@ai-sdk/otel 1.0.123 → 1.0.126

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ai-sdk/otel",
-  "version": "1.0.123",
+  "version": "1.0.126",
   "type": "module",
   "license": "Apache-2.0",
   "sideEffects": false,
@@ -26,12 +26,12 @@
     }
   },
   "dependencies": {
-    "@ai-sdk/provider": "4.0.20",
+    "@ai-sdk/provider": "4.0.21",
     "@opentelemetry/api": "1.9.1",
-    "ai": "7.0.123"
+    "ai": "7.0.126"
   },
   "devDependencies": {
-    "@ai-sdk/provider-utils": "5.0.52",
+    "@ai-sdk/provider-utils": "5.0.53",
     "@opentelemetry/sdk-trace-base": "2.7.1",
     "@types/node": "22.19.19",
     "@vercel/ai-tsconfig": "0.0.0",

package/src/gen-ai-format-messages.ts

@@ -120,6 +120,9 @@ export function mapOperationName(operationId: string): string {
     'ai.embed': 'embeddings',
     'ai.embedMany': 'embeddings',
     'ai.rerank': 'rerank',
+    'ai.generateSpeech': 'ai.generateSpeech',
+    'ai.transcribe': 'ai.transcribe',
+    'ai.streamTranscribe': 'ai.streamTranscribe',
   };
   return mapping[operationId] ?? operationId;
 }

package/src/legacy-open-telemetry.ts

@@ -24,6 +24,8 @@ import type {
   GenerateTextStartEvent,
   GenerateTextStepEndEvent,
   GenerateTextStepStartEvent,
+  GenerateSpeechEndEvent,
+  GenerateSpeechStartEvent,
   ToolExecutionEndEvent,
   ToolExecutionStartEvent,
   Experimental_EvaluateEndEvent as EvaluateEndEvent,
@@ -35,6 +37,10 @@ import type {
   RerankEndEvent,
   RerankStartEvent,
   RerankingModelCallStartEvent,
+  TranscriptionEndEvent,
+  TranscriptionStartEvent,
+  Experimental_StreamTranscriptionEndEvent as StreamTranscriptionEndEvent,
+  Experimental_StreamTranscriptionStartEvent as StreamTranscriptionStartEvent,
   InferTelemetryEvent,
   Telemetry,
   TelemetryOptions,
@@ -42,6 +48,7 @@ import type {
 } from 'ai';
 import { assembleOperationName } from './assemble-operation-name';
 import { getBaseTelemetryAttributes } from './get-base-telemetry-attributes';
+import { getProviderUsageAttributes } from './provider-usage-attributes';
 import { sanitizeAttributeValue } from './sanitize-attribute-value';
 import { stringifyForTelemetry } from './stringify-for-telemetry';
 
@@ -213,8 +220,24 @@ export class LegacyOpenTelemetry implements Telemetry {
       | InferTelemetryEvent<GenerateTextStartEvent>
       | InferTelemetryEvent<GenerateObjectStartEvent>
       | InferTelemetryEvent<EmbedStartEvent>
-      | InferTelemetryEvent<RerankStartEvent>,
+      | InferTelemetryEvent<RerankStartEvent>
+      | InferTelemetryEvent<GenerateSpeechStartEvent>
+      | InferTelemetryEvent<TranscriptionStartEvent>,
   ): void {
+    if (event.operationId === 'ai.generateSpeech') {
+      this.onAudioOperationStart(
+        event as InferTelemetryEvent<GenerateSpeechStartEvent>,
+      );
+      return;
+    }
+
+    if (event.operationId === 'ai.transcribe') {
+      this.onAudioOperationStart(
+        event as InferTelemetryEvent<TranscriptionStartEvent>,
+      );
+      return;
+    }
+
     if (
       event.operationId === 'ai.embed' ||
       event.operationId === 'ai.embedMany'
@@ -243,6 +266,69 @@ export class LegacyOpenTelemetry implements Telemetry {
     this.onGenerateStart(event as InferTelemetryEvent<GenerateTextStartEvent>);
   }
 
