openlit 1.34.19__py3-none-any.whl → 1.34.22__py3-none-any.whl

openlit/__helpers.py

@@ -402,3 +402,43 @@ def record_embedding_metrics(metrics, gen_ai_operation, gen_ai_system, server_ad
     metrics["genai_requests"].add(1, attributes)
     metrics["genai_prompt_tokens"].add(input_tokens, attributes)
     metrics["genai_cost"].record(cost, attributes)
+
+def record_audio_metrics(metrics, gen_ai_operation, gen_ai_system, server_address, server_port,
+    request_model, response_model, environment, application_name, start_time, end_time, cost):
+    """
+    Record audio-specific metrics for the operation.
+    """
+
+    attributes = create_metrics_attributes(
+        operation=gen_ai_operation,
+        system=gen_ai_system,
+        server_address=server_address,
+        server_port=server_port,
+        request_model=request_model,
+        response_model=response_model,
+        service_name=application_name,
+        deployment_environment=environment,
+    )
+    metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
+    metrics["genai_requests"].add(1, attributes)
+    metrics["genai_cost"].record(cost, attributes)
+
+def record_image_metrics(metrics, gen_ai_operation, gen_ai_system, server_address, server_port,
+    request_model, response_model, environment, application_name, start_time, end_time, cost):
+    """
+    Record image-specific metrics for the operation.
+    """
+
+    attributes = create_metrics_attributes(
+        operation=gen_ai_operation,
+        system=gen_ai_system,
+        server_address=server_address,
+        server_port=server_port,
+        request_model=request_model,
+        response_model=response_model,
+        service_name=application_name,
+        deployment_environment=environment,
+    )
+    metrics["genai_client_operation_duration"].record(end_time - start_time, attributes)
+    metrics["genai_requests"].add(1, attributes)
+    metrics["genai_cost"].record(cost, attributes)

openlit/instrumentation/bedrock/__init__.py

@@ -1,4 +1,3 @@
-# pylint: disable=useless-return, bad-staticmethod-argument, disable=duplicate-code
 """Initializer of Auto Instrumentation of AWS Bedrock Functions"""
 
 from typing import Collection
@@ -6,37 +5,43 @@ import importlib.metadata
 from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
 from wrapt import wrap_function_wrapper
 
-from openlit.instrumentation.bedrock.bedrock import converse
+from openlit.instrumentation.bedrock.bedrock import converse, converse_stream
 
 _instruments = ("boto3 >= 1.34.138",)
 
 class BedrockInstrumentor(BaseInstrumentor):
     """
-    An instrumentor for AWS Bedrock's client library.
+    An instrumentor for AWS Bedrock client library.
     """
 
     def instrumentation_dependencies(self) -> Collection[str]:
         return _instruments
 
     def _instrument(self, **kwargs):
-        application_name = kwargs.get("application_name", "default_application")
-        environment = kwargs.get("environment", "default_environment")
+        version = importlib.metadata.version("boto3")
+        environment = kwargs.get("environment", "default")
+        application_name = kwargs.get("application_name", "default")
         tracer = kwargs.get("tracer")
-        event_provider = kwargs.get('event_provider')
-        metrics = kwargs.get("metrics_dict")
         pricing_info = kwargs.get("pricing_info", {})
         capture_message_content = kwargs.get("capture_message_content", False)
+        metrics = kwargs.get("metrics_dict")
         disable_metrics = kwargs.get("disable_metrics")
-        version = importlib.metadata.version("boto3")
 
-        #sync
+        # sync
+        wrap_function_wrapper(
+            "botocore.client",
+            "ClientCreator.create_client",
+            converse(version, environment, application_name, tracer, pricing_info,
+                capture_message_content, metrics, disable_metrics),
+        )
+
+        # streaming
         wrap_function_wrapper(
-            "botocore.client",  
-            "ClientCreator.create_client",  
-            converse(version, environment, application_name,
-                     tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+            "botocore.client",
+            "ClientCreator.create_client",
+            converse_stream(version, environment, application_name, tracer, pricing_info,
+                capture_message_content, metrics, disable_metrics),
         )
 
     def _uninstrument(self, **kwargs):
-        # Proper uninstrumentation logic to revert patched methods
         pass

openlit/instrumentation/bedrock/bedrock.py

@@ -2,64 +2,65 @@
 Module for monitoring Amazon Bedrock API calls.
 """
 
