openlit 1.33.16__tar.gz → 1.33.18__tar.gz

{openlit-1.33.16 → openlit-1.33.18}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: openlit
-Version: 1.33.16
+Version: 1.33.18
 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-1.33.16 → openlit-1.33.18}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "openlit"
-version = "1.33.16"
+version = "1.33.18"
 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-1.33.16 → openlit-1.33.18}/src/openlit/instrumentation/azure_ai_inference/__init__.py

@@ -4,13 +4,11 @@ from typing import Collection
 import importlib.metadata
 from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
 from wrapt import wrap_function_wrapper
-
 from openlit.instrumentation.azure_ai_inference.azure_ai_inference import (
-    complete, embedding
+    complete
 )
-
 from openlit.instrumentation.azure_ai_inference.async_azure_ai_inference import (
-    async_complete, async_embedding
+    async_complete
 )
 
 _instruments = ('azure-ai-inference >= 1.0.0b4',)
@@ -27,6 +25,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
         application_name = kwargs.get('application_name', 'default')
         environment = kwargs.get('environment', '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)
@@ -38,15 +37,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
             'azure.ai.inference',
             'ChatCompletionsClient.complete',
             complete(version, environment, application_name,
-                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
-        )
-
-        # sync embedding
-        wrap_function_wrapper(
-            'azure.ai.inference',
-            'EmbeddingsClient.embed',
-            embedding(version, environment, application_name,
-                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
         # async generate
@@ -54,15 +45,7 @@ class AzureAIInferenceInstrumentor(BaseInstrumentor):
             'azure.ai.inference.aio',
             'ChatCompletionsClient.complete',
             async_complete(version, environment, application_name,
-                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
-        )
-
-        # async embedding
-        wrap_function_wrapper(
-            'azure.ai.inference.aio',
-            'EmbeddingsClient.embed',
-            async_embedding(version, environment, application_name,
-                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
+                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
     def _uninstrument(self, **kwargs):

openlit-1.33.18/src/openlit/instrumentation/azure_ai_inference/async_azure_ai_inference.py

@@ -0,0 +1,144 @@
+"""
+Module for monitoring Azure AI Inference 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.azure_ai_inference.utils import (
+    process_chunk,
+    process_chat_response,
+    process_streaming_chat_response,
+)
+from openlit.semcov import SemanticConvetion
+
+# Initialize logger for logging potential issues and operations
+logger = logging.getLogger(__name__)
+
+def async_complete(version, environment, application_name,
+             tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+    """
+    Generates a telemetry wrapper for GenAI function call
+    """
+
+    class TracedAsyncStream:
+        """
+        Wrapper for streaming responses to collect telemetry.
+        """
+
+        def __init__(
+                self,
+                wrapped,
+                span,
+                span_name,
+                kwargs,
+                server_address,
+                server_port,
+                **args,
+            ):
+            self.__wrapped__ = wrapped
+            self._span = span
+            self._span_name = span_name
+            self._llmresponse = ""
+            self._response_id = ""
+            self._response_model = ""
+            self._finish_reason = ""
+            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
+
+        async def __aenter__(self):
+            await self.__wrapped__.__aenter__()
+            return self
+
+        async def __aexit__(self, exc_type, exc_value, traceback):
+            await self.__wrapped__.__aexit__(exc_type, exc_value, traceback)
+
+        def __aiter__(self):
+            return self
+
+        async def __getattr__(self, name):
+            """Delegate attribute access to the wrapped object."""
+            return getattr(await self.__wrapped__, name)
+
+        async def __anext__(self):
+            try:
+                chunk = await self.__wrapped__.__anext__()
+                process_chunk(self, chunk)
+                return chunk
+            except StopAsyncIteration:
+                try:
+                    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,
+                            event_provider=event_provider,
+                            capture_message_content=capture_message_content,
+                            disable_metrics=disable_metrics,
+                            version=version
+                        )
+
+                except Exception as e:
+                    handle_exception(self._span, e)
+                    logger.error("Error in trace creation: %s", e)
+                raise
+
+    async def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the GenAI function call.
+        """
+
+        streaming = kwargs.get("stream", False)
+        server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+        request_model = kwargs.get("model", "gpt-4o")
+
+        span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
+
+        # pylint: disable=no-else-return
+        if streaming:
+            awaited_wrapped = await wrapped(*args, **kwargs)
+            span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
+
+            return TracedAsyncStream(awaited_wrapped, span, span_name, kwargs, server_address, server_port)
+
+        else:
+            with tracer.start_as_current_span(span_name, kind=SpanKind.CLIENT) as span:
+                start_time = time.time()
+                response = await wrapped(*args, **kwargs)
+                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,
+                    **kwargs
+                )
+
+            return response
+
+    return wrapper

openlit-1.33.18/src/openlit/instrumentation/azure_ai_inference/azure_ai_inference.py