+  experimental_onStreamTranscriptionStart(
+    event: InferTelemetryEvent<StreamTranscriptionStartEvent>,
+  ): void {
+    this.onAudioOperationStart(event);
+  }
+
+  private onAudioOperationStart(
+    event: InferTelemetryEvent<
+      | GenerateSpeechStartEvent
+      | TranscriptionStartEvent
+      | StreamTranscriptionStartEvent
+    >,
+  ): void {
+    const telemetry: TelemetryOptions = {
+      recordInputs: event.recordInputs,
+      recordOutputs: event.recordOutputs,
+      functionId: event.functionId,
+    };
+    const audio = 'audio' in event ? event.audio : undefined;
+    const text = 'text' in event ? event.text : undefined;
+    const baseTelemetryAttributes: Attributes = {
+      'ai.model.provider': event.provider,
+      'ai.model.id': event.modelId,
+      ...Object.fromEntries(
+        Object.entries(event.headers ?? {}).map(([key, value]) => [
+          `ai.request.headers.${key}`,
+          value,
+        ]),
+      ),
+    };
+    const attributes = selectAttributes(telemetry, {
+      ...assembleOperationName({
+        operationId: event.operationId,
+        telemetry,
+      }),
+      ...baseTelemetryAttributes,
+      'ai.request.text': text == null ? undefined : { input: () => text },
+      'ai.request.audio.size':
+        audio?.byteLength == null
+          ? undefined
+          : { input: () => audio.byteLength },
+      'ai.request.audio.mediaType':
+        audio == null ? undefined : { input: () => audio.mediaType },
+    });
+
+    const rootSpan = this.tracer.startSpan(event.operationId, { attributes });
+    const rootContext = trace.setSpan(context.active(), rootSpan);
+    this.callStates.set(event.callId, {
+      operationId: event.operationId,
+      telemetry,
+      rootSpan,
+      rootContext,
+      stepSpan: undefined,
+      stepContext: undefined,
+      embedSpans: new Map(),
+      rerankSpan: undefined,
+      evaluationSpan: undefined,
+      toolSpans: new Map(),
+      baseTelemetryAttributes,
+      settings: {},
+    });
+  }
+
   private onGenerateStart(
     event: InferTelemetryEvent<GenerateTextStartEvent>,
   ): void {
@@ -793,7 +879,9 @@ export class LegacyOpenTelemetry implements Telemetry {
       | GenerateTextEndEvent<ToolSet>
       | GenerateObjectEndEvent<unknown>
       | EmbedEndEvent
-      | RerankEndEvent,
+      | RerankEndEvent
+      | GenerateSpeechEndEvent
+      | TranscriptionEndEvent,
   ): void {
     const state = this.getCallState(event.callId);
     if (!state?.rootSpan) return;
@@ -811,6 +899,16 @@ export class LegacyOpenTelemetry implements Telemetry {
       return;
     }
 
+    if (
+      state.operationId === 'ai.generateSpeech' ||
+      state.operationId === 'ai.transcribe'
+    ) {
+      this.onAudioOperationEnd(
+        event as GenerateSpeechEndEvent | TranscriptionEndEvent,
+      );
+      return;
+    }
+
     if (
       state.operationId === 'ai.generateObject' ||
       state.operationId === 'ai.streamObject'
@@ -822,6 +920,59 @@ export class LegacyOpenTelemetry implements Telemetry {
     this.onGenerateEnd(event as GenerateTextEndEvent<ToolSet>);
   }
 
+  experimental_onStreamTranscriptionEnd(
+    event: InferTelemetryEvent<StreamTranscriptionEndEvent>,
+  ): void {
+    this.onAudioOperationEnd(event);
+  }
+
+  private onAudioOperationEnd(
+    event:
+      | GenerateSpeechEndEvent
+      | TranscriptionEndEvent
+      | StreamTranscriptionEndEvent,
+  ): void {
+    const state = this.getCallState(event.callId);
+    if (!state?.rootSpan) return;
+
+    const isSpeech = event.operationId === 'ai.generateSpeech';
+    const audio = event.audio;
+    state.rootSpan.setAttributes(
+      selectAttributes(state.telemetry, {
+        'ai.request.audio.size': isSpeech
+          ? undefined
+          : { input: () => audio.byteLength },
+        'ai.request.audio.mediaType': isSpeech
+          ? undefined
+          : { input: () => audio.mediaType },
+        'ai.response.text':
+          'segments' in event ? { output: () => event.text } : undefined,
+        'ai.response.audio.size': isSpeech
+          ? { output: () => audio.byteLength }
+          : undefined,
+        'ai.response.audio.mediaType': isSpeech
+          ? { output: () => audio.mediaType }
+          : undefined,
+        'ai.response.audio.format':
+          isSpeech && 'format' in audio
+            ? { output: () => audio.format }
+            : undefined,
+        ...getProviderUsageAttributes({
+          usage: event.usage,
+          prefix: 'ai.usage',
+        }),
+        'ai.response.usage':
+          event.usage == null ? undefined : JSON.stringify(event.usage),
+        'ai.response.providerMetadata':
+          event.providerMetadata == null
+            ? undefined
+            : JSON.stringify(event.providerMetadata),
+      }),
+    );
+    state.rootSpan.end();
+    this.cleanupCallState(event.callId);
+  }
+
   private onGenerateEnd(event: GenerateTextEndEvent<ToolSet>): void {
     const state = this.getCallState(event.callId);
     if (!state?.rootSpan) return;

