openlit 1.34.19__tar.gz → 1.34.22__tar.gz

{openlit-1.34.19 → openlit-1.34.22}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: openlit
-Version: 1.34.19
+Version: 1.34.22
 Summary: OpenTelemetry-native Auto instrumentation library for monitoring LLM Applications and GPUs, facilitating the integration of observability into your GenAI-driven projects
 License: Apache-2.0
 Keywords: OpenTelemetry,otel,otlp,llm,tracing,openai,anthropic,claude,cohere,llm monitoring,observability,monitoring,gpt,Generative AI,chatGPT,gpu

{openlit-1.34.19 → openlit-1.34.22}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "openlit"
-version = "1.34.19"
+version = "1.34.22"
 description = "OpenTelemetry-native Auto instrumentation library for monitoring LLM Applications and GPUs, facilitating the integration of observability into your GenAI-driven projects"
 authors = ["OpenLIT"]
 license = "Apache-2.0"

{openlit-1.34.19 → openlit-1.34.22}/src/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-1.34.19 → openlit-1.34.22}/src/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-1.34.22/src/openlit/instrumentation/bedrock/bedrock.py

@@ -0,0 +1,211 @@
+"""
+Module for monitoring Amazon Bedrock API calls.
+"""
+
+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
+
+def converse(version, environment, application_name, tracer, pricing_info, capture_message_content, metrics, disable_metrics):
+    """
+    Generates a telemetry wrapper for AWS Bedrock converse calls.
+    """
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the ClientCreator.create_client call.
+        """
+
+        def converse_wrapper(original_method, *method_args, **method_kwargs):
+            """
+            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")
+
+            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", {})
+
+                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
+
+        # 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_invoke_model = client.converse
+            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
+
+    return wrapper