@@ -0,0 +1,144 @@
+"""
+Module for monitoring Azure AI Inference 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.azure_ai_inference.utils import (
+    process_chunk,
+    process_chat_response,
+    process_streaming_chat_response,
+)
+from openlit.semcov import SemanticConvetion
+
+# Initialize logger for logging potential issues and operations
+logger = logging.getLogger(__name__)
+
+def complete(version, environment, application_name,
+             tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics):
+    """
+    Generates a telemetry wrapper for GenAI function call
+    """
+
+    class TracedSyncStream:
+        """
+        Wrapper for streaming responses to collect telemetry.
+        """
+
+        def __init__(
+                self,
+                wrapped,
+                span,
+                span_name,
+                kwargs,
+                server_address,
+                server_port,
+                **args,
+            ):
+            self.__wrapped__ = wrapped
+            self._span = span
+            self._span_name = span_name
+            self._llmresponse = ""
+            self._response_id = ""
+            self._response_model = ""
+            self._finish_reason = ""
+            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):
+            self.__wrapped__.__enter__()
+            return self
+
+        def __exit__(self, exc_type, exc_value, traceback):
+            self.__wrapped__.__exit__(exc_type, exc_value, traceback)
+
+        def __iter__(self):
+            return self
+
+        def __getattr__(self, name):
+            """Delegate attribute access to the wrapped object."""
+            return getattr(self.__wrapped__, name)
+
+        def __next__(self):
+            try:
+                chunk = self.__wrapped__.__next__()
+                process_chunk(self, chunk)
+                return chunk
+            except StopIteration:
+                try:
+                    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,
+                            event_provider=event_provider,
+                            capture_message_content=capture_message_content,
+                            disable_metrics=disable_metrics,
+                            version=version
+                        )
+
+                except Exception as e:
+                    handle_exception(self._span, e)
+                    logger.error("Error in trace creation: %s", e)
+                raise
+
+    def wrapper(wrapped, instance, args, kwargs):
+        """
+        Wraps the GenAI function call.
+        """
+
+        streaming = kwargs.get("stream", False)
+        server_address, server_port = set_server_address_and_port(instance, "models.github.ai", 443)
+        request_model = kwargs.get("model", "gpt-4o")
+
+        span_name = f"{SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
+
+        # pylint: disable=no-else-return
+        if streaming:
+            awaited_wrapped = wrapped(*args, **kwargs)
+            span = tracer.start_span(span_name, kind=SpanKind.CLIENT)
+
+            return TracedSyncStream(awaited_wrapped, span, span_name, kwargs, server_address, server_port)
+
+        else:
+            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(
+                    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,
+                    **kwargs
+                )
+
+            return response
+
+    return wrapper

