feat(instrumentation-aws-sdk): add bedrock-runtime extension to apply gen ai conventions (#2700)

Co-authored-by: Trent Mick <[email protected]>

Author: anuraaga

package-lock.json

@@ -1,13 +1,18 @@
-# Note: tests must set `SKIP_TEST_IF_DISABLE=true` to override usage of
-# `mocha --require '../../../scripts/skip-test-if.js' ...` if calling `npm test`.
-
 # Versions [3.363.0, 3.377.0] of all @aws-sdk/client-* were bad releases. See:
 # - https://github.com/open-telemetry/opentelemetry-js-contrib/pull/2464#issuecomment-2403652552
 # - https://github.com/open-telemetry/opentelemetry-js-contrib/issues/1828#issuecomment-1834276719
 
-# node version support in JS SDK v3:
-# - 14.x dropped in v3.567.0 https://github.com/aws/aws-sdk-js-v3/pull/6034
-# - 16.x dropped in v3.723.0 https://github.com/aws/aws-sdk-js-v3/pull/6775
+"@aws-sdk/client-bedrock-runtime":
+  env:
+    - SKIP_TEST_IF_DISABLE=true
+  jobs:
+    - node: ">=18"
+      versions:
+        include: "^3.587.0"
+        exclude: ">=3.363.0 <=3.377.0"
+        mode: "max-7"
+      commands:
+        - mocha --require '@opentelemetry/contrib-test-utils' test/bedrock-runtime.test.ts
 
 "@aws-sdk/client-s3":
   env:
@@ -22,22 +27,6 @@
       commands:
         - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-s3.test.ts
         - mocha --require '@opentelemetry/contrib-test-utils' test/s3.test.ts
-    - node: "16"
-      versions:
-        include: ">=3.6.1 <3.723.0"
-        exclude: "3.529.0 || >=3.363.0 <=3.377.0"
-        mode: "max-7"
-      commands:
-        - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-s3.test.ts
-        - mocha --require '@opentelemetry/contrib-test-utils' test/s3.test.ts
-    - node: "14"
-      versions:
-        include: ">=3.6.1 <3.567.0"
-        exclude: "3.529.0 || >=3.363.0 <=3.377.0"
-        mode: "max-7"
-      commands:
-        - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-s3.test.ts
-        - mocha --require '@opentelemetry/contrib-test-utils' test/s3.test.ts
 
 "@aws-sdk/client-sqs":
   env:
@@ -50,18 +39,3 @@
         mode: "max-7"
       commands:
         - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-sqs.test.ts
-    - node: "16"
-      versions:
-        include: ">=3.24.0 <3.723.0"
-        exclude: ">=3.363.0 <=3.377.0"
-        mode: "max-7"
-      commands:
-        - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-sqs.test.ts
-    - node: "14"
-      versions:
-        include: ">=3.24.0 <3.567.0"
-        exclude: ">=3.363.0 <=3.377.0"
-        mode: "max-7"
-      commands:
-        - mocha --require '@opentelemetry/contrib-test-utils' test/aws-sdk-v3-sqs.test.ts
-

plugins/node/opentelemetry-instrumentation-aws-sdk/.tav.yml

@@ -94,6 +94,7 @@ Specific service logic currently implemented for:
 - [SNS](./doc/sns.md)
 - [Lambda](./doc/lambda.md)
 - DynamoDb
+- Amazon Bedrock Runtime (See the [GenAI semantic conventions](https://opentelemetry.io/docs/specs/semconv/gen-ai/).)
 
 ## Potential Side Effects
 

plugins/node/opentelemetry-instrumentation-aws-sdk/README.md

@@ -35,7 +35,7 @@
     "prewatch": "npm run precompile",
     "prepublishOnly": "npm run compile",
     "tdd": "npm run test -- --watch-extensions ts --watch",
-    "test": "SKIP_TEST_IF_NODE_OLDER_THAN=18 nyc mocha --require '../../../scripts/skip-test-if.js' --require '@opentelemetry/contrib-test-utils' 'test/**/*.test.ts'",
+    "test": "nyc mocha --require '@opentelemetry/contrib-test-utils' 'test/**/*.test.ts'",
     "test-all-versions": "tav",
     "version:update": "node ../../../scripts/version-update.js",
     "watch": "tsc -w"
@@ -50,6 +50,7 @@
     "@opentelemetry/semantic-conventions": "^1.27.0"
   },
   "devDependencies": {
+    "@aws-sdk/client-bedrock-runtime": "^3.587.0",
     "@aws-sdk/client-dynamodb": "^3.85.0",
     "@aws-sdk/client-kinesis": "^3.85.0",
     "@aws-sdk/client-lambda": "^3.85.0",

plugins/node/opentelemetry-instrumentation-aws-sdk/package.json

@@ -323,7 +323,9 @@ export class AwsInstrumentation extends InstrumentationBase<AwsSdkInstrumentatio
         const serviceName =
           clientConfig?.serviceId ??
           removeSuffixFromStringIfExists(
-            awsExecutionContext.clientName,
+            // Use 'AWS' as a fallback serviceName to match type definition.
+            // In practice, `clientName` should always be set.
+            awsExecutionContext.clientName || 'AWS',
             'Client'
           );
         const commandName =