openlit-1.34.22/src/openlit/instrumentation/bedrock/utils.py

@@ -0,0 +1,223 @@
+"""
+AWS Bedrock OpenTelemetry instrumentation utility functions
+"""
+import time
+
+from opentelemetry.trace import Status, StatusCode
+
+from openlit.__helpers import (
+    calculate_ttft,
+    response_as_dict,
+    calculate_tbt,
+    get_chat_model_cost,
+    record_completion_metrics,
+    common_span_attributes,
+    handle_exception
+)
+from openlit.semcov import SemanticConvention
+
+def format_content(messages):
+    """
+    Format the messages into a string for span events.
+    """
+
+    if not messages:
+        return ""
+
+    formatted_messages = []
+    for message in messages:
+        if isinstance(message, dict):
+            role = message.get("role", "user")
+            content = message.get("content", "")
+        else:
+            # Handle Bedrock object format
+            role = getattr(message, "role", "user")
+            content = getattr(message, "content", "")
+
+        if isinstance(content, list):
+            # Handle structured content (e.g., text + images)
+            text_parts = []
+            for part in content:
+                if isinstance(part, dict):
+                    # Bedrock format: {"text": "content"} or generic format: {"type": "text", "text": "content"}
+                    if "text" in part:
+                        text_parts.append(part.get("text", ""))
+                    elif part.get("type") == "text":
+                        text_parts.append(part.get("text", ""))
+            content = " ".join(text_parts)
+        elif not isinstance(content, str):
+            content = str(content)
+
+        formatted_messages.append(f"{role}: {content}")
+
+    return "\n".join(formatted_messages)
+
+def process_chunk(self, chunk):
+    """
+    Process a chunk of response data and update state.
+    """
+
+    end_time = time.time()
+    # Record the timestamp for the current chunk
+    self._timestamps.append(end_time)
+
+    if len(self._timestamps) == 1:
+        # Calculate time to first chunk
+        self._ttft = calculate_ttft(self._timestamps, self._start_time)
+
+    chunked = response_as_dict(chunk)
+
+    # Handle Bedrock messageStart event
+    if "messageStart" in chunked:
+        message_start = chunked.get("messageStart", {})
+        self._response_role = message_start.get("role", "assistant")
+
+    # Handle Bedrock contentBlockDelta event
+    if "contentBlockDelta" in chunked:
+        content_delta = chunked.get("contentBlockDelta", {})
+        delta = content_delta.get("delta", {})
+        if "text" in delta:
+            self._llmresponse += delta.get("text", "")
+
+    # Handle Bedrock messageStop event
+    if "messageStop" in chunked:
+        message_stop = chunked.get("messageStop", {})
+        self._finish_reason = message_stop.get("stopReason", "")
+
+    # Handle Bedrock metadata event (final event with usage info)
+    if "metadata" in chunked:
+        metadata = chunked.get("metadata", {})
+        usage = metadata.get("usage", {})
+        self._input_tokens = usage.get("inputTokens", 0)
+        self._output_tokens = usage.get("outputTokens", 0)
+        self._end_time = end_time
+
+def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+                        capture_message_content, disable_metrics, version, llm_config, is_stream):
+    """
+    Process chat request and generate Telemetry
+    """
+
+    scope._end_time = time.time()
+    if len(scope._timestamps) > 1:
+        scope._tbt = calculate_tbt(scope._timestamps)
+
+    formatted_messages = format_content(scope._kwargs.get("messages", []))
+    request_model = scope._kwargs.get("modelId", "amazon.titan-text-express-v1")
+
+    cost = get_chat_model_cost(request_model, pricing_info, scope._input_tokens, scope._output_tokens)
+
+    # Common Span Attributes
+    common_span_attributes(scope,
+        SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT, SemanticConvention.GEN_AI_SYSTEM_AWS_BEDROCK,
+        scope._server_address, scope._server_port, request_model, scope._response_model,
+        environment, application_name, is_stream, scope._tbt, scope._ttft, version)
+
+    # Bedrock-specific attributes from llm_config
+    bedrock_attributes = [
+        (SemanticConvention.GEN_AI_REQUEST_FREQUENCY_PENALTY, "frequencyPenalty"),
+        (SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, "maxTokens"),
+        (SemanticConvention.GEN_AI_REQUEST_PRESENCE_PENALTY, "presencePenalty"),
+        (SemanticConvention.GEN_AI_REQUEST_STOP_SEQUENCES, "stopSequences"),
+        (SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, "temperature"),
+        (SemanticConvention.GEN_AI_REQUEST_TOP_P, "topP"),
+        (SemanticConvention.GEN_AI_REQUEST_TOP_K, "topK"),
+    ]
+
+    # Set each bedrock-specific attribute if the corresponding value exists and is not None
+    for attribute, key in bedrock_attributes:
+        value = llm_config.get(key)
+        if value is not None:
+            scope._span.set_attribute(attribute, value)
+
+    # Span Attributes for Response parameters
+    scope._span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_ID, scope._response_id)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_FINISH_REASON, [scope._finish_reason])
+    scope._span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, "text" if isinstance(scope._llmresponse, str) else "json")
+
+    # Span Attributes for Cost and Tokens
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, scope._input_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, scope._output_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE, scope._input_tokens + scope._output_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
+
+    # Span Attributes for Content
+    if capture_message_content:
+        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_PROMPT, formatted_messages)
+        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_COMPLETION, scope._llmresponse)
+
+        # To be removed once the change to span_attributes (from span events) is complete
+        scope._span.add_event(
+            name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
+            attributes={
+                SemanticConvention.GEN_AI_CONTENT_PROMPT: formatted_messages,
+            },
+        )
+        scope._span.add_event(
+            name=SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT,
+            attributes={
+                SemanticConvention.GEN_AI_CONTENT_COMPLETION: scope._llmresponse,
+            },
+        )
+
+    scope._span.set_status(Status(StatusCode.OK))
+
+    # Record metrics
+    if not disable_metrics:
+        record_completion_metrics(metrics, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT, SemanticConvention.GEN_AI_SYSTEM_AWS_BEDROCK,
+            scope._server_address, scope._server_port, request_model, scope._response_model, environment,
+            application_name, scope._start_time, scope._end_time, scope._input_tokens, scope._output_tokens,
+            cost, scope._tbt, scope._ttft)
+
+def process_streaming_chat_response(scope, pricing_info, environment, application_name, metrics,
+                                    capture_message_content=False, disable_metrics=False, version="", llm_config=None):
+    """
+    Process streaming chat response and generate telemetry.
+    """
+
+    try:
+        if llm_config is None:
+            llm_config = {}
+
+        common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+                            capture_message_content, disable_metrics, version, llm_config, is_stream=True)
+    except Exception as e:
+        handle_exception(scope._span, e)
+        raise
+
+def process_chat_response(response, request_model, pricing_info, server_port, server_address, environment,
+                          application_name, metrics, start_time, span, capture_message_content=False,
+                          disable_metrics=False, version="1.0.0", llm_config=None, **kwargs):
+    """
+    Process non-streaming chat response and generate telemetry.
+    """
+
+    try:
+        if llm_config is None:
+            llm_config = {}
+
+        scope = type("GenericScope", (), {})()
+        response_dict = response_as_dict(response)
+
+        scope._start_time = start_time
+        scope._end_time = time.time()
+        scope._span = span
+        scope._llmresponse = response_dict.get("output", {}).get("message", {}).get("content", [{}])[0].get("text", "")
+        scope._response_role = response_dict.get("output", {}).get("message", {}).get("role", "assistant")
+        scope._input_tokens = response_dict.get("usage", {}).get("inputTokens", 0)
+        scope._output_tokens = response_dict.get("usage", {}).get("outputTokens", 0)
+        scope._response_model = request_model
+        scope._finish_reason = response_dict.get("stopReason", "")
+        scope._response_id = response_dict.get("RequestId", "")
+        scope._timestamps = []
+        scope._ttft, scope._tbt = scope._end_time - scope._start_time, 0
+        scope._server_address, scope._server_port = server_address, server_port
+        scope._kwargs = kwargs
+
+        common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+                            capture_message_content, disable_metrics, version, llm_config, is_stream=False)
+
+        return response
+    except Exception as e:
+        handle_exception(span, e)
+        raise

{openlit-1.34.19 → openlit-1.34.22}/src/openlit/instrumentation/litellm/async_litellm.py

@@ -68,9 +68,9 @@ def acompletion(version, environment, application_name, tracer, pricing_info,
         def __aiter__(self):
             return self
 
-        def __getattr__(self, name):
+        async def __getattr__(self, name):
             """Delegate attribute access to the wrapped object."""
-            return getattr(self.__wrapped__, name)
+            return getattr(await self.__wrapped__, name)
 
         async def __anext__(self):
             try: