Merge branch 'main' into add-checksum

Author: harrryr

.github/workflows/application-signals-e2e-test.yml

@@ -117,6 +117,7 @@ jobs:
       python-version: '3.8'
 
   eks-v3-9-amd64:
+    if: ${{ always() }}
     needs: eks-v3-8-amd64
     uses: aws-observability/aws-application-signals-test-framework/.github/workflows/python-eks-test.yml@main
     secrets: inherit
@@ -128,6 +129,7 @@ jobs:
       python-version: '3.9'
 
   eks-v3-10-amd64:
+    if: ${{ always() }}
     needs: eks-v3-9-amd64
     uses: aws-observability/aws-application-signals-test-framework/.github/workflows/python-eks-test.yml@main
     secrets: inherit
@@ -139,6 +141,7 @@ jobs:
       python-version: '3.10'
 
   eks-v3-11-amd64:
+    if: ${{ always() }}
     needs: eks-v3-10-amd64
     uses: aws-observability/aws-application-signals-test-framework/.github/workflows/python-eks-test.yml@main
     secrets: inherit
@@ -150,6 +153,7 @@ jobs:
       python-version: '3.11'
 
   eks-v3-12-amd64:
+    if: ${{ always() }}
     needs: eks-v3-11-amd64
     uses: aws-observability/aws-application-signals-test-framework/.github/workflows/python-eks-test.yml@main
     secrets: inherit

.github/workflows/release_lambda.yml

@@ -3,6 +3,9 @@ name: Release Lambda layer
 on:
   workflow_dispatch:
     inputs:
+      version:
+        description: The version to tag the lambda release with, e.g., 1.2.0
+        required: true
       aws_region:
         description: 'Deploy to aws regions'
         required: true
@@ -98,7 +101,7 @@ jobs:
             aws lambda publish-layer-version \
               --layer-name ${{ env.LAYER_NAME }} \
               --content S3Bucket=${{ env.BUCKET_NAME }},S3Key=aws-opentelemetry-python-layer.zip \
-              --compatible-runtimes python3.10 python3.11 python3.12 \
+              --compatible-runtimes python3.10 python3.11 python3.12 python3.13 \
               --compatible-architectures "arm64" "x86_64" \
               --license-info "Apache-2.0" \
               --description "AWS Distro of OpenTelemetry Lambda Layer for Python Runtime" \
@@ -184,45 +187,13 @@ jobs:
         with:
           name: layer.tf
           path: layer.tf
-      - name: Commit changes
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        run: |
-          git config user.name "github-actions[bot]"
-          git config user.email "github-actions[bot]@users.noreply.github.com"
-          mv layer.tf lambda-layer/terraform/lambda/
-          git add lambda-layer/terraform/lambda/layer.tf
-          git commit -m "Update Lambda layer ARNs for releasing" || echo "No changes to commit"
-          git push
-  create-release:
-    runs-on: ubuntu-latest
-    needs: generate-release-note
-    steps:
-      - name: Checkout Repo @ SHA - ${{ github.sha }}
-        uses: actions/checkout@v4
-      - name: Get latest commit SHA
-        run: |
-          echo "COMMIT_SHA=${GITHUB_SHA}" >> $GITHUB_ENV
-          SHORT_SHA=$(echo $GITHUB_SHA | cut -c1-7)
-          echo "SHORT_SHA=${SHORT_SHA}" >> $GITHUB_ENV
-      - name: Create Tag
-        run: |
-          git config user.name "github-actions[bot]"
-          git config user.email "github-actions[bot]@users.noreply.github.com"
-          TAG_NAME="lambda-${SHORT_SHA}"
-          git tag -a "$TAG_NAME" -m "Release Lambda layer based on commit $TAG_NAME"
-          git push origin "$TAG_NAME"
-          echo "TAG_NAME=${TAG_NAME}" >> $GITHUB_ENV
-        env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-      - name: Create Release
+      - name: Create GH release
         id: create_release
-        uses: actions/create-release@v1
         env:
-          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
-        with:
-          tag_name: ${{ env.TAG_NAME }}
-          release_name: "Release AWSOpenTelemetryDistroPython Lambda Layer"
-          body_path: lambda-layer/terraform/lambda/layer.tf
-          draft: true
-          prerelease: false
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} # This token is provided by Actions, you do not need to create your own token
+        run: |
+          gh release create --target "$GITHUB_REF_NAME" \
+             --title "Release lambda-v${{ github.event.inputs.version }}" \
+             --draft \
+             "lambda-v${{ github.event.inputs.version }}" \
+             layer.tf

