openlit 1.34.12__py3-none-any.whl → 1.34.14__py3-none-any.whl

openlit/instrumentation/transformers/__init__.py

@@ -30,12 +30,19 @@ class TransformersInstrumentor(BaseInstrumentor):
         version = importlib.metadata.version("transformers")
 
         wrap_function_wrapper(
-            "transformers",  
-            "TextGenerationPipeline.__call__",  
-            pipeline_wrapper(version, environment, application_name,
-                 tracer, pricing_info, capture_message_content, metrics, disable_metrics),
+            "transformers",
+            "TextGenerationPipeline.__call__",
+            pipeline_wrapper(
+                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/transformers/transformers.py

@@ -2,22 +2,15 @@
 Module for monitoring HF Transformers API calls.
 """
 
-import logging
 import time
 from opentelemetry.trace import SpanKind
-from openlit.__helpers import (
-    set_server_address_and_port
-)
-from openlit.instrumentation.transformers.utils import (
-    process_chat_response,
-)
+from openlit.__helpers import set_server_address_and_port
+from openlit.instrumentation.transformers.utils import process_chat_response
 from openlit.semcov import SemanticConvention
 
-# Initialize logger for logging potential issues and operations
-logger = logging.getLogger(__name__)
 
-def pipeline_wrapper(version, environment, application_name,
-    tracer, pricing_info, capture_message_content, metrics, disable_metrics):
+def pipeline_wrapper(version, environment, application_name, tracer, pricing_info,
+                     capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -32,27 +25,27 @@ def pipeline_wrapper(version, environment, application_name,
 
         span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
-        with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
+        with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
             start_time = time.time()
             response = wrapped(*args, **kwargs)
 
             response = process_chat_response(
-                    instance = instance,
-                    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,
-                    args=args,
-                    kwargs=kwargs,
-                    capture_message_content=capture_message_content,
-                    disable_metrics=disable_metrics,
-                    version=version,
+                instance=instance,
+                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,
+                args=args,
+                kwargs=kwargs,
+                capture_message_content=capture_message_content,
+                disable_metrics=disable_metrics,
+                version=version,
             )
 
         return response

openlit/instrumentation/transformers/utils.py

@@ -3,19 +3,61 @@ HF Transformers OpenTelemetry instrumentation utility functions
 """
 import time
 
-from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
 from opentelemetry.trace import Status, StatusCode
 
 from openlit.__helpers import (
-    response_as_dict,
-    calculate_tbt,
     general_tokens,
     get_chat_model_cost,
-    create_metrics_attributes,
-    format_and_concatenate
+    common_span_attributes,
+    record_completion_metrics,
 )
 from openlit.semcov import SemanticConvention
 
+def format_content(content):
+    """
+    Format content to a consistent structure.
+    """
+    if isinstance(content, str):
+        return content
+    elif isinstance(content, list):
+        # Check if its a list of chat messages (like in the test case)
+        if (len(content) > 0 and isinstance(content[0], dict) and
+            "role" in content[0] and "content" in content[0]):
+            # Handle chat message format like Groq
+            formatted_messages = []
+            for message in content:
+                role = message["role"]
+                msg_content = message["content"]
+
+                if isinstance(msg_content, list):
+                    content_str = ", ".join(
+                        f'{item["type"]}: {item["text"] if "text" in item else item.get("image_url", str(item))}'
+                        if isinstance(item, dict) and "type" in item
+                        else str(item)
+                        for item in msg_content
+                    )
+                    formatted_messages.append(f"{role}: {content_str}")
+                else:
+                    formatted_messages.append(f"{role}: {msg_content}")
+            return "\n".join(formatted_messages)
+        else:
+            # Handle other list formats (transformers responses)
+            formatted_content = []
+            for item in content:
+                if isinstance(item, str):
+                    formatted_content.append(item)
+                elif isinstance(item, dict):
+                    # Handle dict format for transformers
+                    if "generated_text" in item:
+                        formatted_content.append(str(item["generated_text"]))
+                    else:
+                        formatted_content.append(str(item))
+                else:
+                    formatted_content.append(str(item))
+            return " ".join(formatted_content)
+    else:
+        return str(content)
+
 def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
     capture_message_content, disable_metrics, version, args, kwargs, is_stream):
 