package/src/open-telemetry.ts

@@ -23,6 +23,8 @@ import type {
   GenerateTextStartEvent,
   GenerateTextStepEndEvent,
   GenerateTextStepStartEvent,
+  GenerateSpeechEndEvent,
+  GenerateSpeechStartEvent,
   ToolExecutionEndEvent,
   ToolExecutionStartEvent,
   Experimental_EvaluateEndEvent as EvaluateEndEvent,
@@ -33,6 +35,10 @@ import type {
   RerankEndEvent,
   RerankStartEvent,
   RerankingModelCallStartEvent,
+  TranscriptionEndEvent,
+  TranscriptionStartEvent,
+  Experimental_StreamTranscriptionEndEvent as StreamTranscriptionEndEvent,
+  Experimental_StreamTranscriptionStartEvent as StreamTranscriptionStartEvent,
   InferTelemetryEvent,
   Telemetry,
   TelemetryOptions,
@@ -48,6 +54,7 @@ import {
   mapProviderName,
 } from './gen-ai-format-messages';
 import { recordErrorOnSpan } from './record-span';
+import { getProviderUsageAttributes } from './provider-usage-attributes';
 import { sanitizeAttributes } from './sanitize-attribute-value';
 import { selectAttributes } from './select-attributes';
 import {
@@ -209,8 +216,24 @@ export class OpenTelemetry implements Telemetry {
       | InferTelemetryEvent<GenerateTextStartEvent>
       | InferTelemetryEvent<GenerateObjectStartEvent>
       | InferTelemetryEvent<EmbedStartEvent>
-      | InferTelemetryEvent<RerankStartEvent>,
+      | InferTelemetryEvent<RerankStartEvent>
+      | InferTelemetryEvent<GenerateSpeechStartEvent>
+      | InferTelemetryEvent<TranscriptionStartEvent>,
   ): void {
+    if (event.operationId === 'ai.generateSpeech') {
+      this.onAudioOperationStart(
+        event as InferTelemetryEvent<GenerateSpeechStartEvent>,
+      );
+      return;
+    }
+
+    if (event.operationId === 'ai.transcribe') {
+      this.onAudioOperationStart(
+        event as InferTelemetryEvent<TranscriptionStartEvent>,
+      );
+      return;
+    }
+
     if (
       event.operationId === 'ai.embed' ||
       event.operationId === 'ai.embedMany'
@@ -239,6 +262,102 @@ export class OpenTelemetry implements Telemetry {
     this.onGenerateStart(event as InferTelemetryEvent<GenerateTextStartEvent>);
   }
 