-import logging
 import time
 from opentelemetry.trace import SpanKind
 from openlit.__helpers import (
+    handle_exception,
     set_server_address_and_port
 )
 from openlit.instrumentation.bedrock.utils import (
+    process_chunk,
     process_chat_response,
+    process_streaming_chat_response,
 )
 from openlit.semcov import SemanticConvention
 
-# Initialize logger for logging potential issues and operations
-logger = logging.getLogger(__name__)
-
-def converse(version, environment, application_name, tracer, event_provider,
-         pricing_info, capture_message_content, metrics, disable_metrics):
+def converse(version, environment, application_name, tracer, pricing_info, capture_message_content, metrics, disable_metrics):
     """
-    Generates a telemetry wrapper for GenAI function call
+    Generates a telemetry wrapper for AWS Bedrock converse calls.
     """
 
     def wrapper(wrapped, instance, args, kwargs):
         """
-        Wraps the GenAI function call.
+        Wraps the ClientCreator.create_client call.
         """
 
         def converse_wrapper(original_method, *method_args, **method_kwargs):
-
             """
-            Wraps the GenAI function call.
+            Wraps the individual converse method call.
             """
 
-            server_address, server_port = set_server_address_and_port(instance, 'aws.amazon.com', 443)
-            request_model = method_kwargs.get('modelId', 'amazon.titan-text-express-v1')
+            server_address, server_port = set_server_address_and_port(instance, "aws.amazon.com", 443)
+            request_model = method_kwargs.get("modelId", "amazon.titan-text-express-v1")
 
-            span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}'
+            span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
             with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
                 start_time = time.time()
                 response = original_method(*method_args, **method_kwargs)
-                llm_config = method_kwargs.get('inferenceConfig', {})
-                response = process_chat_response(
-                    response=response,
-                    request_model=request_model,
-                    pricing_info=pricing_info,
-                    server_port=server_port,
-                    server_address=server_address,
-                    environment=environment,
-                    application_name=application_name,
-                    metrics=metrics,
-                    event_provider=event_provider,
-                    start_time=start_time,
-                    span=span,
-                    capture_message_content=capture_message_content,
-                    disable_metrics=disable_metrics,
-                    version=version,
-                    llm_config=llm_config,
-                    **method_kwargs
-                )
+                llm_config = method_kwargs.get("inferenceConfig", {})
+
+                try:
+                    response = process_chat_response(
+                        response=response,
+                        request_model=request_model,
+                        pricing_info=pricing_info,
+                        server_port=server_port,
+                        server_address=server_address,
+                        environment=environment,
+                        application_name=application_name,
+                        metrics=metrics,
+                        start_time=start_time,
+                        span=span,
+                        capture_message_content=capture_message_content,
+                        disable_metrics=disable_metrics,
+                        version=version,
+                        llm_config=llm_config,
+                        **method_kwargs
+                    )
+
+                except Exception as e:
+                    handle_exception(span, e)
 
                 return response
 
@@ -67,10 +68,143 @@ def converse(version, environment, application_name, tracer, event_provider,
         client = wrapped(*args, **kwargs)
 