plugins/node/opentelemetry-instrumentation-aws-sdk/src/aws-sdk.ts

@@ -0,0 +1,140 @@
+/*
+ * Copyright The OpenTelemetry Authors
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      https://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+/*
+ * This file contains a copy of unstable semantic convention definitions
+ * used by this package.
+ * @see https://github.com/open-telemetry/opentelemetry-js/tree/main/semantic-conventions#unstable-semconv
+ */
+
+/**
+ * The name of the operation being performed.
+ *
+ * @note If one of the predefined values applies, but specific system uses a different name it's **RECOMMENDED** to document it in the semantic conventions for specific GenAI system and use system-specific name in the instrumentation. If a different name is not documented, instrumentation libraries **SHOULD** use applicable predefined value.
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_OPERATION_NAME = 'gen_ai.operation.name' as const;
+
+/**
+ * The maximum number of tokens the model generates for a request.
+ *
+ * @example 100
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_REQUEST_MAX_TOKENS =
+  'gen_ai.request.max_tokens' as const;
+
+/**
+ * The name of the GenAI model a request is being made to.
+ *
+ * @example "gpt-4"
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_REQUEST_MODEL = 'gen_ai.request.model' as const;
+
+/**
+ * List of sequences that the model will use to stop generating further tokens.
+ *
+ * @example ["forest", "lived"]
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_REQUEST_STOP_SEQUENCES =
+  'gen_ai.request.stop_sequences' as const;
+
+/**
+ * The temperature setting for the GenAI request.
+ *
+ * @example 0.0
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_REQUEST_TEMPERATURE =
+  'gen_ai.request.temperature' as const;
+
+/**
+ * The top_p sampling setting for the GenAI request.
+ *
+ * @example 1.0
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_REQUEST_TOP_P = 'gen_ai.request.top_p' as const;
+
+/**
+ * Array of reasons the model stopped generating tokens, corresponding to each generation received.
+ *
+ * @example ["stop"]
+ * @example ["stop", "length"]
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_RESPONSE_FINISH_REASONS =
+  'gen_ai.response.finish_reasons' as const;
+
+/**
+ * The Generative AI product as identified by the client or server instrumentation.
+ *
+ * @example "openai"
+ *
+ * @note The `gen_ai.system` describes a family of GenAI models with specific model identified
+ * by `gen_ai.request.model` and `gen_ai.response.model` attributes.
+ *
+ * The actual GenAI product may differ from the one identified by the client.
+ * Multiple systems, including Azure OpenAI and Gemini, are accessible by OpenAI client
+ * libraries. In such cases, the `gen_ai.system` is set to `openai` based on the
+ * instrumentation's best knowledge, instead of the actual system. The `server.address`
+ * attribute may help identify the actual system in use for `openai`.
+ *
+ * For custom model, a custom friendly name **SHOULD** be used.
+ * If none of these options apply, the `gen_ai.system` **SHOULD** be set to `_OTHER`.
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_SYSTEM = 'gen_ai.system' as const;
+
+/**
+ * The number of tokens used in the GenAI input (prompt).
+ *
+ * @example 100
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_USAGE_INPUT_TOKENS =
+  'gen_ai.usage.input_tokens' as const;
+
+/**
+ * The number of tokens used in the GenAI response (completion).
+ *
+ * @example 180
+ *
+ * @experimental This attribute is experimental and is subject to breaking changes in minor releases of `@opentelemetry/semantic-conventions`.
+ */
+export const ATTR_GEN_AI_USAGE_OUTPUT_TOKENS =
+  'gen_ai.usage.output_tokens' as const;
+
+/**
+ * Enum value "chat" for attribute {@link ATTR_GEN_AI_OPERATION_NAME}.
+ */
+export const GEN_AI_OPERATION_NAME_VALUE_CHAT = 'chat' as const;
+
+/**
+ * Enum value "aws.bedrock" for attribute {@link ATTR_GEN_AI_SYSTEM}.
+ */
+export const GEN_AI_SYSTEM_VALUE_AWS_BEDROCK = 'aws.bedrock' as const;

plugins/node/opentelemetry-instrumentation-aws-sdk/src/semconv.ts

@@ -21,6 +21,7 @@ import {
   NormalizedRequest,
   NormalizedResponse,
 } from '../types';
+import { BedrockRuntimeServiceExtension } from './bedrock-runtime';
 import { DynamodbServiceExtension } from './dynamodb';
 import { SnsServiceExtension } from './sns';
 import { LambdaServiceExtension } from './lambda';
@@ -37,6 +38,7 @@ export class ServicesExtensions implements ServiceExtension {
     this.services.set('Lambda', new LambdaServiceExtension());
     this.services.set('S3', new S3ServiceExtension());
     this.services.set('Kinesis', new KinesisServiceExtension());
+    this.services.set('BedrockRuntime', new BedrockRuntimeServiceExtension());
   }
 
   requestPreSpanHook(

plugins/node/opentelemetry-instrumentation-aws-sdk/src/services/ServicesExtensions.ts