openlit-1.33.18/src/openlit/instrumentation/azure_ai_inference/utils.py

@@ -0,0 +1,225 @@
+"""
+Azure AI Inference 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_ttft,
+    response_as_dict,
+    calculate_tbt,
+    extract_and_format_input,
+    get_chat_model_cost,
+    create_metrics_attributes,
+    otel_event,
+    concatenate_all_contents
+)
+from openlit.semcov import SemanticConvetion
+
+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)
+
+    # Collect message IDs and aggregated response from events
+    if (len(chunked.get('choices')) > 0 and ('delta' in chunked.get('choices')[0] and
+        'content' in chunked.get('choices')[0].get('delta'))):
+
+        if content := chunked.get('choices')[0].get('delta').get('content'):
+            self._llmresponse += content
+
+    if chunked.get('choices')[0].get('finish_reason') is not None:
+        self._finish_reason = chunked.get('choices')[0].get('finish_reason')
+
+    if chunked.get('usage') is not None:
+        self._input_tokens = chunked.get('usage').get('prompt_tokens')
+        self._response_id = chunked.get('id')
+        self._response_model = chunked.get('model')
+        self._output_tokens = chunked.get('usage').get('completion_tokens')
+
+def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+                        event_provider, 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)
+
+    formatted_messages = extract_and_format_input(scope._kwargs.get('messages', ''))
+    request_model = scope._kwargs.get('model', 'claude-3-opus-20240229')
+
+    cost = get_chat_model_cost(request_model, pricing_info, scope._input_tokens, scope._output_tokens)
+
+    # Set Span attributes (OTel Semconv)
+    scope._span.set_attribute(TELEMETRY_SDK_NAME, 'openlit')
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_OPERATION, SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_SYSTEM, SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MODEL, request_model)
+    scope._span.set_attribute(SemanticConvetion.SERVER_PORT, scope._server_port)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_MAX_TOKENS, scope._kwargs.get('max_tokens', -1))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_STOP_SEQUENCES, scope._kwargs.get('stop', []))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TEMPERATURE, scope._kwargs.get('temperature', 1.0))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_K, scope._kwargs.get('top_k', 1.0))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_TOP_P, scope._kwargs.get('top_p', 1.0))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_FREQUENCY_PENALTY,
+                                                          scope._kwargs.get("frequency_penalty", 0.0))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_PRESENCE_PENALTY,
+                                                          scope._kwargs.get("presence_penalty", 0.0))
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_FINISH_REASON, [scope._finish_reason])
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_ID, scope._response_id)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_RESPONSE_MODEL, scope._response_model)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_INPUT_TOKENS, scope._input_tokens)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_OUTPUT_TOKENS, scope._output_tokens)
+    scope._span.set_attribute(SemanticConvetion.SERVER_ADDRESS, scope._server_address)
+
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_OUTPUT_TYPE,
+                              'text' if isinstance(scope._llmresponse, str) else 'json')
+
+    scope._span.set_attribute(DEPLOYMENT_ENVIRONMENT, environment)
+    scope._span.set_attribute(SERVICE_NAME, application_name)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_REQUEST_IS_STREAM, is_stream)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_CLIENT_TOKEN_USAGE, scope._input_tokens + scope._output_tokens)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_USAGE_COST, cost)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TBT, scope._tbt)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_SERVER_TTFT, scope._ttft)
+    scope._span.set_attribute(SemanticConvetion.GEN_AI_SDK_VERSION, version)
+
+    # To be removed one the change to log events (from span events) is complete
+    prompt = concatenate_all_contents(formatted_messages)
+    if capture_message_content:
+        scope._span.add_event(
+            name=SemanticConvetion.GEN_AI_CONTENT_PROMPT_EVENT,
+            attributes={
+                SemanticConvetion.GEN_AI_CONTENT_PROMPT: prompt,
+            },
+        )
+        scope._span.add_event(
+            name=SemanticConvetion.GEN_AI_CONTENT_COMPLETION_EVENT,
+            attributes={
+                SemanticConvetion.GEN_AI_CONTENT_COMPLETION: scope._llmresponse,
+            },
+        )
+
+    choice_event_body = {
+        'finish_reason': scope._finish_reason,
+        'index': 0,
+        'message': {
+            **({'content': scope._llmresponse} if capture_message_content else {}),
+            'role': 'assistant'
+        }
+    }
+
+    # Emit events
+    for role in ['user', 'system', 'assistant', 'tool']:
+        if formatted_messages.get(role, {}).get('content', ''):
+            event = otel_event(
+                name=getattr(SemanticConvetion, f'GEN_AI_{role.upper()}_MESSAGE'),
+                attributes={
+                    SemanticConvetion.GEN_AI_SYSTEM: SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE
+                },
+                body = {
+                    # pylint: disable=line-too-long
+                    **({'content': formatted_messages.get(role, {}).get('content', '')} if capture_message_content else {}),
+                    'role': formatted_messages.get(role, {}).get('role', []),
+                    **({
+                        'tool_calls': {
+                            'function': {
+                                # pylint: disable=line-too-long
+                                'name': (scope._tool_calls[0].get('function', {}).get('name', '') if scope._tool_calls else ''),
+                                'arguments': (scope._tool_calls[0].get('function', {}).get('arguments', '') if scope._tool_calls else '')
+                            },
+                            'id': (scope._tool_calls[0].get('id', '') if scope._tool_calls else ''),
+                            'type': 'function'
+                        }
+                    } if role == 'assistant' else {}),
+                    **({
+                        'id': (scope._tool_calls[0].get('id', '') if scope._tool_calls else '')
+                    } if role == 'tool' else {})
+                }
+            )
+            event_provider.emit(event)
+
+    choice_event = otel_event(
+        name=SemanticConvetion.GEN_AI_CHOICE,
+        attributes={
+            SemanticConvetion.GEN_AI_SYSTEM: SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE
+        },
+        body=choice_event_body
+    )
+    event_provider.emit(choice_event)
+
+    scope._span.set_status(Status(StatusCode.OK))
+
+    if not disable_metrics:
+        metrics_attributes = create_metrics_attributes(
+            service_name=application_name,
+            deployment_environment=environment,
+            operation=SemanticConvetion.GEN_AI_OPERATION_TYPE_CHAT,
+            system=SemanticConvetion.GEN_AI_SYSTEM_AZURE_AI_INFERENCE,
+            request_model=request_model,
+            server_address=scope._server_address,
+            server_port=scope._server_port,
+            response_model=scope._response_model,
+        )
+
+        metrics['genai_client_usage_tokens'].record(scope._input_tokens + scope._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(scope._output_tokens, metrics_attributes)
+        metrics['genai_prompt_tokens'].add(scope._input_tokens, metrics_attributes)
+        metrics['genai_cost'].record(cost, metrics_attributes)
+
+def process_streaming_chat_response(self, pricing_info, environment, application_name, metrics,
+                                    event_provider, capture_message_content=False, disable_metrics=False, version=''):
+    """
+    Process chat request and generate Telemetry
+    """
+
+    common_chat_logic(self, pricing_info, environment, application_name, metrics,
+                        event_provider, capture_message_content, disable_metrics, version, is_stream=True)
+
+def process_chat_response(response, request_model, pricing_info, server_port, server_address,
+                          environment, application_name, metrics, event_provider, start_time,
+                          span, capture_message_content=False, disable_metrics=False, version='1.0.0', **kwargs):
+    """
+    Process chat request and generate Telemetry
+    """
+
+    self = type('GenericScope', (), {})()
+    response_dict = response_as_dict(response)
+
+    # pylint: disable = no-member
+    self._start_time = start_time
+    self._end_time = time.time()
+    self._span = span
+    self._llmresponse = response_dict.get('choices', {})[0].get('message', '').get('content', '')
+    self._input_tokens = response_dict.get('usage').get('prompt_tokens')
+    self._output_tokens = response_dict.get('usage').get('completion_tokens')
+    self._response_model = response_dict.get('model', '')
+    self._finish_reason = response_dict.get('choices', {})[0].get('finish_reason', '')
+    self._response_id = response_dict.get('id', '')
+    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
+
+    common_chat_logic(self, pricing_info, environment, application_name, metrics,
+                        event_provider, capture_message_content, disable_metrics, version, is_stream=False)
+
+    return response