+  experimental_onStreamTranscriptionStart(
+    event: InferTelemetryEvent<StreamTranscriptionStartEvent>,
+  ): void {
+    this.onAudioOperationStart(event);
+  }
+
+  private onAudioOperationStart(
+    event: InferTelemetryEvent<
+      | GenerateSpeechStartEvent
+      | TranscriptionStartEvent
+      | StreamTranscriptionStartEvent
+    >,
+  ): void {
+    const telemetry: TelemetryOptions = {
+      recordInputs: event.recordInputs,
+      recordOutputs: event.recordOutputs,
+      functionId: event.functionId,
+    };
+    const providerName = mapProviderName(event.provider);
+    const operationName = mapOperationName(event.operationId);
+    const audio = 'audio' in event ? event.audio : undefined;
+    const text = 'text' in event ? event.text : undefined;
+    const baseSupplementalAttributes = selectSupplementalAttributes(
+      telemetry,
+      this.supplementalAttributes,
+      {
+        headers: getHeaderAttributes(event.headers),
+      },
+    );
+
+    const attributes = selectAttributes(telemetry, {
+      'gen_ai.operation.name': operationName,
+      'gen_ai.provider.name': providerName,
+      'gen_ai.request.model': event.modelId,
+      'gen_ai.agent.name': telemetry.functionId,
+      'gen_ai.output.type':
+        event.operationId === 'ai.generateSpeech' ? 'speech' : 'text',
+      'gen_ai.request.stream': event.operationId === 'ai.streamTranscribe',
+      'gen_ai.input.messages':
+        text == null
+          ? undefined
+          : {
+              input: () =>
+                JSON.stringify([
+                  {
+                    role: 'user',
+                    parts: [{ type: 'text', content: text }],
+                  },
+                ]),
+            },
+      'ai.request.text': text == null ? undefined : { input: () => text },
+      'ai.request.audio.size':
+        audio?.byteLength == null
+          ? undefined
+          : { input: () => audio.byteLength },
+      'ai.request.audio.media_type':
+        audio == null ? undefined : { input: () => audio.mediaType },
+      ...baseSupplementalAttributes,
+    });
+
+    const rootSpan = this.tracer.startSpan(
+      `${operationName} ${event.modelId}`,
+      {
+        attributes: this.getSpanAttributes({
+          attributes,
+          spanType: 'operation',
+          operationId: event.operationId,
+          callId: event.callId,
+          runtimeContext: undefined,
+        }),
+        kind: SpanKind.CLIENT,
+      },
+    );
+    const rootContext = trace.setSpan(context.active(), rootSpan);
+
+    this.callStates.set(event.callId, {
+      operationId: event.operationId,
+      telemetry,
+      rootSpan,
+      rootContext,
+      stepSpan: undefined,
+      stepContext: undefined,
+      inferenceSpan: undefined,
+      inferenceContext: undefined,
+      embedSpans: new Map(),
+      rerankSpan: undefined,
+      evaluationSpan: undefined,
+      toolSpans: new Map(),
+      settings: {},
+      provider: event.provider,
+      modelId: event.modelId,
+      runtimeContext: undefined,
+      baseSupplementalAttributes,
+    });
+  }
+
   private onGenerateStart(
     event: InferTelemetryEvent<GenerateTextStartEvent>,
   ): void {
@@ -1032,7 +1151,9 @@ export class OpenTelemetry implements Telemetry {
       | GenerateTextEndEvent<ToolSet>
       | GenerateObjectEndEvent<unknown>
       | EmbedEndEvent
-      | RerankEndEvent,
+      | RerankEndEvent
+      | GenerateSpeechEndEvent
+      | TranscriptionEndEvent,
   ): void {
     const state = this.getCallState(event.callId);
     if (!state?.rootSpan) return;
@@ -1050,6 +1171,16 @@ export class OpenTelemetry implements Telemetry {
       return;
     }
 
+    if (
+      state.operationId === 'ai.generateSpeech' ||
+      state.operationId === 'ai.transcribe'
+    ) {
+      this.onAudioOperationEnd(
+        event as GenerateSpeechEndEvent | TranscriptionEndEvent,
+      );
+      return;
+    }
+
     if (
       state.operationId === 'ai.generateObject' ||
       state.operationId === 'ai.streamObject'
@@ -1061,6 +1192,90 @@ export class OpenTelemetry implements Telemetry {
     this.onGenerateEnd(event as GenerateTextEndEvent<ToolSet>);
   }
 