@@ -0,0 +1,134 @@
+/*
+ * Copyright The OpenTelemetry Authors
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      https://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+import { Attributes, DiagLogger, Span, Tracer } from '@opentelemetry/api';
+import { RequestMetadata, ServiceExtension } from './ServiceExtension';
+import {
+  ATTR_GEN_AI_SYSTEM,
+  ATTR_GEN_AI_OPERATION_NAME,
+  ATTR_GEN_AI_REQUEST_MODEL,
+  ATTR_GEN_AI_REQUEST_MAX_TOKENS,
+  ATTR_GEN_AI_REQUEST_TEMPERATURE,
+  ATTR_GEN_AI_REQUEST_TOP_P,
+  ATTR_GEN_AI_REQUEST_STOP_SEQUENCES,
+  ATTR_GEN_AI_USAGE_INPUT_TOKENS,
+  ATTR_GEN_AI_USAGE_OUTPUT_TOKENS,
+  ATTR_GEN_AI_RESPONSE_FINISH_REASONS,
+  GEN_AI_OPERATION_NAME_VALUE_CHAT,
+  GEN_AI_SYSTEM_VALUE_AWS_BEDROCK,
+} from '../semconv';
+import {
+  AwsSdkInstrumentationConfig,
+  NormalizedRequest,
+  NormalizedResponse,
+} from '../types';
+
+export class BedrockRuntimeServiceExtension implements ServiceExtension {
+  requestPreSpanHook(
+    request: NormalizedRequest,
+    config: AwsSdkInstrumentationConfig,
+    diag: DiagLogger
+  ): RequestMetadata {
+    switch (request.commandName) {
+      case 'Converse':
+        return this.requestPreSpanHookConverse(request, config, diag);
+    }
+
+    return {
+      isIncoming: false,
+    };
+  }
+
+  private requestPreSpanHookConverse(
+    request: NormalizedRequest,
+    config: AwsSdkInstrumentationConfig,
+    diag: DiagLogger
+  ): RequestMetadata {
+    let spanName = GEN_AI_OPERATION_NAME_VALUE_CHAT;
+    const spanAttributes: Attributes = {
+      [ATTR_GEN_AI_SYSTEM]: GEN_AI_SYSTEM_VALUE_AWS_BEDROCK,
+      [ATTR_GEN_AI_OPERATION_NAME]: GEN_AI_OPERATION_NAME_VALUE_CHAT,
+    };
+
+    const modelId = request.commandInput.modelId;
+    if (modelId) {
+      spanAttributes[ATTR_GEN_AI_REQUEST_MODEL] = modelId;
+      if (spanName) {
+        spanName += ` ${modelId}`;
+      }
+    }
+
+    const inferenceConfig = request.commandInput.inferenceConfig;
+    if (inferenceConfig) {
+      const { maxTokens, temperature, topP, stopSequences } = inferenceConfig;
+      if (maxTokens !== undefined) {
+        spanAttributes[ATTR_GEN_AI_REQUEST_MAX_TOKENS] = maxTokens;
+      }
+      if (temperature !== undefined) {
+        spanAttributes[ATTR_GEN_AI_REQUEST_TEMPERATURE] = temperature;
+      }
+      if (topP !== undefined) {
+        spanAttributes[ATTR_GEN_AI_REQUEST_TOP_P] = topP;
+      }
+      if (stopSequences !== undefined) {
+        spanAttributes[ATTR_GEN_AI_REQUEST_STOP_SEQUENCES] = stopSequences;
+      }
+    }
+
+    return {
+      spanName,
+      isIncoming: false,
+      spanAttributes,
+    };
+  }
+
+  responseHook(
+    response: NormalizedResponse,
+    span: Span,
+    tracer: Tracer,
+    config: AwsSdkInstrumentationConfig
+  ) {
+    if (!span.isRecording()) {
+      return;
+    }
+
+    switch (response.request.commandName) {
+      case 'Converse':
+        return this.responseHookConverse(response, span, tracer, config);
+    }
+  }
+
+  private responseHookConverse(
+    response: NormalizedResponse,
+    span: Span,
+    tracer: Tracer,
+    config: AwsSdkInstrumentationConfig
+  ) {
+    const { stopReason, usage } = response.data;
+    if (usage) {
+      const { inputTokens, outputTokens } = usage;
+      if (inputTokens !== undefined) {
+        span.setAttribute(ATTR_GEN_AI_USAGE_INPUT_TOKENS, inputTokens);
+      }
+      if (outputTokens !== undefined) {
+        span.setAttribute(ATTR_GEN_AI_USAGE_OUTPUT_TOKENS, outputTokens);
+      }
+    }
+
+    if (stopReason !== undefined) {
+      span.setAttribute(ATTR_GEN_AI_RESPONSE_FINISH_REASONS, [stopReason]);
+    }
+  }
+}