{openlit-1.33.16 → openlit-1.33.18}/src/openlit/instrumentation/langchain/langchain.py

@@ -14,6 +14,7 @@ from openlit.__helpers import (
     calculate_ttft,
     calculate_tbt,
     create_metrics_attributes,
+    extract_and_format_input,
 )
 from openlit.semcov import SemanticConvetion
 
@@ -249,34 +250,7 @@ def chat(gen_ai_endpoint, version, environment, application_name,
             end_time = time.time()
 
             try:
-                # Format 'messages' into a single string
-                message_prompt = kwargs.get("messages", "") or args[0]
-                formatted_messages = []
-
-                for message in message_prompt:
-                    # Handle the case where message is a tuple
-                    if isinstance(message, tuple) and len(message) == 2:
-                        role, content = message
-                    # Handle the case where message is a dictionary
-                    elif isinstance(message, dict):
-                        role = message["role"]
-                        content = message["content"]
-                    else:
-                        continue
-
-                    # Check if the content is a list
-                    if isinstance(content, list):
-                        content_str = ", ".join(
-                            f'{item["type"]}: {item["text"] if "text" in item else item["image_url"]}'
-                            if "type" in item else f'text: {item["text"]}'
-                            for item in content
-                        )
-                        formatted_messages.append(f"{role}: {content_str}")
-                    else:
-                        formatted_messages.append(f"{role}: {content}")
-
-                # Join all formatted messages with newline
-                prompt = "\n".join(formatted_messages)
+                prompt = str(kwargs.get('messages', '') or args[0])
 
                 input_tokens = general_tokens(str(prompt))
                 output_tokens = general_tokens(str(response))