         # Replace the original method with the instrumented one
-        if kwargs.get('service_name') == 'bedrock-runtime':
+        if kwargs.get("service_name") == "bedrock-runtime":
             original_invoke_model = client.converse
-            client.converse = lambda *args, **kwargs: converse_wrapper(original_invoke_model,
-                                                                            *args, **kwargs)
+            client.converse = lambda *args, **kwargs: converse_wrapper(original_invoke_model, *args, **kwargs)
+
+        return client
+
+    return wrapper
+
+def converse_stream(version, environment, application_name, tracer, pricing_info, capture_message_content, metrics, disable_metrics):
+    """
+    Generates a telemetry wrapper for AWS Bedrock converse_stream calls.
+    """
+
+    class TracedSyncStream:
+        """
+        Wrapper for streaming responses to collect telemetry.
+        """
+
+        def __init__(
+                self,
+                wrapped_response,
+                span,
+                span_name,
+                kwargs,
+                server_address,
+                server_port,
+                **args,
+            ):
+            self.__wrapped_response = wrapped_response
+            # Extract the actual stream iterator from the response
+            if isinstance(wrapped_response, dict) and "stream" in wrapped_response:
+                self.__wrapped_stream = iter(wrapped_response["stream"])
+            else:
+                self.__wrapped_stream = iter(wrapped_response)
+
+            self._span = span
+            self._span_name = span_name
+            self._llmresponse = ""
+            self._response_id = ""
+            self._response_model = ""
+            self._finish_reason = ""
+            self._tools = None
+            self._input_tokens = 0
+            self._output_tokens = 0
+
+            self._args = args
+            self._kwargs = kwargs
+            self._start_time = time.time()
+            self._end_time = None
+            self._timestamps = []
+            self._ttft = 0
+            self._tbt = 0
+            self._server_address = server_address
+            self._server_port = server_port
+
+        def __enter__(self):
+            if hasattr(self.__wrapped_stream, "__enter__"):
+                self.__wrapped_stream.__enter__()
+            return self
+
+        def __exit__(self, exc_type, exc_value, traceback):
+            if hasattr(self.__wrapped_stream, "__exit__"):
+                self.__wrapped_stream.__exit__(exc_type, exc_value, traceback)
+
+        def __iter__(self):
+            return self
+
+        def __getattr__(self, name):
+            """Delegate attribute access to the wrapped response."""
+            return getattr(self.__wrapped_response, name)
+
+        def get(self, key, default=None):
+            """Delegate get method to the wrapped response if its a dict."""
+            if isinstance(self.__wrapped_response, dict):
+                return self.__wrapped_response.get(key, default)
+            return getattr(self.__wrapped_response, key, default)
+
+        def __getitem__(self, key):
+            """Delegate item access to the wrapped response if its a dict."""
+            if isinstance(self.__wrapped_response, dict):
+                return self.__wrapped_response[key]
+            return getattr(self.__wrapped_response, key)
+
+        def __next__(self):
+            try:
+                chunk = next(self.__wrapped_stream)
+                process_chunk(self, chunk)
+                return chunk
+            except StopIteration:
+                try:
+                    llm_config = self._kwargs.get("inferenceConfig", {})
+                    with tracer.start_as_current_span(self._span_name, kind=SpanKind.CLIENT) as self._span:
+                        process_streaming_chat_response(
+                            self,
+                            pricing_info=pricing_info,
+                            environment=environment,
+                            application_name=application_name,
+                            metrics=metrics,
+                            capture_message_content=capture_message_content,
+                            disable_metrics=disable_metrics,
+                            version=version,
+                            llm_config=llm_config
+                        )
+
+                except Exception as e:
+                    handle_exception(self._span, e)
+
+                raise
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the ClientCreator.create_client call.
+        """
+
+        def converse_stream_wrapper(original_method, *method_args, **method_kwargs):
+            """
+            Wraps the individual converse_stream method call.
+            """
+
+            server_address, server_port = set_server_address_and_port(instance, "aws.amazon.com", 443)
+            request_model = method_kwargs.get("modelId", "amazon.titan-text-express-v1")
+
+            span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
+
+            # Get the streaming response
+            stream_response = original_method(*method_args, **method_kwargs)
+            span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
+
+            return TracedSyncStream(stream_response, span, span_name, method_kwargs, server_address, server_port)
+
+        # Get the original client instance from the wrapper
+        client = wrapped(*args, **kwargs)
+
+        # Replace the original method with the instrumented one
+        if kwargs.get("service_name") == "bedrock-runtime":
+            original_stream_model = client.converse_stream
+            client.converse_stream = lambda *args, **kwargs: converse_stream_wrapper(original_stream_model, *args, **kwargs)
 
         return client