+  experimental_onStreamTranscriptionEnd(
+    event: InferTelemetryEvent<StreamTranscriptionEndEvent>,
+  ): void {
+    this.onAudioOperationEnd(event);
+  }
+
+  private onAudioOperationEnd(
+    event:
+      | GenerateSpeechEndEvent
+      | TranscriptionEndEvent
+      | StreamTranscriptionEndEvent,
+  ): void {
+    const state = this.getCallState(event.callId);
+    if (!state?.rootSpan) return;
+
+    const outputAudio =
+      event.operationId === 'ai.generateSpeech' ? event.audio : undefined;
+    const inputAudio =
+      event.operationId === 'ai.generateSpeech' ? undefined : event.audio;
+    const transcript = 'segments' in event ? event.text : undefined;
+
+    state.rootSpan.setAttributes(
+      selectAttributes(state.telemetry, {
+        'ai.request.audio.size':
+          inputAudio == null
+            ? undefined
+            : { input: () => inputAudio.byteLength },
+        'ai.request.audio.media_type':
+          inputAudio == null
+            ? undefined
+            : { input: () => inputAudio.mediaType },
+        'ai.response.text':
+          transcript == null ? undefined : { output: () => transcript },
+        'gen_ai.output.messages':
+          transcript == null
+            ? undefined
+            : {
+                output: () =>
+                  JSON.stringify([
+                    {
+                      role: 'assistant',
+                      parts: [{ type: 'text', content: transcript }],
+                    },
+                  ]),
+              },
+        'ai.response.audio.size':
+          outputAudio == null
+            ? undefined
+            : { output: () => outputAudio.byteLength },
+        'ai.response.audio.media_type':
+          outputAudio == null
+            ? undefined
+            : { output: () => outputAudio.mediaType },
+        'ai.response.audio.format':
+          outputAudio == null || !('format' in outputAudio)
+            ? undefined
+            : { output: () => String(outputAudio.format) },
+        ...getProviderUsageAttributes({
+          usage: event.usage,
+          prefix: 'gen_ai.usage',
+        }),
+        ...selectSupplementalAttributes(
+          state.telemetry,
+          this.supplementalAttributes,
+          {
+            usage: {
+              'ai.response.usage':
+                event.usage == null ? undefined : JSON.stringify(event.usage),
+            },
+            providerMetadata: {
+              'ai.response.providerMetadata':
+                event.providerMetadata == null
+                  ? undefined
+                  : JSON.stringify(event.providerMetadata),
+            },
+          },
+        ),
+      }),
+    );
+
+    state.rootSpan.end();
+    this.cleanupCallState(event.callId);
+  }
+
   private onGenerateEnd(event: GenerateTextEndEvent<ToolSet>): void {
     const state = this.getCallState(event.callId);
     if (!state?.rootSpan) return;

package/src/provider-usage-attributes.ts

@@ -0,0 +1,69 @@
+import type { Attributes } from '@opentelemetry/api';
+import type { JSONObject, JSONValue } from '@ai-sdk/provider';
+
+const usageKeyAliases: Record<string, string> = {
+  inputtokens: 'input_tokens',
+  prompttokens: 'input_tokens',
+  prompttokencount: 'input_tokens',
+  totalinputtokens: 'input_tokens',
+  outputtokens: 'output_tokens',
+  completiontokens: 'output_tokens',
+  completiontokencount: 'output_tokens',
+  candidatestokencount: 'output_tokens',
+  totaltokens: 'total_tokens',
+  totaltokencount: 'total_tokens',
+};
+
+function normalizeUsageKey(key: string): string {
+  const aliasKey = key.replace(/[^a-zA-Z0-9]/g, '').toLowerCase();
+  const alias = usageKeyAliases[aliasKey];
+
+  if (alias != null) {
+    return alias;
+  }
+
+  return key
+    .replace(/([a-z0-9])([A-Z])/g, '$1_$2')
+    .replace(/[^a-zA-Z0-9]+/g, '_')
+    .replace(/^_+|_+$/g, '')
+    .toLowerCase();
+}
+
+/**
+ * Converts provider-native usage objects into numeric span attributes.
+ *
+ * Usage units such as tokens, characters, and seconds remain distinct so
+ * observability backends can apply the provider/model-specific price unit.
+ */
+export function getProviderUsageAttributes({
+  usage,
+  prefix,
+}: {
+  usage: JSONObject | undefined;
+  prefix: string;
+}): Attributes {
+  const attributes: Attributes = {};
+
+  const addValue = (value: JSONValue | undefined, path: string[]) => {
+    if (typeof value === 'number') {
+      if (Number.isFinite(value) && path.length > 0) {
+        attributes[`${prefix}.${path.join('.')}`] = value;
+      }
+      return;
+    }
+
+    if (value == null || typeof value !== 'object' || Array.isArray(value)) {
+      return;
+    }
+
+    for (const [key, nestedValue] of Object.entries(value)) {
+      const normalizedKey = normalizeUsageKey(key);
+      if (normalizedKey !== '') {
+        addValue(nestedValue, [...path, normalizedKey]);
+      }
+    }
+  };
+
+  addValue(usage, []);
+  return attributes;
+}