openlit 1.33.20__py3-none-any.whl → 1.33.21__py3-none-any.whl

openlit/__helpers.py

@@ -240,6 +240,11 @@ def extract_and_format_input(messages):
     fixed_roles = ['user', 'assistant', 'system', 'tool', 'developer']
     formatted_messages = {role_key: {'role': '', 'content': ''} for role_key in fixed_roles}
 
+    # Check if input is a simple string
+    if isinstance(messages, str):
+        formatted_messages['user'] = {'role': 'user', 'content': messages}
+        return formatted_messages
+
     for message in messages:
         message = response_as_dict(message)
 
@@ -276,3 +281,55 @@ def concatenate_all_contents(formatted_messages):
         for message_data in formatted_messages.values()
         if message_data['content']
     )
+
+def format_and_concatenate(messages):
+    """
+    Process a list of messages to extract content, categorize them by role,
+    and concatenate all 'content' fields into a single string with role: content format.
+    """
+
+    formatted_messages = {}
+
+    # Check if input is a simple string
+    if isinstance(messages, str):
+        formatted_messages['user'] = {'role': 'user', 'content': messages}
+    elif isinstance(messages, list) and all(isinstance(m, str) for m in messages):
+        # If it's a list of strings, each string is 'user' input
+        user_content = ' '.join(messages)
+        formatted_messages['user'] = {'role': 'user', 'content': user_content}
+    else:
+        for message in messages:
+            message = response_as_dict(message)
+            role = message.get('role', 'unknown')  # Default to 'unknown' if no role is specified
+            content = message.get('content', '')
+
+            # Initialize role in formatted messages if not present
+            if role not in formatted_messages:
+                formatted_messages[role] = {'role': role, 'content': ''}
+
+            # Handle list of dictionaries in content
+            if isinstance(content, list):
+                content_str = []
+                for item in content:
+                    if isinstance(item, dict):
+                        # Collect text or other attributes as needed
+                        text = item.get('text', '')
+                        image_url = item.get('image_url', '')
+                        content_str.append(text)
+                        content_str.append(image_url)
+                content_str = ", ".join(filter(None, content_str))
+            else:
+                content_str = content
+
+            # Concatenate content
+            if formatted_messages[role]['content']:
+                formatted_messages[role]['content'] += ' ' + content_str
+            else:
+                formatted_messages[role]['content'] = content_str
+
+    # Concatenate role and content for all messages
+    return ' '.join(
+        f"{message_data['role']}: {message_data['content']}"
+        for message_data in formatted_messages.values()
+        if message_data['content']
+    )

openlit/instrumentation/ollama/__init__.py

@@ -16,6 +16,29 @@ from openlit.instrumentation.ollama.async_ollama import (
 
 _instruments = ("ollama >= 0.2.0",)
 
+# Dispatch wrapper to route instrumentation to chat or embeddings based on path
+def _dispatch(sync_chat_wrap, sync_emb_wrap):
+    def wrapper(wrapped, instance, args, kwargs):
+        if len(args) > 2 and isinstance(args[2], str):
+            op = args[2].rstrip("/").split("/")[-1]
+            if op == "chat":
+                return sync_chat_wrap(wrapped, instance, args, kwargs)
+            if op == "embeddings":
+                return sync_emb_wrap(wrapped, instance, args, kwargs)
+        return wrapped(*args, **kwargs)
+    return wrapper
+
+def _dispatch_async(async_chat_wrap, async_emb_wrap):
+    async def wrapper(wrapped, instance, args, kwargs):
+        if len(args) > 2 and isinstance(args[2], str):
+            op = args[2].rstrip("/").split("/")[-1]
+            if op == "chat":
+                return await async_chat_wrap(wrapped, instance, args, kwargs)
+            if op == "embeddings":
+                return await async_emb_wrap(wrapped, instance, args, kwargs)
+        return await wrapped(*args, **kwargs)
+    return wrapper
+
 class OllamaInstrumentor(BaseInstrumentor):
     """
     An instrumentor for Ollama's client library.
@@ -35,48 +58,38 @@ class OllamaInstrumentor(BaseInstrumentor):
         disable_metrics = kwargs.get("disable_metrics")
         version = importlib.metadata.version("ollama")
 
-        # sync chat
-        wrap_function_wrapper(
-            "ollama",
-            "chat",
-            chat(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+        # Build wrapper factories for chat and embeddings
+        sync_chat_wrap = chat(
+            version, environment, application_name,
+            tracer, event_provider, pricing_info,
+            capture_message_content, metrics, disable_metrics
         )
-        wrap_function_wrapper(
-            "ollama",
-            "Client.chat",
-            chat(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+        sync_emb_wrap = embeddings(
+            version, environment, application_name,
+            tracer, event_provider, pricing_info,
+            capture_message_content, metrics, disable_metrics
         )
-
-        # sync embeddings
-        wrap_function_wrapper(
-            "ollama",
-            "embeddings",
-            embeddings(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+        async_chat_wrap = async_chat(
+            version, environment, application_name,
+            tracer, event_provider, pricing_info,
+            capture_message_content, metrics, disable_metrics
         )
-        wrap_function_wrapper(
-            "ollama",
-            "Client.embeddings",
-            embeddings(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+        async_emb_wrap = async_embeddings(
+            version, environment, application_name,
+            tracer, event_provider, pricing_info,
+            capture_message_content, metrics, disable_metrics
         )
 
-        # async chat
+        # Patch underlying request methods to ensure instrumentation regardless of import order
         wrap_function_wrapper(
-            "ollama",
-            "AsyncClient.chat",
-            async_chat(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+            "ollama._client",
+            "Client._request",
+            _dispatch(sync_chat_wrap, sync_emb_wrap),
         )
-
-        # async embeddings
         wrap_function_wrapper(
-            "ollama",
-            "AsyncClient.embeddings",
-            async_embeddings(version, environment, application_name,
-                  tracer, event_provider, pricing_info, capture_message_content, metrics, disable_metrics),
+            "ollama._client",
+            "AsyncClient._request",
+            _dispatch_async(async_chat_wrap, async_emb_wrap),
         )
 
     def _uninstrument(self, **kwargs):

openlit/instrumentation/ollama/async_ollama.py

@@ -106,7 +106,8 @@ def async_chat(version, environment, application_name,
         streaming = kwargs.get("stream", False)
 
         server_address, server_port = set_server_address_and_port(instance, "127.0.0.1", 11434)
-        request_model = kwargs.get("model", "gpt-4o")
+        json_body = kwargs.get("json", {}) or {}
+        request_model = json_body.get("model") or kwargs.get("model")
 
         span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
@@ -154,7 +155,8 @@ def async_embeddings(version, environment, application_name,
         """
 
         server_address, server_port = set_server_address_and_port(instance, '127.0.0.1', 11434)
-        request_model = kwargs.get('model', 'all-minilm')
+        json_body = kwargs.get('json', {}) or {}
+        request_model = json_body.get('model') or kwargs.get('model')
 
         span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING} {request_model}'
 

openlit/instrumentation/ollama/ollama.py

@@ -106,7 +106,8 @@ def chat(version, environment, application_name,
         streaming = kwargs.get("stream", False)
 
         server_address, server_port = set_server_address_and_port(instance, "127.0.0.1", 11434)
-        request_model = kwargs.get("model", "gpt-4o")
+        json_body = kwargs.get("json", {}) or {}
+        request_model = json_body.get("model") or kwargs.get("model")
 
         span_name = f"{SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT} {request_model}"
 
@@ -154,7 +155,8 @@ def embeddings(version, environment, application_name,
         """
 
         server_address, server_port = set_server_address_and_port(instance, '127.0.0.1', 11434)
-        request_model = kwargs.get('model', 'all-minilm')
+        json_body = kwargs.get('json', {}) or {}
+        request_model = json_body.get('model') or kwargs.get('model')
 
         span_name = f'{SemanticConvention.GEN_AI_OPERATION_TYPE_EMBEDDING} {request_model}'
 

openlit/instrumentation/ollama/utils.py

@@ -57,8 +57,10 @@ def common_chat_logic(scope, pricing_info, environment, application_name, metric
     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", "gpt-4o")
+    json_body = scope._kwargs.get("json", {}) or {}
+    request_model = json_body.get("model") or scope._kwargs.get("model")
+    messages = json_body.get("messages", scope._kwargs.get("messages", ""))
+    formatted_messages = extract_and_format_input(messages)
 
     cost = get_chat_model_cost(request_model, pricing_info, scope._input_tokens, scope._output_tokens)
 
@@ -252,7 +254,9 @@ def process_embedding_response(response, request_model, pricing_info, server_por
     end_time = time.time()
 
     try:
-        input_tokens = general_tokens(str(kwargs.get('prompt')))
+        json_body = kwargs.get("json", {}) or {}
+        prompt_val = json_body.get('prompt', kwargs.get('prompt', ''))
+        input_tokens = general_tokens(str(prompt_val))
 
         # Calculate cost of the operation
         cost = get_embed_model_cost(request_model,
@@ -293,7 +297,7 @@ def process_embedding_response(response, request_model, pricing_info, server_por
                 SemanticConvention.GEN_AI_SYSTEM: SemanticConvention.GEN_AI_SYSTEM_OLLAMA
             },
             body={
-                **({"content": kwargs.get('prompt', '')} if capture_message_content else {}),
+                **({"content": prompt_val} if capture_message_content else {}),
                 "role":  'user'
             }
         )

openlit/instrumentation/transformers/__init__.py

@@ -1,16 +1,20 @@
-# pylint: disable=useless-return, bad-staticmethod-argument, disable=duplicate-code
-"""Initializer of Auto Instrumentation of HuggingFace Transformer Functions"""
+"""
+Initializer of Auto Instrumentation of HuggingFace Transformer Functions
+"""
+
 from typing import Collection
 import importlib.metadata
 from opentelemetry.instrumentation.instrumentor import BaseInstrumentor
 from wrapt import wrap_function_wrapper
 
-from openlit.instrumentation.transformers.transformers import text_wrap
+from openlit.instrumentation.transformers.transformers import pipeline_wrapper
 
-_instruments = ("transformers >= 4.39.3",)
+_instruments = ("transformers >= 4.48.0",)
 
 class TransformersInstrumentor(BaseInstrumentor):
-    """An instrumentor for HuggingFace Transformer Functions."""
+    """
+    An instrumentor for HuggingFace Transformer library.
+    """
 
     def instrumentation_dependencies(self) -> Collection[str]:
         return _instruments
@@ -28,10 +32,10 @@ class TransformersInstrumentor(BaseInstrumentor):
         wrap_function_wrapper(
             "transformers",  
             "TextGenerationPipeline.__call__",  
-            text_wrap(version, environment, application_name,
+            pipeline_wrapper(version, environment, application_name,
                  tracer, pricing_info, capture_message_content, metrics, disable_metrics),
         )
 
-    @staticmethod
     def _uninstrument(self, **kwargs):
+        # Proper uninstrumentation logic to revert patched methods
         pass

openlit/instrumentation/transformers/transformers.py

@@ -1,63 +1,31 @@
 """
-Module for monitoring ChromaDB.
+Module for monitoring HF Transformers API calls.
 """
 
 import logging
 import time
-from opentelemetry.trace import SpanKind, Status, StatusCode
-from opentelemetry.sdk.resources import SERVICE_NAME, TELEMETRY_SDK_NAME, DEPLOYMENT_ENVIRONMENT
+from opentelemetry.trace import SpanKind
 from openlit.__helpers import (
-    get_chat_model_cost,
-    handle_exception,
-    general_tokens,
-    create_metrics_attributes,
     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 text_wrap(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):
     """
-    Creates a wrapper around a function call to trace and log its execution metrics.
-
-    This function wraps any given function to measure its execution time,
-    log its operation, and trace its execution using OpenTelemetry.
-    
-    Parameters:
-    - version (str): The version of the Langchain application.
-    - environment (str): The deployment environment (e.g., 'production', 'development').
-    - application_name (str): Name of the Langchain application.
-    - tracer (opentelemetry.trace.Tracer): The tracer object used for OpenTelemetry tracing.
-    - pricing_info (dict): Information about the pricing for internal metrics (currently not used).
-    - capture_message_content (bool): Flag indicating whether to trace the content of the response.
-
-    Returns:
-    - function: A higher-order function that takes a function 'wrapped' and returns
-                a new function that wraps 'wrapped' with additional tracing and logging.
+    Generates a telemetry wrapper for GenAI function call
     """
 
     def wrapper(wrapped, instance, args, kwargs):
         """
-        An inner wrapper function that executes the wrapped function, measures execution
-        time, and records trace data using OpenTelemetry.
-
-        Parameters:
-        - wrapped (Callable): The original function that this wrapper will execute.
-        - instance (object): The instance to which the wrapped function belongs. This
-                             is used for instance methods. For static and classmethods,
-                             this may be None.
-        - args (tuple): Positional arguments passed to the wrapped function.
-        - kwargs (dict): Keyword arguments passed to the wrapped function.
-
-        Returns:
-        - The result of the wrapped function call.
-        
-        The wrapper initiates a span with the provided tracer, sets various attributes
-        on the span based on the function's execution and response, and ensures
-        errors are handled and logged appropriately.
+        Wraps the GenAI function call.
         """
 
         server_address, server_port = set_server_address_and_port(instance, "127.0.0.1", 80)
@@ -68,130 +36,26 @@ def text_wrap(version, environment, application_name,
         with tracer.start_as_current_span(span_name, kind= SpanKind.CLIENT) as span:
             start_time = time.time()
             response = wrapped(*args, **kwargs)
-            end_time = time.time()
-
-            # pylint: disable=protected-access
-            forward_params = instance._forward_params
-
-            try:
-                if args and len(args) > 0:
-                    prompt = args[0]
-                else:
-                    prompt = kwargs.get("args", "")
-
-                input_tokens = general_tokens(prompt[0])
-
-                span.set_attribute(TELEMETRY_SDK_NAME, "openlit")
-                span.set_attribute(SemanticConvention.GEN_AI_OPERATION,
-                                   SemanticConvention.GEN_AI_OPERATION_TYPE_CHAT)
-                span.set_attribute(SemanticConvention.GEN_AI_SYSTEM,
-                                   SemanticConvention.GEN_AI_SYSTEM_HUGGING_FACE)
-                span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MODEL,
-                                   request_model)
-                span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TEMPERATURE,
-                                   forward_params.get("temperature", "null"))
-                span.set_attribute(SemanticConvention.GEN_AI_REQUEST_TOP_P,
-                                   forward_params.get("top_p", "null"))
-                span.set_attribute(SemanticConvention.GEN_AI_REQUEST_MAX_TOKENS,
-                                   forward_params.get("max_length", -1))
-                span.set_attribute(SemanticConvention.GEN_AI_USAGE_INPUT_TOKENS,
-                                   input_tokens)
-                span.set_attribute(SemanticConvention.SERVER_ADDRESS,
-                                    server_address)
-                span.set_attribute(SemanticConvention.SERVER_PORT,
-                                    server_port)
-                span.set_attribute(SemanticConvention.GEN_AI_RESPONSE_MODEL,
-                                    request_model)
-
-                span.set_attribute(DEPLOYMENT_ENVIRONMENT,
-                                   environment)
-                span.set_attribute(SERVICE_NAME,
-                                   application_name)
-                span.set_attribute(SemanticConvention.GEN_AI_REQUEST_IS_STREAM,
-                                    False)
-                span.set_attribute(SemanticConvention.GEN_AI_SERVER_TTFT,
-                                    end_time - start_time)
-                span.set_attribute(SemanticConvention.GEN_AI_SDK_VERSION,
-                                    version)
-                if capture_message_content:
-                    span.add_event(
-                        name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
-                        attributes={
-                            SemanticConvention.GEN_AI_CONTENT_PROMPT: prompt,
-                        },
-                    )
-
-                i = 0
-                output_tokens = 0
-                for completion in response:
-                    if len(response) > 1:
-                        attribute_name = f"gen_ai.content.completion.{i}"
-                    else:
-                        attribute_name = SemanticConvention.GEN_AI_CONTENT_COMPLETION_EVENT
-                    if capture_message_content:
-                        # pylint: disable=bare-except
-                        try:
-                            llm_response = completion.get('generated_text', '')
-                        except:
-                            llm_response = completion[i].get('generated_text', '')
-
-                        span.add_event(
-                            name=attribute_name,
-                            attributes={
-                                SemanticConvention.GEN_AI_CONTENT_COMPLETION: llm_response,
-                            },
-                        )
-                    output_tokens += general_tokens(llm_response)
-
-                    i=i+1
-                span.set_attribute(SemanticConvention.GEN_AI_USAGE_OUTPUT_TOKENS,
-                                   output_tokens)
-                span.set_attribute(SemanticConvention.GEN_AI_USAGE_TOTAL_TOKENS,
-                                   input_tokens + output_tokens)
-
-                # Calculate cost of the operation
-                cost = get_chat_model_cost(request_model,
-                                            pricing_info, input_tokens,
-                                            output_tokens)
-                span.set_attribute(SemanticConvention.GEN_AI_USAGE_COST,
-                                    cost)
-
-                span.set_status(Status(StatusCode.OK))
-
-                if disable_metrics is False:
-                    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=server_address,
-                        server_port=server_port,
-                        response_model=request_model,
-                    )
-
-                    metrics["genai_client_usage_tokens"].record(
-                        input_tokens + output_tokens, attributes
-                    )
-                    metrics["genai_client_operation_duration"].record(
-                        end_time - start_time, attributes
-                    )
-                    metrics["genai_server_ttft"].record(
-                        end_time - start_time, attributes
-                    )
-                    metrics["genai_requests"].add(1, attributes)
-                    metrics["genai_completion_tokens"].add(output_tokens, attributes)
-                    metrics["genai_prompt_tokens"].add(input_tokens, attributes)
-                    metrics["genai_cost"].record(cost, attributes)
-
-                # Return original response
-                return response
-
-            except Exception as e:
-                handle_exception(span, e)
-                logger.error("Error in trace creation: %s", e)
 
-                # Return original response
-                return response
+            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,
+            )
+
+        return response
 
     return wrapper

openlit/instrumentation/transformers/utils.py

@@ -0,0 +1,183 @@
+"""
+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
+)
+from openlit.semcov import SemanticConvention
+
+def common_chat_logic(scope, pricing_info, environment, application_name, metrics,
+    capture_message_content, disable_metrics, version, args, kwargs, is_stream):
+
+    """
+    Process chat request and generate Telemetry
+    """
+
+    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)
+
+    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)
+    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
+    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.add_event(
+            name=SemanticConvention.GEN_AI_CONTENT_PROMPT_EVENT,
+            attributes={
+                SemanticConvention.GEN_AI_CONTENT_PROMPT: scope._prompt,
+            },
+        )
+        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))
+
+    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)
+
+def process_chat_response(instance, response, request_model, pricing_info, server_port, server_address,
+                          environment, application_name, metrics, start_time,
+                          span, args, kwargs, capture_message_content=False, disable_metrics=False, version="1.0.0"):
+    """
+    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]
+    else:
+        self._prompt = (
+            kwargs.get("text_inputs") or
+            (kwargs.get("image") and kwargs.get("question") and
+            ("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)
+        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
+            ]
+
+            # Join all non-empty responses into a single string
+            self._llmresponse = " ".join(filter(None, self._llmresponse))
+
+    elif self._kwargs.get("task", "text-generation") == "automatic-speech-recognition":
+        self._llmresponse = response_dict.get("text", "")
+
+    elif self._kwargs.get("task", "text-generation") == "image-classification":
+        self._llmresponse = str(response_dict[0])
+
+    elif self._kwargs.get("task", "text-generation") == "visual-question-answering":
+        self._llmresponse = str(response_dict[0]).get("answer")
+
+    common_chat_logic(self, pricing_info, environment, application_name, metrics,
+            capture_message_content, disable_metrics, version, args, kwargs, is_stream=False)
+
+    return response

{openlit-1.33.20.dist-info → openlit-1.33.21.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: openlit
-Version: 1.33.20
+Version: 1.33.21
 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-1.33.20.dist-info → openlit-1.33.21.dist-info}/RECORD

@@ -1,4 +1,4 @@
-openlit/__helpers.py,sha256=mbcQvTwjf0R3POo2vDPc0Ms94pNCmAAu9OWXeenFHC4,9068
+openlit/__helpers.py,sha256=sg0EGJGC_OlZePR84cLK77l_lZRBPJwdjWjq_RuaYS0,11444
 openlit/__init__.py,sha256=1OzJQmiZrTlT3Aze_l8GOf1GXH7dAHztJn0Uzd1LAPc,23924
 openlit/evals/__init__.py,sha256=nJe99nuLo1b5rf7pt9U9BCdSDedzbVi2Fj96cgl7msM,380
 openlit/evals/all.py,sha256=oWrue3PotE-rB5WePG3MRYSA-ro6WivkclSHjYlAqGs,7154
@@ -90,10 +90,10 @@ openlit/instrumentation/mistral/mistral.py,sha256=_2qM8v4RCL-S0Mm1vbW77m5vUm8aPD
 openlit/instrumentation/multion/__init__.py,sha256=Wr3lcDyG_YbOLkCUzBFhraAedF6E113tce8eSWlcz10,3149
 openlit/instrumentation/multion/async_multion.py,sha256=XutZnayCJOZ_NA9bvE1NUoej41KOGR7FRn2tpoGKMEU,6092
 openlit/instrumentation/multion/multion.py,sha256=-WqRAcu5qiEMY9XDmlJTQHuQiWfdwms9JDn127QCNb8,6074
-openlit/instrumentation/ollama/__init__.py,sha256=JjxSqEegmRoRqIVz7ZAq9dLyXPZ2DqV2wGmgXCENNpw,3004
-openlit/instrumentation/ollama/async_ollama.py,sha256=rQ637cpOenezcbyJPV16LLQ2UKuROrO31-0lbfyGhoA,6541
-openlit/instrumentation/ollama/ollama.py,sha256=9C-XtUZ9FoR1cjpFdO3BHjF6NZqRhzNLt-z2z0cJIyE,6437
-openlit/instrumentation/ollama/utils.py,sha256=vAFCWv4qWd-_jThR0IQGoBhjp99uy1QUtJk6Kj6y_Js,14604
+openlit/instrumentation/ollama/__init__.py,sha256=v7VhVxHw_c6QtMznxe6a7z6QrYHZsH_NSXfiXao83Ns,3707
+openlit/instrumentation/ollama/async_ollama.py,sha256=zJPDr2ROh1nvFGoxgdTbe04Zr1KhmgJUYFPeuRLQGLk,6667
+openlit/instrumentation/ollama/ollama.py,sha256=MNUltiP9XVT4azmO_-E2vjhFaoHQyJ0Z6c-HnB0_jCE,6563
+openlit/instrumentation/ollama/utils.py,sha256=41uvYaYkGwWfRyHYqhOOwrFy6cMzBlG1urJYUat9Q24,14819
 openlit/instrumentation/openai/__init__.py,sha256=FiL4OHDhs957spa3k9sNC_VLt0-txtwbnujQwnevQ5I,5564
 openlit/instrumentation/openai/async_openai.py,sha256=CiyBpn8Evnd_gh3Cm1WbfkN7eUpDmFh4KMvxka-B4og,71764
 openlit/instrumentation/openai/openai.py,sha256=r8ZNVoAFTPuCUf18a5v1Lp48LXwCeT9paEB-3USSiiU,71507
@@ -114,8 +114,9 @@ openlit/instrumentation/reka/reka.py,sha256=L6gH7j94tcYlc_FCkQP6SrxH7yBr4uSgtN8B
 openlit/instrumentation/together/__init__.py,sha256=MLLL2t8FyrytpfMueqcwekiqTKn-JN40HBD_LbZS_jQ,2661
 openlit/instrumentation/together/async_together.py,sha256=ToSeYqE0mCgSsCNSO0pqoyS7WU6YarHxa3I7ZrzH-d8,30634
 openlit/instrumentation/together/together.py,sha256=7Da9fjHaZk_ObXMnSZA79-RktgwHRVYevsZAA-OpcXY,30530
-openlit/instrumentation/transformers/__init__.py,sha256=f-kWX6_VdiOXqDa64iK6oJyYF3xkHuKxIBPTll-W7Lw,1467
-openlit/instrumentation/transformers/transformers.py,sha256=naSIvynfI16RWboB4cZb728AqubU8lUt-zYSRUg6X_k,9174
+openlit/instrumentation/transformers/__init__.py,sha256=9Ubss5nlumcypxprxff8Fv3sst7II27SsvCzqkBX9Kg,1457
+openlit/instrumentation/transformers/transformers.py,sha256=zCAwfXu77HPlhy7vuU-nvNvsmmU4fs4aaFMCBG5AOLA,1993
+openlit/instrumentation/transformers/utils.py,sha256=UP-aB_hP4SVLQ1A0I-PrNXC3mPJkVZnS7UOkQGe6OXc,8087
 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
@@ -125,7 +126,7 @@ openlit/otel/events.py,sha256=VrMjTpvnLtYRBHCiFwJojTQqqNpRCxoD4yJYeQrtPsk,3560
 openlit/otel/metrics.py,sha256=urpadRfC_BjLCPxWgk5J6NGStECeJA55LFkyTD43Jd4,6837
 openlit/otel/tracing.py,sha256=tjV2bEbEDPUB1Z46gE-UsJsb04sRdFrfbhIDkxViZc0,3103
 openlit/semcov/__init__.py,sha256=JF9MwflazC8jHOiQdPYshfv1q5Z9bhB4OGa0N_fr9d4,13305
-openlit-1.33.20.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
-openlit-1.33.20.dist-info/METADATA,sha256=GZtU_l6KXa6MR28a4KEa0TzHsNh5d0mwmPnZ2dNFMfQ,23470
-openlit-1.33.20.dist-info/WHEEL,sha256=fGIA9gx4Qxk2KDKeNJCbOEwSrmLtjWCwzBz351GyrPQ,88
-openlit-1.33.20.dist-info/RECORD,,
+openlit-1.33.21.dist-info/LICENSE,sha256=xx0jnfkXJvxRnG63LTGOxlggYnIysveWIZ6H3PNdCrQ,11357
+openlit-1.33.21.dist-info/METADATA,sha256=RP9M2CRa9gXnTVN6I_YjyjoGS8C3zUI04w4lbI6q3yE,23470
+openlit-1.33.21.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+openlit-1.33.21.dist-info/RECORD,,

{openlit-1.33.20.dist-info → openlit-1.33.21.dist-info}/WHEEL

@@ -1,4 +1,4 @@
 Wheel-Version: 1.0
-Generator: poetry-core 2.1.2
+Generator: poetry-core 2.1.3
 Root-Is-Purelib: true
 Tag: py3-none-any