Dockerfile

@@ -21,7 +21,7 @@ RUN sed -i "/opentelemetry-exporter-otlp-proto-grpc/d" ./aws-opentelemetry-distr
 RUN mkdir workspace && pip install --target workspace ./aws-opentelemetry-distro
 
 # Stage 2: Build the cp-utility binary
-FROM public.ecr.aws/docker/library/rust:1.75 as builder
+FROM public.ecr.aws/docker/library/rust:1.81 as builder
 
 WORKDIR /usr/src/cp-utility
 COPY ./tools/cp-utility .

aws-opentelemetry-distro/src/amazon/opentelemetry/distro/_aws_span_processing_util.py

@@ -29,6 +29,12 @@
 # TODO: Use Semantic Conventions once upgrade to 0.47b0
 GEN_AI_REQUEST_MODEL: str = "gen_ai.request.model"
 GEN_AI_SYSTEM: str = "gen_ai.system"
+GEN_AI_REQUEST_MAX_TOKENS: str = "gen_ai.request.max_tokens"
+GEN_AI_REQUEST_TEMPERATURE: str = "gen_ai.request.temperature"
+GEN_AI_REQUEST_TOP_P: str = "gen_ai.request.top_p"
+GEN_AI_RESPONSE_FINISH_REASONS: str = "gen_ai.response.finish_reasons"
+GEN_AI_USAGE_INPUT_TOKENS: str = "gen_ai.usage.input_tokens"
+GEN_AI_USAGE_OUTPUT_TOKENS: str = "gen_ai.usage.output_tokens"
 
 
 # Get dialect keywords retrieved from dialect_keywords.json file.

aws-opentelemetry-distro/src/amazon/opentelemetry/distro/patches/_bedrock_patches.py

@@ -2,7 +2,13 @@
 # SPDX-License-Identifier: Apache-2.0
 import abc
 import inspect
-from typing import Dict, Optional
+import io
+import json
+import logging
+import math
+from typing import Any, Dict, Optional
+
+from botocore.response import StreamingBody
 
 from amazon.opentelemetry.distro._aws_attribute_keys import (
     AWS_BEDROCK_AGENT_ID,
@@ -11,7 +17,16 @@
     AWS_BEDROCK_GUARDRAIL_ID,
     AWS_BEDROCK_KNOWLEDGE_BASE_ID,
 )
-from amazon.opentelemetry.distro._aws_span_processing_util import GEN_AI_REQUEST_MODEL, GEN_AI_SYSTEM
+from amazon.opentelemetry.distro._aws_span_processing_util import (
+    GEN_AI_REQUEST_MAX_TOKENS,
+    GEN_AI_REQUEST_MODEL,
+    GEN_AI_REQUEST_TEMPERATURE,
+    GEN_AI_REQUEST_TOP_P,
+    GEN_AI_RESPONSE_FINISH_REASONS,
+    GEN_AI_SYSTEM,
+    GEN_AI_USAGE_INPUT_TOKENS,
+    GEN_AI_USAGE_OUTPUT_TOKENS,
+)
 from opentelemetry.instrumentation.botocore.extensions.types import (
     _AttributeMapT,
     _AwsSdkCallContext,
@@ -26,7 +41,11 @@
 _GUARDRAIL_ID: str = "guardrailId"
 _GUARDRAIL_ARN: str = "guardrailArn"
 _MODEL_ID: str = "modelId"
-_AWS_BEDROCK_SYSTEM: str = "aws_bedrock"
+_AWS_BEDROCK_SYSTEM: str = "aws.bedrock"
+
+_logger = logging.getLogger(__name__)
+# Set logger level to DEBUG
+_logger.setLevel(logging.DEBUG)
 