@@ -24,56 +66,42 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
     """
 
     scope._end_time = time.time()
-    if len(scope._timestamps) > 1:
-        scope._tbt = calculate_tbt(scope._timestamps)
-
     forward_params = scope._instance._forward_params
     request_model = scope._instance.model.config.name_or_path
 
     input_tokens = general_tokens(scope._prompt)
-    output_tokens = general_tokens(scope._llmresponse)
+    output_tokens = general_tokens(scope._completion)
 
     cost = get_chat_model_cost(request_model, pricing_info, input_tokens, output_tokens)
 
-    # Set Span attributes (OTel Semconv)
-    scope._span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
-    scope._span.set_attribute(SemanticConvention.GEN_AI_OPERATION, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SYSTEM, SemanticConvention.GEN_AI_SYSTEM_HUGGING_FACE)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MODEL, request_model)
-    scope._span.set_attribute(SemanticConvention.SERVER_PORT, scope._server_port)
-
-    # List of attributes and their config keys
-    attributes = [
-        (SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, "temperature"),
-        (SemanticConvention.GEN_AI_REQUEST_TOP_K, "top_k"),
-        (SemanticConvention.GEN_AI_REQUEST_TOP_P, "top_p"),
-        (SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, "max_length"),
-    ]
-
-    # Set each attribute if the corresponding value exists and is not None
-    for attribute, key in attributes:
-        value = forward_params.get(key)
-        if value is not None:
-            scope._span.set_attribute(attribute, value)
-
-    scope._span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_MODEL, request_model)
+    # Common Span Attributes
+    common_span_attributes(scope,
+        SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT, SemanticConvention.GEN_AI_SYSTEM_HUGGING_FACE,
+        scope._server_address, scope._server_port, request_model, request_model,
+        environment, application_name, is_stream, scope._tbt, scope._ttft, version)
+
+    # Set request parameters from forward_params
+    if forward_params.get("temperature") is not None:
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, forward_params["temperature"])
+    if forward_params.get("top_k") is not None:
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_K, forward_params["top_k"])
+    if forward_params.get("top_p") is not None:
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_P, forward_params["top_p"])
+    if forward_params.get("max_length") is not None:
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, forward_params["max_length"])
+
+    # Set token usage and cost attributes
     scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, input_tokens)
     scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, output_tokens)
-    scope._span.set_attribute(SemanticConvention.SERVER_ADDRESS, scope._server_address)
-    scope._span.set_attribute(DEPLOYMENT_ENVIRONMENT, environment)
-    scope._span.set_attribute(SERVICE_NAME, application_name)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_IS_STREAM, is_stream)
     scope._span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE, input_tokens + output_tokens)
     scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SERVER_TBT, scope._tbt)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SERVER_TTFT, scope._ttft)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SDK_VERSION, version)
 
-    # To be removed one the change to span_attributes (from span events) is complete
+    # Span Attributes for Content
     if capture_message_content:
         scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_PROMPT, scope._prompt)
-        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_COMPLETION, scope._llmresponse)
+        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_COMPLETION, scope._completion)
 
+        # 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={
@@ -83,32 +111,18 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
         scope._span.add_event(
             name=SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT,
             attributes={
-                SemanticConvention.GEN_AI_CONTENT_COMPLETION: scope._llmresponse,
+                SemanticConvention.GEN_AI_CONTENT_COMPLETION: scope._completion,
             },
         )
 
     scope._span.set_status(Status(StatusCode.OK))
 
+    # Record metrics using the standardized helper function
     if not disable_metrics:
-        metrics_attributes = create_metrics_attributes(
-            service_name=application_name,
-            deployment_environment=environment,
-            operation=SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
-            system=SemanticConvention.GEN_AI_SYSTEM_HUGGING_FACE,
-            request_model=request_model,
-            server_address=scope._server_address,
-            server_port=scope._server_port,
-            response_model=request_model,
-        )
-
-        metrics["genai_client_usage_tokens"].record(input_tokens + output_tokens, metrics_attributes)
-        metrics["genai_client_operation_duration"].record(scope._end_time - scope._start_time, metrics_attributes)
-        metrics["genai_server_tbt"].record(scope._tbt, metrics_attributes)
-        metrics["genai_server_ttft"].record(scope._ttft, metrics_attributes)
-        metrics["genai_requests"].add(1, metrics_attributes)
-        metrics["genai_completion_tokens"].add(output_tokens, metrics_attributes)
-        metrics["genai_prompt_tokens"].add(input_tokens, metrics_attributes)
-        metrics["genai_cost"].record(cost, metrics_attributes)
+        record_completion_metrics(metrics, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
+            SemanticConvention.GEN_AI_SYSTEM_HUGGING_FACE, scope._server_address, scope._server_port,
+            request_model, request_model, environment, application_name, scope._start_time, scope._end_time,
+            cost, input_tokens, output_tokens, scope._tbt, scope._ttft)
 
 def process_chat_response(instance, response, request_model, pricing_info, server_port, server_address,
     environment, application_name, metrics, start_time,
@@ -117,67 +131,69 @@ def process_chat_response(instance, response, request_model, pricing_info, serve
     Process chat request and generate Telemetry
     """
 
-    self = type("GenericScope", (), {})()
-    response_dict = response_as_dict(response)
-
-    # pylint: disable = no-member
-    self._instance = instance
-    self._start_time = start_time
-    self._end_time = time.time()
-    self._span = span
-    self._timestamps = []
-    self._ttft, self._tbt = self._end_time - self._start_time, 0
-    self._server_address, self._server_port = server_address, server_port
-    self._kwargs = kwargs
-    self._args = args
-
-    if self._args and len(self._args) > 0:
-        self._prompt = args[0]
+    scope = type("GenericScope", (), {})()
+    scope._instance = instance
+    scope._start_time = start_time
+    scope._end_time = time.time()
+    scope._span = span
+    scope._server_address = server_address
+    scope._server_port = server_port
+    scope._kwargs = kwargs
+    scope._args = args
+
+    # Extract prompt from args or kwargs
+    if args and len(args) > 0:
+        scope._prompt = args[0]
     else:
-        self._prompt = (
+        scope._prompt = (
             kwargs.get("text_inputs") or
             (kwargs.get("image") and kwargs.get("question") and
-            ("image: " + kwargs.get("image") + " question:" + kwargs.get("question"))) or
+             ("image: " + kwargs.get("image") + " question:" + kwargs.get("question"))) or
             kwargs.get("fallback") or
             ""
         )
-    self._prompt = format_and_concatenate(self._prompt)
-
-    self._llmresponse = []
-    if self._kwargs.get("task", "text-generation") == "text-generation":
-        first_entry = response_dict[0]
-
-        if isinstance(first_entry, dict) and isinstance(first_entry.get("generated_text"), list):
-            last_element = first_entry.get("generated_text")[-1]
-            self._llmresponse = last_element.get("content", last_element)
+    scope._prompt = format_content(scope._prompt)
+
+    # Process response based on task type
+    task = kwargs.get("task", "text-generation")
+
+    if task == "text-generation":
+        # Handle text generation responses
+        if isinstance(response, list) and len(response) > 0:
+            first_entry = response[0]
+            if isinstance(first_entry, dict):
+                if isinstance(first_entry.get("generated_text"), list):
+                    # Handle nested list format
+                    last_element = first_entry.get("generated_text")[-1]
+                    scope._completion = last_element.get("content", str(last_element))
+                else:
+                    # Handle standard format
+                    scope._completion = first_entry.get("generated_text", "")
+            else:
+                scope._completion = str(first_entry)
         else:
-            def extract_text(entry):
-                if isinstance(entry, dict):
-                    return entry.get("generated_text")
-                if isinstance(entry, list):
-                    return " ".join(
-                        extract_text(sub_entry) for sub_entry in entry if isinstance(sub_entry, dict)
-                    )
-                return ""
-
-            # Process and collect all generated texts
-            self._llmresponse = [
-                extract_text(entry) for entry in response_dict
-            ]
+            scope._completion = ""
 
-            # Join all non-empty responses into a single string
-            self._llmresponse = " ".join(filter(None, self._llmresponse))
+    elif task == "automatic-speech-recognition":
+        scope._completion = response.get("text", "") if isinstance(response, dict) else ""
 
-    elif self._kwargs.get("task", "text-generation") == "automatic-speech-recognition":
-        self._llmresponse = response_dict.get("text", "")
+    elif task == "image-classification":
+        scope._completion = str(response[0]) if isinstance(response, list) and len(response) > 0 else ""
 
-    elif self._kwargs.get("task", "text-generation") == "image-classification":
-        self._llmresponse = str(response_dict[0])
+    elif task == "visual-question-answering":
+        if isinstance(response, list) and len(response) > 0 and isinstance(response[0], dict):
+            scope._completion = response[0].get("answer", "")
+        else:
+            scope._completion = ""
+    else:
+        # Default handling for other tasks
+        scope._completion = format_content(response)
 
-    elif self._kwargs.get("task", "text-generation") == "visual-question-answering":
-        self._llmresponse = str(response_dict[0]).get("answer")
+    # Initialize timing attributes
+    scope._tbt = 0
+    scope._ttft = scope._end_time - scope._start_time
 
-    common_chat_logic(self, pricing_info, environment, application_name, metrics,
-            capture_message_content, disable_metrics, version, args, kwargs, is_stream=False)
+    common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+                     capture_message_content, disable_metrics, version, args, kwargs, is_stream=False)
 
     return response

openlit/instrumentation/vllm/__init__.py

@@ -1,4 +1,3 @@
-# pylint: disable=useless-return, bad-staticmethod-argument, disable=duplicate-code
 """Initializer of Auto Instrumentation of vLLM Functions"""
 
 from typing import Collection
@@ -14,15 +13,15 @@ _instruments = ("vllm >= 0.5.4",)
 
 class VLLMInstrumentor(BaseInstrumentor):
     """
-    An instrumentor for vLLM's client library.
+    An instrumentor for vLLM 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")
+        application_name = kwargs.get("application_name", "default")
+        environment = kwargs.get("environment", "default")
         tracer = kwargs.get("tracer")
         metrics = kwargs.get("metrics_dict")
         pricing_info = kwargs.get("pricing_info", {})
@@ -30,14 +29,13 @@ class VLLMInstrumentor(BaseInstrumentor):
         disable_metrics = kwargs.get("disable_metrics")
         version = importlib.metadata.version("vllm")
 
-        # sync chat
+        # Chat completions
         wrap_function_wrapper(
             "vllm.entrypoints.llm",
             "LLM.generate",
             generate(version, environment, application_name,
-                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
+                     tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
     def _uninstrument(self, **kwargs):
-        # Proper uninstrumentation logic to revert patched methods
         pass

openlit/instrumentation/vllm/utils.py

@@ -1,15 +1,15 @@
 """
-Utility functions for vLLM instrumentation.
+vLLM OpenTelemetry instrumentation utility functions
 """
-
 import time
-from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+
 from opentelemetry.trace import Status, StatusCode
+
 from openlit.__helpers import (
-    calculate_tbt,
-    get_chat_model_cost,
     general_tokens,
-    create_metrics_attributes,
+    get_chat_model_cost,
+    common_span_attributes,
+    record_completion_metrics,
 )
 from openlit.semcov import SemanticConvention
 
@@ -24,77 +24,81 @@ def get_inference_config(args, kwargs):
         return args[1]
     return None
 
+def format_content(prompts):
+    """
+    Process a list of prompts to extract content.
+    """
+
+    if isinstance(prompts, str):
+        return prompts
+    elif isinstance(prompts, list):
+        return "\n".join(str(prompt) for prompt in prompts)
+    else:
+        return str(prompts)
+
 def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
     capture_message_content, disable_metrics, version, is_stream):
     """
     Process chat request and generate Telemetry
     """
 
-    scope._end_time = time.time()
-    if len(scope._timestamps) > 1:
-        scope._tbt = calculate_tbt(scope._timestamps)
-
-    # Set base span attributes
-    scope._span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
-    scope._span.set_attribute(SemanticConvention.GEN_AI_OPERATION, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SYSTEM, SemanticConvention.GEN_AI_SYSTEM_VLLM)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MODEL, scope._request_model)
-    scope._span.set_attribute(SemanticConvention.SERVER_PORT, scope._server_port)
-    scope._span.set_attribute(SemanticConvention.SERVER_ADDRESS, scope._server_address)
-
-    # Handle inference configuration
+    request_model = scope._request_model
+
+    # Extract prompts and completions from vLLM response
+    input_tokens = 0
+    output_tokens = 0
+    prompt = ""
+    completion = ""
+
+    for output in scope._response:
+        prompt += output.prompt + "\n"
+        if output.outputs and len(output.outputs) > 0:
+            completion += output.outputs[0].text + "\n"
+        input_tokens += general_tokens(output.prompt)
+        output_tokens += general_tokens(output.outputs[0].text)
+
+    cost = get_chat_model_cost(request_model, pricing_info, input_tokens, output_tokens)
+
+    # Common Span Attributes
+    common_span_attributes(scope,
+        SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT, SemanticConvention.GEN_AI_SYSTEM_VLLM,
+        scope._server_address, scope._server_port, request_model, request_model,
+        environment, application_name, is_stream, scope._tbt, scope._ttft, version)
+
+    # Span Attributes for Request parameters
     inference_config = get_inference_config(scope._args, scope._kwargs)
     if inference_config:
-        attributes = [
-            (SemanticConvention.GEN_AI_REQUEST_FREQUENCY_PENALTY, 'frequency_penalty'),
-            (SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, 'max_tokens'),
-            (SemanticConvention.GEN_AI_REQUEST_PRESENCE_PENALTY, 'presence_penalty'),
-            (SemanticConvention.GEN_AI_REQUEST_STOP_SEQUENCES, 'stop_sequences'),
-            (SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, 'temperature'),
-            (SemanticConvention.GEN_AI_REQUEST_TOP_P, 'top_p'),
-            (SemanticConvention.GEN_AI_REQUEST_TOP_K, 'top_k'),
-        ]
-
-        for attribute, key in attributes:
-            value = getattr(inference_config, key, None)
-            if value is not None:
-                scope._span.set_attribute(attribute, value)
-
-    scope._span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_MODEL, scope._request_model)
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS, getattr(inference_config, 'max_tokens', -1))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_STOP_SEQUENCES, getattr(inference_config, 'stop_sequences', []))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TEMPERATURE, getattr(inference_config, 'temperature', 1.0))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_P, getattr(inference_config, 'top_p', 1.0))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_K, getattr(inference_config, 'top_k', -1))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_PRESENCE_PENALTY,
+            getattr(inference_config, 'presence_penalty', 0.0))
+        scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+            getattr(inference_config, 'frequency_penalty', 0.0))
+
+    # Span Attributes for Response parameters
     scope._span.set_attribute(SemanticConvention.GEN_AI_OUTPUT_TYPE, "text")
 
-    # Set base span attributes (Extras)
-    scope._span.set_attribute(DEPLOYMENT_ENVIRONMENT, environment)
-    scope._span.set_attribute(SERVICE_NAME, application_name)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_REQUEST_IS_STREAM, is_stream)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SERVER_TBT, scope._tbt)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SERVER_TTFT, scope._ttft)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_SDK_VERSION, version)
-
-    input_tokens = 0
-    output_tokens = 0
-    cost = 0
+    # Span Attributes for Cost and Tokens
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS, input_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS, output_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE, input_tokens + output_tokens)
+    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
 
+    # Span Attributes for Content
     if capture_message_content:
-        prompt = ""
-        completion = ""
+        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_PROMPT, prompt)
+        scope._span.set_attribute(SemanticConvention.GEN_AI_CONTENT_COMPLETION, completion)
 
-        for output in scope._response:
-            prompt += output.prompt + "\n"
-            if output.outputs and len(output.outputs) > 0:
-                completion += output.outputs[0].text + "\n"
-            input_tokens += general_tokens(output.prompt)
-            output_tokens += general_tokens(output.outputs[0].text)
-
-        # Add a single event for prompt
+        # 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: prompt,
             },
         )
-
-        # Add a single event for completion
         scope._span.add_event(
             name=SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT,
             attributes={
@@ -102,39 +106,14 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
             },
         )
 
-    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS,
-                        input_tokens)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS,
-                        output_tokens)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_CLIENT_TOKEN_USAGE,
-                        input_tokens + output_tokens)
-
-    # Calculate cost of the operation
-    cost = get_chat_model_cost(scope._request_model, pricing_info, input_tokens, output_tokens)
-    scope._span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST, cost)
-
     scope._span.set_status(Status(StatusCode.OK))
 
-    if disable_metrics is False:
-        metrics_attributes = create_metrics_attributes(
-            service_name=application_name,
-            deployment_environment=environment,
-            operation=SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT,
-            system=SemanticConvention.GEN_AI_SYSTEM_VLLM,
-            request_model=scope._request_model,
-            server_address=scope._server_address,
-            server_port=scope._server_port,
-            response_model=scope._request_model,
-        )
-        metrics['genai_client_operation_duration'].record(scope._end_time - scope._start_time, metrics_attributes)
-        metrics['genai_server_tbt'].record(scope._tbt, metrics_attributes)
-        metrics['genai_server_ttft'].record(scope._ttft, metrics_attributes)
-        metrics['genai_requests'].add(1, metrics_attributes)
-        metrics['genai_completion_tokens'].add(output_tokens, metrics_attributes)
-        metrics['genai_prompt_tokens'].add(input_tokens, metrics_attributes)
-        metrics['genai_cost'].record(cost, metrics_attributes)
-        metrics['genai_client_usage_tokens'].record(
-            input_tokens + output_tokens, metrics_attributes)
+    # Metrics
+    if not disable_metrics:
+        record_completion_metrics(metrics, SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT, SemanticConvention.GEN_AI_SYSTEM_VLLM,
+            scope._server_address, scope._server_port, request_model, request_model, environment,
+            application_name, scope._start_time, scope._end_time, input_tokens, output_tokens,
+            cost, scope._tbt, scope._ttft)
 
 def process_chat_response(instance, response, request_model, pricing_info, server_port, server_address,
     environment, application_name, metrics, start_time, span, args, kwargs,
@@ -142,20 +121,23 @@ def process_chat_response(instance, response, request_model, pricing_info, serve
     """
     Process chat request and generate Telemetry
     """
-    self = type('GenericScope', (), {})()
-    self._response = response
-    self._start_time = start_time
-    self._end_time = time.time()
-    self._span = span
-    self._ttft, self._tbt = self._end_time - self._start_time, 0
-    self._server_address = server_address
-    self._server_port = server_port
-    self._request_model = request_model
-    self._timestamps = []
-    self._args = args
-    self._kwargs = kwargs
-
-    common_chat_logic(self, pricing_info, environment, application_name, metrics,
+
+    # Create scope object
+    scope = type("GenericScope", (), {})()
+
+    scope._response = response
+    scope._start_time = start_time
+    scope._end_time = time.time()
+    scope._span = span
+    scope._ttft, scope._tbt = scope._end_time - scope._start_time, 0
+    scope._server_address = server_address
+    scope._server_port = server_port
+    scope._request_model = request_model
+    scope._timestamps = []
+    scope._args = args
+    scope._kwargs = kwargs
+
+    common_chat_logic(scope, pricing_info, environment, application_name, metrics,
         capture_message_content, disable_metrics, version, is_stream=False)
 
     return response

openlit/instrumentation/vllm/vllm.py

@@ -2,7 +2,6 @@
 Module for monitoring vLLM API calls.
 """
 
-import logging
 import time
 from opentelemetry.trace import SpanKind
 from openlit.__helpers import (
@@ -14,11 +13,8 @@ from openlit.instrumentation.vllm.utils import (
 )
 from openlit.semcov import SemanticConvention
 
-# Initialize logger for logging potential issues and operations
-logger = logging.getLogger(__name__)
-
-def generate(version, environment, application_name,
-    tracer, pricing_info, capture_message_content, metrics, disable_metrics):
+def generate(version, environment, application_name, tracer, pricing_info,
+             capture_message_content, metrics, disable_metrics):
     """
     Generates a telemetry wrapper for GenAI function call
     """
@@ -27,7 +23,6 @@ def generate(version, environment, application_name,
         """
         Wraps the GenAI function call.
         """
-
         server_address, server_port = set_server_address_and_port(instance, "http://127.0.0.1", 443)
         request_model = instance.llm_engine.model_config.model or "facebook/opt-125m"
 
@@ -56,9 +51,9 @@ def generate(version, environment, application_name,
                     disable_metrics=disable_metrics,
                     version=version,
                 )
+
             except Exception as e:
                 handle_exception(span, e)
-                logger.error("Error in trace creation: %s", e)
 
             return response
 

{openlit-1.34.12.dist-info → openlit-1.34.14.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: openlit
-Version: 1.34.12
+Version: 1.34.14
 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-1.34.12.dist-info → openlit-1.34.14.dist-info}/RECORD

@@ -125,20 +125,20 @@ openlit/instrumentation/together/__init__.py,sha256=0UmUqQtppyK3oopb4lTjX2LITgVC
 openlit/instrumentation/together/async_together.py,sha256=0-h5fKw6rIwN_fvWVpGuvVqizIuM9xFCzz8Z4oGgOj0,6822
 openlit/instrumentation/together/together.py,sha256=nY6mzHmHgoMbbnB_9eL0EBQjP0ltJVdkQj4pbamHAj0,6723
 openlit/instrumentation/together/utils.py,sha256=n7r_pM_sqFnJEAkL7OhPydr0Uct0A74vXdcYELdbeW0,14368
-openlit/instrumentation/transformers/__init__.py,sha256=9Ubss5nlumcypxprxff8Fv3sst7II27SsvCzqkBX9Kg,1457
-openlit/instrumentation/transformers/transformers.py,sha256=y--t7PXhUfPC81w-aEE7qowMah3os9gnKBQ5bN4QLGc,1980
-openlit/instrumentation/transformers/utils.py,sha256=3f-ewpUpduaBrTVIFJKaabACjz-6Vf8K7NEU0EzQ4Nk,8042
+openlit/instrumentation/transformers/__init__.py,sha256=hXq0WUZNl6Sz0Ihk29kA9i8Q1j0e1URFb7v7etnQpxI,1511
+openlit/instrumentation/transformers/transformers.py,sha256=MHnHVo_6NP0gSIqxen6qQpCrZ0fs8Ec80EdZumMpVNo,1797
+openlit/instrumentation/transformers/utils.py,sha256=MMy_SyRyDI4X-0mqbBwStac0xabmw0ZRvv_VWLA_Nkg,8426
 openlit/instrumentation/vertexai/__init__.py,sha256=mT28WCBvQfRCkAWGL6bd0EjEPHvMjaNcz6T3jsLZh8k,3745
 openlit/instrumentation/vertexai/async_vertexai.py,sha256=-kpg-eiL76O5_XopUPghCYwJHf0Nrxi00_Z5tCwq6zM,23086
 openlit/instrumentation/vertexai/vertexai.py,sha256=5NB090aWlm9DnlccNNLRO6A97P_RN-JnHb5JS01tYyw,23000
-openlit/instrumentation/vllm/__init__.py,sha256=VUWsjtYEe1_u4hJRDAZI5abrRfZ2L85LxZmc76irBrk,1524
-openlit/instrumentation/vllm/utils.py,sha256=hPVG_UKLY7xTvmmHbBdPy8HT7y_8VIILn37a5zOTYzU,6822
-openlit/instrumentation/vllm/vllm.py,sha256=SZosSwnkBUKspPtsm_k6VQaAWnD4kdcFWj2n-StWJus,2175
+openlit/instrumentation/vllm/__init__.py,sha256=uaSzQmgDuKJ-sh61sfVdzVt2qAZaozZIQ8sbmQ0XpZE,1357
+openlit/instrumentation/vllm/utils.py,sha256=HuCPNBgChWg9vA7DHNFCij_y8qj27DjZxdZ0Nvdt2fg,5751
+openlit/instrumentation/vllm/vllm.py,sha256=VzazF2f4LLwjZDO_G8lIN_d622oSJM0fIO9wjxXbhyg,2004
 openlit/otel/events.py,sha256=VrMjTpvnLtYRBHCiFwJojTQqqNpRCxoD4yJYeQrtPsk,3560
 openlit/otel/metrics.py,sha256=GM2PDloBGRhBTkHHkYaqmOwIAQkY124ZhW4sEqW1Fgk,7086
 openlit/otel/tracing.py,sha256=tjV2bEbEDPUB1Z46gE-UsJsb04sRdFrfbhIDkxViZc0,3103
 openlit/semcov/__init__.py,sha256=ptyo37PY-FHDx_PShEvbdns71cD4YvvXw15bCRXKCKM,13461
-openlit-1.34.12.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-openlit-1.34.12.dist-info/METADATA,sha256=1GWoYLlOr6o5nSG3t4_n58xc9QJjcTSPUGNEGJJfEn8,23470
-openlit-1.34.12.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
-openlit-1.34.12.dist-info/RECORD,,
+openlit-1.34.14.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+openlit-1.34.14.dist-info/METADATA,sha256=qaOh__y9R5tT0z7qveai1LH4KWY7ampN3PzVhhm7D0M,23470
+openlit-1.34.14.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+openlit-1.34.14.dist-info/RECORD,,