitiveDataDefault; |
| 111 | + |
| 112 | + /// <inheritdoc/> |
| 113 | + public override object? GetService(Type serviceType, object? serviceKey = null) => |
| 114 | + serviceType == typeof(ActivitySource) ? _activitySource : |
| 115 | + base.GetService(serviceType, serviceKey); |
| 116 | + |
| 117 | + /// <inheritdoc/> |
| 118 | + public override async Task<SpeechToTextResponse> GetTextAsync(Stream audioSpeechStream, SpeechToTextOptions? options = null, CancellationToken cancellationToken = default) |
| 119 | + { |
| 120 | + _ = Throw.IfNull(audioSpeechStream); |
| 121 | + |
| 122 | + using Activity? activity = CreateAndConfigureActivity(options); |
| 123 | + Stopwatch? stopwatch = _operationDurationHistogram.Enabled ? Stopwatch.StartNew() : null; |
| 124 | + string? requestModelId = options?.ModelId ?? _defaultModelId; |
| 125 | + |
| 126 | + SpeechToTextResponse? response = null; |
| 127 | + Exception? error = null; |
| 128 | + try |
| 129 | + { |
| 130 | + response = await base.GetTextAsync(audioSpeechStream, options, cancellationToken); |
| 131 | + return response; |
| 132 | + } |
| 133 | + catch (Exception ex) |
| 134 | + { |
| 135 | + error = ex; |
| 136 | + throw; |
| 137 | + } |
| 138 | + finally |
| 139 | + { |
| 140 | + TraceResponse(activity, requestModelId, response, error, stopwatch); |
| 141 | + } |
| 142 | + } |
| 143 | + |
| 144 | + /// <inheritdoc/> |
| 145 | + public override async IAsyncEnumerable<SpeechToTextResponseUpdate> GetStreamingTextAsync( |
| 146 | + Stream audioSpeechStream, SpeechToTextOptions? options = null, [EnumeratorCancellation] CancellationToken cancellationToken = default) |
| 147 | + { |
| 148 | + _ = Throw.IfNull(audioSpeechStream); |
| 149 | + |
| 150 | + using Activity? activity = CreateAndConfigureActivity(options); |
| 151 | + Stopwatch? stopwatch = _operationDurationHistogram.Enabled ? Stopwatch.StartNew() : null; |
| 152 | + string? requestModelId = options?.ModelId ?? _defaultModelId; |
| 153 | + |
| 154 | + IAsyncEnumerable<SpeechToTextResponseUpdate> updates; |
| 155 | + try |
| 156 | + { |
| 157 | + updates = base.GetStreamingTextAsync(audioSpeechStream, options, cancellationToken); |
| 158 | + } |
| 159 | + catch (Exception ex) |
| 160 | + { |
| 161 | + TraceResponse(activity, requestModelId, response: null, ex, stopwatch); |
| 162 | + throw; |
| 163 | + } |
| 164 | + |
| 165 | + var responseEnumerator = updates.GetAsyncEnumerator(cancellationToken); |
| 166 | + List<SpeechToTextResponseUpdate> trackedUpdates = []; |
| 167 | + Exception? error = null; |
| 168 | + try |
| 169 | + { |
| 170 | + while (true) |
| 171 | + { |
| 172 | + SpeechToTextResponseUpdate update; |
| 173 | + try |
| 174 | + { |
| 175 | + if (!await responseEnumerator.MoveNextAsync()) |
| 176 | + { |
| 177 | + break; |
| 178 | + } |
| 179 | + |
| 180 | + update = responseEnumerator.Current; |
| 181 | + } |
| 182 | + catch (Exception ex) |
| 183 | + { |
| 184 | + error = ex; |
| 185 | + throw; |
| 186 | + } |
| 187 | + |
| 188 | + trackedUpdates.Add(update); |
| 189 | + yield return update; |
| 190 | + Activity.Current = activity; // workaround for https://github.com/dotnet/runtime/issues/47802 |
| 191 | + } |
| 192 | + } |
| 193 | + finally |
| 194 | + { |
| 195 | + TraceResponse(activity, requestModelId, trackedUpdates.ToSpeechToTextResponse(), error, stopwatch); |
| 196 | + |
| 197 | + await responseEnumerator.DisposeAsync(); |
| 198 | + } |
| 199 | + } |
| 200 | + |
| 201 | + /// <summary>Creates an activity for a speech-to-text request, or returns <see langword="null"/> if not enabled.</summary> |
| 202 | + private Activity? CreateAndConfigureActivity(SpeechToTextOptions? options) |
| 203 | + { |
| 204 | + Activity? activity = null; |
| 205 | + if (_activitySource.HasListeners()) |
| 206 | + { |
| 207 | + string? modelId = options?.ModelId ?? _defaultModelId; |
| 208 | + |
| 209 | + activity = _activitySource.StartActivity( |
| 210 | + string.IsNullOrWhiteSpace(modelId) ? OpenTelemetryConsts.GenAI.GenerateContentName : $"{OpenTelemetryConsts.GenAI.GenerateContentName} {modelId}", |
| 211 | + ActivityKind.Client); |
| 212 | + |
| 213 | + if (activity is { IsAllDataRequested: true }) |
| 214 | + { |
| 215 | + _ = activity |
| 216 | + .AddTag(OpenTelemetryConsts.GenAI.Operation.Name, OpenTelemetryConsts.GenAI.GenerateContentName) |
| 217 | + .AddTag(OpenTelemetryConsts.GenAI.Request.Model, modelId) |
| 218 | + .AddTag(OpenTelemetryConsts.GenAI.Provider.Name, _providerName) |
| 219 | + .AddTag(OpenTelemetryConsts.GenAI.Output.Type, OpenTelemetryConsts.TypeText); |
| 220 | + |
| 221 | + if (_serverAddress is not null) |
| 222 | + { |
| 223 | + _ = activity |
| 224 | + .AddTag(OpenTelemetryConsts.Server.Address, _serverAddress) |
| 225 | + .AddTag(OpenTelemetryConsts.Server.Port, _serverPort); |
| 226 | + } |
| 227 | + |
| 228 | + if (options is not null) |
| 229 | + { |
| 230 | + if (EnableSensitiveData) |
| 231 | + { |
| 232 | + // Log all additional request options as raw values on the span. |
| 233 | + // Since AdditionalProperties has undefined meaning, we treat it as potentially sensitive data. |
| 234 | + if (options.AdditionalProperties is { } props) |
| 235 | + { |
| 236 | + foreach (KeyValuePair<string, object?> prop in props) |
| 237 | + { |
| 238 | + _ = activity.AddTag(prop.Key, prop.Value); |
| 239 | + } |
| 240 | + } |
| 241 | + } |
| 242 | + } |
| 243 | + } |
| 244 | + } |
| 245 | + |
| 246 | + return activity; |
| 247 | + } |
| 248 | + |
| 249 | + /// <summary>Adds speech-to-text response information to the activity.</summary> |
| 250 | + private void TraceResponse( |
| 251 | + Activity? activity, |
| 252 | + string? requestModelId, |
| 253 | + SpeechToTextResponse? response, |
| 254 | + Exception? error, |
| 255 | + Stopwatch? stopwatch) |
| 256 | + { |
| 257 | + if (_operationDurationHistogram.Enabled && stopwatch is not null) |
| 258 | + { |
| 259 | + TagList tags = default; |
| 260 | + |
| 261 | + AddMetricTags(ref tags, requestModelId, response); |
| 262 | + if (error is not null) |
| 263 | + { |
| 264 | + tags.Add(OpenTelemetryConsts.Error.Type, error.GetType().FullName); |
| 265 | + } |
| 266 | + |
| 267 | + _operationDurationHistogram.Record(stopwatch.Elapsed.TotalSeconds, tags); |
| 268 | + } |
| 269 | + |
| 270 | + if (_tokenUsageHistogram.Enabled && response?.Usage is { } usage) |
| 271 | + { |
| 272 | + if (usage.InputTokenCount is long inputTokens) |
| 273 | + { |
| 274 | + TagList tags = default; |
| 275 | + tags.Add(OpenTelemetryConsts.GenAI.Token.Type, OpenTelemetryConsts.TokenTypeInput); |
| 276 | + AddMetricTags(ref tags, requestModelId, response); |
| 277 | + _tokenUsageHistogram.Record((int)inputTokens, tags); |
| 278 | + } |
| 279 | + |
| 280 | + if (usage.OutputTokenCount is long outputTokens) |
| 281 | + { |
| 282 | + TagList tags = default; |
| 283 | + tags.Add(OpenTelemetryConsts.GenAI.Token.Type, OpenTelemetryConsts.TokenTypeOutput); |
| 284 | + AddMetricTags(ref tags, requestModelId, response); |
| 285 | + _tokenUsageHistogram.Record((int)outputTokens, tags); |
| 286 | + } |
| 287 | + } |
| 288 | + |
| 289 | + if (error is not null) |
| 290 | + { |
| 291 | + _ = activity? |
| 292 | + .AddTag(OpenTelemetryConsts.Error.Type, error.GetType().FullName) |
| 293 | + .SetStatus(ActivityStatusCode.Error, error.Message); |
| 294 | + } |
| 295 | + |
| 296 | + if (response is not null) |
| 297 | + { |
| 298 | + AddOutputMessagesTags(response, activity); |
| 299 | + |
| 300 | + if (activity is not null) |
| 301 | + { |
| 302 | + if (!string.IsNullOrWhiteSpace(response.ResponseId)) |
| 303 | + { |
| 304 | + _ = activity.AddTag(OpenTelemetryConsts.GenAI.Response.Id, response.ResponseId); |
| 305 | + } |
| 306 | + |
| 307 | + if (response.ModelId is not null) |
| 308 | + { |
| 309 | + _ = activity.AddTag(OpenTelemetryConsts.GenAI.Response.Model, response.ModelId); |
| 310 | + } |
| 311 | + |
| 312 | + if (response.Usage?.InputTokenCount is long inputTokens) |
| 313 | + { |
| 314 | + _ = activity.AddTag(OpenTelemetryConsts.GenAI.Usage.InputTokens, (int)inputTokens); |
| 315 | + } |
| 316 | + |
| 317 | + if (response.Usage?.OutputTokenCount is long outputTokens) |
| 318 | + { |
| 319 | + _ = activity.AddTag(OpenTelemetryConsts.GenAI.Usage.OutputTokens, (int)outputTokens); |
| 320 | + } |
| 321 | + |
| 322 | + // Log all additional response properties as raw values on the span. |
| 323 | + // Since AdditionalProperties has undefined meaning, we treat it as potentially sensitive data. |
| 324 | + if (EnableSensitiveData && response.AdditionalProperties is { } props) |
| 325 | + { |
| 326 | + foreach (KeyValuePair<string, object?> prop in props) |
| 327 | + { |
| 328 | + _ = activity.AddTag(prop.Key, prop.Value); |
| 329 | + } |
| 330 | + } |
| 331 | + } |
| 332 | + } |
| 333 | + |
| 334 | + void AddMetricTags(ref TagList tags, string? requestModelId, SpeechToTextResponse? response) |
| 335 | + { |
| 336 | + tags.Add(OpenTelemetryConsts.GenAI.Operation.Name, OpenTelemetryConsts.GenAI.GenerateContentName); |
| 337 | + |
| 338 | + if (requestModelId is not null) |
| 339 | + { |
| 340 | + tags.Add(OpenTelemetryConsts.GenAI.Request.Model, requestModelId); |
| 341 | + } |
| 342 | + |
| 343 | + tags.Add(OpenTelemetryConsts.GenAI.Provider.Name, _providerName); |
| 344 | + |
| 345 | + if (_serverAddress is string endpointAddress) |
| 346 | + { |
| 347 | + tags.Add(OpenTelemetryConsts.Server.Address, endpointAddress); |
| 348 | + tags.Add(OpenTelemetryConsts.Server.Port, _serverPort); |
| 349 | + } |
| 350 | + |
| 351 | + if (response?.ModelId is string responseModel) |
| 352 | + { |
| 353 | + tags.Add(OpenTelemetryConsts.GenAI.Response.Model, responseModel); |
| 354 | + } |
| 355 | + } |
| 356 | + } |
| 357 | + |
| 358 | + private void AddOutputMessagesTags(SpeechToTextResponse response, Activity? activity) |
| 359 | + { |
| 360 | + if (EnableSensitiveData && activity is { IsAllDataRequested: true }) |
| 361 | + { |
| 362 | + _ = activity.AddTag( |
| 363 | + OpenTelemetryConsts.GenAI.Output.Messages, |
| 364 | + OpenTelemetryChatClient.SerializeChatMessages([new(ChatRole.Assistant, response.Contents)])); |
| 365 | + } |
| 366 | + } |
| 367 | +} |
0 commit comments