 
 class _BedrockAgentOperation(abc.ABC):
@@ -240,3 +259,190 @@ def extract_attributes(self, attributes: _AttributeMapT):
         model_id = self._call_context.params.get(_MODEL_ID)
         if model_id:
             attributes[GEN_AI_REQUEST_MODEL] = model_id
+
+            # Get the request body if it exists
+            body = self._call_context.params.get("body")
+            if body:
+                try:
+                    request_body = json.loads(body)
+
+                    if "amazon.titan" in model_id:
+                        self._extract_titan_attributes(attributes, request_body)
+                    if "amazon.nova" in model_id:
+                        self._extract_nova_attributes(attributes, request_body)
+                    elif "anthropic.claude" in model_id:
+                        self._extract_claude_attributes(attributes, request_body)
+                    elif "meta.llama" in model_id:
+                        self._extract_llama_attributes(attributes, request_body)
+                    elif "cohere.command" in model_id:
+                        self._extract_cohere_attributes(attributes, request_body)
+                    elif "ai21.jamba" in model_id:
+                        self._extract_ai21_attributes(attributes, request_body)
+                    elif "mistral" in model_id:
+                        self._extract_mistral_attributes(attributes, request_body)
+
+                except json.JSONDecodeError:
+                    _logger.debug("Error: Unable to parse the body as JSON")
+
+    def _extract_titan_attributes(self, attributes, request_body):
+        config = request_body.get("textGenerationConfig", {})
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, config.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, config.get("topP"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, config.get("maxTokenCount"))
+
+    def _extract_nova_attributes(self, attributes, request_body):
+        config = request_body.get("inferenceConfig", {})
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, config.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, config.get("top_p"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, config.get("max_new_tokens"))
+
+    def _extract_claude_attributes(self, attributes, request_body):
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, request_body.get("max_tokens"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, request_body.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, request_body.get("top_p"))
+
+    def _extract_cohere_attributes(self, attributes, request_body):
+        prompt = request_body.get("message")
+        if prompt:
+            attributes[GEN_AI_USAGE_INPUT_TOKENS] = math.ceil(len(prompt) / 6)
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, request_body.get("max_tokens"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, request_body.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, request_body.get("p"))
+
+    def _extract_ai21_attributes(self, attributes, request_body):
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, request_body.get("max_tokens"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, request_body.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, request_body.get("top_p"))
+
+    def _extract_llama_attributes(self, attributes, request_body):
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, request_body.get("max_gen_len"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, request_body.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, request_body.get("top_p"))
+
+    def _extract_mistral_attributes(self, attributes, request_body):
+        prompt = request_body.get("prompt")
+        if prompt:
+            attributes[GEN_AI_USAGE_INPUT_TOKENS] = math.ceil(len(prompt) / 6)
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_MAX_TOKENS, request_body.get("max_tokens"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TEMPERATURE, request_body.get("temperature"))
+        self._set_if_not_none(attributes, GEN_AI_REQUEST_TOP_P, request_body.get("top_p"))
+
+    @staticmethod
+    def _set_if_not_none(attributes, key, value):
+        if value is not None:
+            attributes[key] = value
+
+    # pylint: disable=too-many-branches
+    def on_success(self, span: Span, result: Dict[str, Any]):
+        model_id = self._call_context.params.get(_MODEL_ID)
+
+        if not model_id:
+            return
+
+        if "body" in result and isinstance(result["body"], StreamingBody):
+            original_body = None
+            try:
+                original_body = result["body"]
+                body_content = original_body.read()
+
+                # Use one stream for telemetry
+                stream = io.BytesIO(body_content)
+                telemetry_content = stream.read()
+                response_body = json.loads(telemetry_content.decode("utf-8"))
+                if "amazon.titan" in model_id:
+                    self._handle_amazon_titan_response(span, response_body)
+                if "amazon.nova" in model_id:
+                    self._handle_amazon_nova_response(span, response_body)
+                elif "anthropic.claude" in model_id:
+                    self._handle_anthropic_claude_response(span, response_body)
+                elif "meta.llama" in model_id:
+                    self._handle_meta_llama_response(span, response_body)
+                elif "cohere.command" in model_id:
+                    self._handle_cohere_command_response(span, response_body)
+                elif "ai21.jamba" in model_id:
+                    self._handle_ai21_jamba_response(span, response_body)
+                elif "mistral" in model_id:
+                    self._handle_mistral_mistral_response(span, response_body)
+                # Replenish stream for downstream application use
+                new_stream = io.BytesIO(body_content)
+                result["body"] = StreamingBody(new_stream, len(body_content))
+
+            except json.JSONDecodeError:
+                _logger.debug("Error: Unable to parse the response body as JSON")
+            except Exception as e:  # pylint: disable=broad-exception-caught, invalid-name
+                _logger.debug("Error processing response: %s", e)
+            finally:
+                if original_body is not None:
+                    original_body.close()
+
+    # pylint: disable=no-self-use
+    def _handle_amazon_titan_response(self, span: Span, response_body: Dict[str, Any]):
+        if "inputTextTokenCount" in response_body:
+            span.set_attribute(GEN_AI_USAGE_INPUT_TOKENS, response_body["inputTextTokenCount"])
+            if "results" in response_body and response_body["results"]:
+                result = response_body["results"][0]
+                if "tokenCount" in result:
+                    span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, result["tokenCount"])
+                if "completionReason" in result:
+                    span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [result["completionReason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_amazon_nova_response(self, span: Span, response_body: Dict[str, Any]):
+        if "usage" in response_body:
+            usage = response_body["usage"]
+            if "inputTokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_INPUT_TOKENS, usage["inputTokens"])
+            if "outputTokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, usage["outputTokens"])
+        if "stopReason" in response_body:
+            span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [response_body["stopReason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_anthropic_claude_response(self, span: Span, response_body: Dict[str, Any]):
+        if "usage" in response_body:
+            usage = response_body["usage"]
+            if "input_tokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_INPUT_TOKENS, usage["input_tokens"])
+            if "output_tokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, usage["output_tokens"])
+        if "stop_reason" in response_body:
+            span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [response_body["stop_reason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_cohere_command_response(self, span: Span, response_body: Dict[str, Any]):
+        # Output tokens: Approximate from the response text
+        if "text" in response_body:
+            span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, math.ceil(len(response_body["text"]) / 6))
+        if "finish_reason" in response_body:
+            span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [response_body["finish_reason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_ai21_jamba_response(self, span: Span, response_body: Dict[str, Any]):
+        if "usage" in response_body:
+            usage = response_body["usage"]
+            if "prompt_tokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_INPUT_TOKENS, usage["prompt_tokens"])
+            if "completion_tokens" in usage:
+                span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, usage["completion_tokens"])
+        if "choices" in response_body:
+            choices = response_body["choices"][0]
+            if "finish_reason" in choices:
+                span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [choices["finish_reason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_meta_llama_response(self, span: Span, response_body: Dict[str, Any]):
+        if "prompt_token_count" in response_body:
+            span.set_attribute(GEN_AI_USAGE_INPUT_TOKENS, response_body["prompt_token_count"])
+        if "generation_token_count" in response_body:
+            span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, response_body["generation_token_count"])
+        if "stop_reason" in response_body:
+            span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [response_body["stop_reason"]])
+
+    # pylint: disable=no-self-use
+    def _handle_mistral_mistral_response(self, span: Span, response_body: Dict[str, Any]):
+        if "outputs" in response_body:
+            outputs = response_body["outputs"][0]
+            if "text" in outputs:
+                span.set_attribute(GEN_AI_USAGE_OUTPUT_TOKENS, math.ceil(len(outputs["text"]) / 6))
+        if "stop_reason" in outputs:
+            span.set_attribute(GEN_AI_RESPONSE_FINISH_REASONS, [outputs["stop_reason"]])